WO2022143562A1

WO2022143562A1 - Cargo container state detection method and apparatus

Info

Publication number: WO2022143562A1
Application number: PCT/CN2021/141762
Authority: WO
Inventors: 戴海能; 周维; 王进
Original assignee: 虹软科技股份有限公司
Priority date: 2020-12-28
Filing date: 2021-12-27
Publication date: 2022-07-07
Also published as: CN114693588A; DE112021006736T5

Abstract

A cargo container state detection method and apparatus. Said method comprises: acquiring a cargo container image (S102); processing the cargo container image using a first network model, so as to obtain a cargo container state confidence (S104); and on the basis of the cargo container state confidence, determining a cargo container state (S106). Said method solves the technical problem in the related art that a cargo container state detection method cannot detect whether a cargo container is loaded with cargo.

Description

Method and device for detecting state of cargo box

This application claims the priority of a Chinese patent application with a priority number of 202011589677.5 and an invention titled "Method and Device for Detecting the Condition of a Cargo Box" filed with the China Patent Office on December 28, 2020, the entire contents of which are incorporated herein by reference Applying.

technical field

The present disclosure relates to the field of cargo box state detection, and in particular, to a method and device for detecting the state of a cargo box.

Background technique

At present, the muck trucks will be overloaded and the muck will be scattered during the transportation process. If they cannot be found in time, it will not only affect the cleanliness of the road, but also cause many hidden safety hazards.

In order to solve the above problems, in the related art, the first time difference and the second time difference are determined respectively by setting the first radar and the second radar, and the microcontroller determines the first side cover and the second time difference according to the first time difference and the second time difference. The airtightness of the side roof can identify whether the roof of the muck is effectively closed, and reduce the environmental problems caused by the ineffective closure of the roof of the muck. In this method, it can be identified whether the roof cover of the muck truck is effectively closed, but it cannot be identified whether the cargo box of the muck truck is loaded with goods; and the use of laser radar in this method is costly.

For the above problems, no effective solution has been proposed yet.

SUMMARY OF THE INVENTION

The embodiments of the present disclosure provide a method and device for detecting the state of a cargo box, so as to at least solve the technical problem that the method for detecting the state of a cargo box in the related art cannot detect whether goods are loaded in the cargo box.

According to an aspect of the embodiments of the present disclosure, a method for detecting the state of a cargo box is provided, including: acquiring an image of the cargo box; processing the image of the cargo box by using a first network model to obtain a confidence level of the state of the cargo box; Confidence, determine the status of the cargo box, where the status of the cargo box includes one of the following: the top cover of the cargo box is closed, no cargo is loaded in the cargo box, the amount of cargo loaded in the cargo box meets the preset amount, the camera is abnormal, the cargo box is abnormal The amount of cargo loaded does not meet the preset amount.

Optionally, using the first network model to process the image of the cargo box to obtain the confidence level of the state of the cargo box includes: trimming the image of the cargo box to obtain the image of the target area; preprocessing the image of the target area to obtain the processed image ; Input the processed image into the first network model to obtain the confidence level of the state of the container.

Optionally, the confidence level of the state of the cargo box includes: a first confidence level for representing the closure of the top cover of the cargo box, a second confidence level for representing no cargo in the cargo box, and a cargo for representing the cargo loaded in the cargo box. The third confidence level is used to represent the abnormality of the camera, and the fifth confidence level is used to represent that the amount of goods loaded in the cargo box does not meet the preset amount.

Optionally, using the first network model to process the image of the container to obtain the confidence of the state of the container, further comprising: extracting the confidence of the state of the multi-frame container corresponding to the images of the multi-frame container; Enter the second network model to obtain the confidence level of the container state.

Optionally, determining the state of the container based on the confidence level of the container state includes: extracting the container corresponding to the highest value among the first confidence level, the second confidence level, the third confidence level, the fourth confidence level, and the fifth confidence level The state is the state of the container; or, extracting the state of the container whose first confidence level, the second confidence level, the third confidence level, the fourth confidence level and the fifth confidence level are greater than the preset threshold is the container state; or, extracting the first The highest value among the first confidence level, the second confidence level, the third confidence level, the fourth confidence level, and the fifth confidence level, and the container state corresponding to the one greater than the preset threshold value is the container state.

Optionally, determining the status of the container based on the confidence of the status of the container includes: determining the status of the container based on the confidence of the status of the container and a preset rule, wherein the preset rule is used to represent the priority of the confidence of the status of the container .

Optionally, determining the state of the container based on the confidence of the state of the container includes: judging whether the fourth confidence is greater than a fifth preset value; if the fourth confidence is greater than the fifth preset, determining that the state of the container is a camera Abnormal; if the fourth confidence level is less than or equal to the fifth preset value, it is determined whether the first confidence level is greater than the second preset value; if the first confidence level is greater than the second preset value, it is determined that the state of the container is the top of the container The cover is closed; if the first confidence level is less than or equal to the second preset value, it is determined whether the second confidence level is greater than the third preset value; if the second confidence level is greater than the third preset value, it is determined that the state of the cargo box is a cargo box If the second confidence level is less than or equal to the third preset value, it is determined whether the third confidence level is greater than the fourth preset value; if the third confidence level is greater than the fourth preset value, it is determined that the state of the cargo box is The amount of goods loaded in the container meets the preset amount; if the third confidence level is less than or equal to the fourth preset value, it is determined whether the fifth confidence level is greater than the sixth preset value; if the fifth confidence level is greater than the sixth preset value value, then it is determined that the status of the container is that the quantity of goods loaded in the container does not meet the preset quantity.

Optionally, the preprocessing includes at least one of the following: scaling processing and normalization processing.

Optionally, the method further includes: acquiring an image of an original container; processing the image of the original container to obtain multiple sets of training samples; and using multiple sets of training samples to train an initial model to obtain a first network model.

Optionally, processing the original cargo box images to obtain multiple sets of training samples, including: performing perturbation processing on the original cargo box images to obtain multiple sets of original cargo box images; trimming multiple sets of original cargo box images to obtain multiple sets of original cargo box images. group target area images; preprocess multiple groups of target area images to obtain multiple groups of training samples.

Optionally, the perturbation processing includes at least one of the following: rotation, translation, scaling, noise addition, blurring, illumination change and channel change.

Optionally, acquiring the image of the cargo box includes: detecting whether the current light intensity is greater than a first preset value; if the current light intensity is greater than the first preset value, using the first camera to acquire the image of the cargo box; if the current light intensity is less than or equal to If the first preset value is used, the image of the cargo box is obtained by using the second camera.

Optionally, before using multiple sets of training samples to train the initial model to obtain the first network model, the method further includes: adding a normalization layer to the initial layer of the initial model, wherein the normalization layer is used to Group training samples are normalized.

According to another aspect of the embodiments of the present disclosure, there is also provided an apparatus for detecting the state of a cargo box, including: an acquiring component, configured to acquire an image of the cargo box; and a processing component, configured to process the image of the cargo box by using a first network model , obtain the confidence of the state of the cargo box; the determining component is configured to determine the state of the cargo box based on the confidence of the state of the cargo box, wherein the state of the cargo box includes one of the following: the top cover of the cargo box is closed, the cargo is not loaded in the cargo box, the cargo box The amount of goods loaded in the container meets the preset amount, the camera is abnormal, and the amount of goods loaded in the container does not meet the preset amount.

According to another aspect of the embodiments of the present disclosure, a computer-readable storage medium is also provided, where the computer-readable storage medium includes a stored program, wherein when the program is executed, a device where the computer-readable storage medium is located is controlled to execute the above-mentioned cargo box Status detection method.

According to another aspect of the embodiments of the present disclosure, a processor is also provided, and the processor is configured to run a program, wherein the above-mentioned method for detecting the state of a cargo box is executed when the program runs.

Description of drawings

The accompanying drawings described herein are used to provide a further understanding of the present disclosure and constitute a part of the present application. The exemplary embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation of the present disclosure. In the attached image:

FIG. 1 is a flowchart of a method for detecting the state of a cargo box according to an embodiment of the present disclosure;

2 is a schematic diagram of a state of a cargo box with a top cover of the cargo box closed according to an embodiment of the present disclosure;

FIG. 3 is a schematic diagram of the state of a cargo box with no goods loaded in the cargo box according to an embodiment of the present disclosure;

FIG. 4 is a schematic diagram of the state of a cargo box in which the amount of goods loaded in a cargo box meets a preset amount according to an embodiment of the present disclosure;

FIG. 5 is a schematic diagram of the state of a cargo box with an abnormal camera according to an embodiment of the present disclosure;

6 is a schematic diagram of an optional method for detecting the state of a cargo box according to an embodiment of the present disclosure;

FIG. 7 is a schematic diagram of another optional method for detecting the state of a cargo box according to an embodiment of the present disclosure;

FIG. 8 is a schematic diagram of an apparatus for detecting the state of a cargo box according to an embodiment of the present disclosure.

Detailed ways

In order to make those skilled in the art better understand the solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only Embodiments are part of the present disclosure, but not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present disclosure.

It should be noted that the terms "first", "second" and the like in the description and claims of the present disclosure and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific sequence or sequence. It is to be understood that the data so used may be interchanged under appropriate circumstances such that the embodiments of the disclosure described herein can be practiced in sequences other than those illustrated or described herein. Furthermore, the terms "comprising" and "having" and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those expressly listed Rather, those steps or units may include other steps or units not expressly listed or inherent to these processes, methods, products or devices.

Example 1

According to an embodiment of the present disclosure, an embodiment of a method for detecting the state of a cargo box is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings may be executed in a computer system such as a set of computer-executable instructions, and , although a logical order is shown in the flowcharts, in some cases steps shown or described may be performed in an order different from that herein.

FIG. 1 is a method for detecting the state of a cargo box according to an embodiment of the present disclosure. As shown in FIG. 1 , the method includes the following steps:

Step S102, acquiring the image of the cargo box;

Step S104, using the first network model to process the image of the container to obtain the confidence level of the state of the container;

Step S106: Determine the status of the cargo box based on the confidence level of the status of the cargo box, wherein the status of the cargo box includes one of the following: the top cover of the cargo box is closed, the cargo box is not loaded with goods, and the amount of goods loaded in the cargo box meets a preset amount , The camera is abnormal, and the amount of goods loaded in the cargo box does not meet the preset amount.

Through the above embodiments of the present disclosure, after the image of the cargo box is obtained, the first network model is used to process the image of the cargo box to obtain the confidence level of the status of the cargo box, and then the status of the cargo box is determined based on the confidence level of the status of the cargo box, thereby realizing the detection When the cargo box is in the state, it can not only detect whether the top cover of the shipping box is effectively closed, but also detect whether the cargo box is loaded with goods, and detect the capacity of the goods loaded in the cargo box, and the obtained cargo box image does not show the status of the cargo box. When the camera is abnormal, it can detect the status of the container effectively. In the present disclosure, only a camera is used to realize detection. Compared with the solution of using radar to realize detection in the prior art, the cost is reduced, and the method of detecting the state of the cargo box in the related art can not detect whether the cargo is loaded in the cargo box. question.

The following describes in detail with reference to the above implementation steps.

Step S102, acquiring the image of the cargo box.

The image of the cargo box in the above steps may be an image of a cargo box of a muck truck, an image of a cargo box of a truck, or an image of a cargo box placed in a warehouse, which is not limited herein. It should be noted that the image of the cargo box may be an image of the cargo box obtained by a common camera, or may be an image of an infrared cargo box obtained by an infrared camera. In an environment with insufficient light such as at night, the images obtained by ordinary cameras are dim and blurred, and the quality is low, which will affect the accuracy of subsequent detection of the state of the container. Therefore, infrared cameras can be used to collect images in an environment with insufficient light, and high-quality images can be collected in the early stage. The image ensures that the subsequent detection of the condition of the cargo box is realized with high precision.

In the above steps, the acquired image of the cargo box may be the current image of the cargo box, or may be the image of the historically stored cargo box.

The image of the cargo box in the above steps can be acquired by a camera mounted on the cargo box, or acquired by a camera mounted on the vehicle body. It should be noted that the camera can acquire an image of the container at a fixed position, and the image of the container can display the state of the top cover of the shipping container and the state of the goods loaded in the container.

In an optional embodiment, an image of the current cargo box of the muck truck may be acquired through a camera installed on the muck truck.

Step S104 , using the first network model to process the image of the container to obtain the confidence level of the state of the container.

The first network model in the above steps can be a CNN (Convolution neural network, convolutional neural network), and the first network model can simulate the visual cortex to decompose and analyze image data.

It should be noted that the first network model of the present application can directly obtain the confidence levels of all the container states, that is, the present application only needs one first network model to obtain the confidence levels of the states of multiple containers according to the container image, thereby increasing the confidence level. to determine the speed at which the status of the container is determined.

In the above steps, the status of the cargo box may be that the top cover of the cargo box is closed, no goods are loaded in the cargo box, the amount of goods loaded in the cargo box meets the preset amount, the camera is abnormal, and the amount of goods loaded in the cargo box does not meet the preset amount. . Wherein, the unloaded goods in the container may refer to the empty container, and the amount of goods loaded in the container that meets the preset amount may mean that the container is fully loaded, the container is overloaded, etc. value to be determined. In the present application, the output of the confidence level of the first network model includes not only the opening and closing status of the top cover of the cargo box, but also the detection of the status of the amount of goods in the cargo box and the status of the camera.

Confidence, also known as reliability and confidence level, can be the probability that the estimated value and the overall parameter are within a certain allowable error range. The confidence of the state of the container in the above steps may refer to the probability of each state of the container obtained by analyzing the image of the container through the first network model.

The target area in the above steps may refer to the ROI (Region of Interest) in the image of the cargo box. Considering that the camera shooting scene is unchanged, the identification of the status of the cargo box can be based on the preset interest. area carried out. For example, it may refer to the area where the top cover of the cargo box is located in the cargo box image, which may be determined according to actual detection needs.

Specifically, the ROI can be used to trim the image to obtain the target area image; the ROI can be used to specify the target area in the read-in image, so that only the target area image where the container is located can be classified and identified, which can reduce the detection processing time and increase the The detection accuracy can bring convenience to image processing. Among them, you can use Range (range function) to specify the range of rows and columns of interest. Range is a continuous sequence from the start index to the end index; you can also use the rectangle Rect (rectangle function) to frame, specify the coordinates of the upper left corner of the rectangle Rect and the length and width of the rectangle.

Optionally, the above-mentioned preprocessing includes at least one of the following: scaling processing and normalization processing.

The scaling process in the above steps refers to the process of adjusting the size of the digital image. Image scaling requires a trade-off between processing efficiency and the smoothness and clarity of the result. As the size of an image increases, the pixels that make up the image become more visible; conversely, shrinking an image will enhance its smoothness and clarity. The selection of the fixed size of the image is based on the actual requirements of the recognition speed and recognition accuracy of the first network model, and the same standard is used in the recognition process for different collected images, so as to ensure that the information output by the network is more accurate. Exemplarily, the target area image can be scaled to a fixed size: 128*128.

The normalization processing in the above steps refers to limiting the data to a certain range. The normalization process does not change the image information, but it can eliminate the initial invalid effect of the network caused by the singular image, and can prevent the gradient explosion after the image enters the first network model. The normalization method can be maximum-minimum normalization, Z-score normalization, function transformation, etc. Among them, maximum-minimum normalization refers to the linear change of the original data; Z-score normalization is based on the mean and standard deviation of the original data. Data normalization performed.

Specifically, the image of the cargo box can be cropped through the ROI to obtain the target area, and then the target area is scaled to a fixed size: 128*128, and then the scaled target area is normalized, specifically, the pixels of the target area are The value is subtracted from the mean value of 128, and then divided by the variance of 256. Finally, the processed image is input into the convolutional neural network to obtain the confidence level of the state of the container.

Optionally, the above-mentioned confidence level of the state of the container includes: a first confidence level for characterizing the closure of the top cover of the container, a second confidence level for characterizing the unloaded goods in the container, and a second confidence level for characterizing the loaded goods in the container. The third confidence level is used to characterize the abnormality of the camera, and the fifth confidence level is used to indicate that the quantity of goods loaded in the cargo box does not meet the preset amount.

In an optional embodiment, a convolutional neural network can be used to process the current cargo box image, and the output result of the first network model is that the confidence level of the top cover of the cargo box being closed is 80%, and the cargo box is not loaded with goods. The confidence level is 5%, the confidence level of the amount of goods loaded in the container meets the preset amount is 5%, the confidence level of the camera is abnormal is 5%, the confidence level of the cargo volume loaded in the container does not meet the preset amount is 5%.

In step S106, the state of the container is determined based on the confidence level of the state of the container.

The status of the cargo box includes one of the following: the top cover of the cargo box is closed, no cargo is loaded in the cargo box, the amount of goods loaded in the cargo box meets the preset amount, the camera is abnormal, and the amount of goods loaded in the cargo box does not meet the predetermined amount. Set the amount.

Fig. 2 shows a schematic diagram of the state of the cargo box with the top cover of the cargo box closed; Fig. 3 shows a schematic diagram of the state of the cargo box with no goods loaded in the cargo box; Fig. 4 shows that the amount of goods loaded in the cargo box meets the preset amount Schematic diagram of the status of the cargo box; Figure 5 shows the schematic diagram of the status of the cargo box with an abnormal camera.

The closing of the top cover of the cargo box means that the cargo box can be normally closed after the loading of goods does not meet the preset amount or the cargo box is not loaded with goods; When the preset amount is met, the top cover of the cargo box is normally opened, or the amount of goods loaded in the cargo box is too large, so that the top cover of the cargo box cannot be closed normally; the abnormality of the camera may be that the camera is blocked or the camera is deflected, resulting in the inability to effectively detect the cargo box. state within.

In an optional embodiment, the highest confidence level among the confidence levels of the states of each cargo box output by the first network model may be selected, and the state of the cargo box may be determined to be the container state corresponding to the highest confidence level, so as to ensure the state of the cargo box. accuracy.

In another optional embodiment, a preset threshold may be set, and when the confidence level is greater than the preset threshold, it is determined that the container state is the container state corresponding to the confidence level, so as to ensure the accuracy of the obtained container state . When the first confidence level is greater than the second preset value, the state of the cargo box is determined to be that the top cover of the cargo box is closed; when the second confidence level is greater than the third preset value, the state of the cargo box is determined to be that no goods are loaded in the cargo box; When the third confidence level is greater than the fourth preset value, it is determined that the state of the cargo box is that the amount of goods loaded in the cargo box meets the preset amount; when the fourth confidence level is greater than the fifth preset value, it is determined that the state of the cargo box is the camera abnormality; When the fifth confidence level is greater than the sixth preset value, it is determined that the state of the cargo box is that the amount of goods loaded in the cargo box does not meet the preset amount.

In yet another optional embodiment, the state of the container may be determined according to the highest confidence level and a preset threshold, and when the confidence level satisfies these two conditions, the state of the container is determined as the state of the container corresponding to the confidence level , to further improve the accuracy of the status of the cargo box.

Specifically, the judgment may be made according to the judgment priorities of the first confidence level, the second confidence level, the third confidence level, the fourth confidence level, and the fifth confidence level, and then output the status of the container according to the judgment result. The judgment priority may be set by the user, and the judgment priority may also be randomly generated. The priority setting can be based on the degree of danger corresponding to the status of the container. For example, if the camera is blocked or faulty, it will not be possible to determine the status of the top cover and the amount of goods. If the top cover cannot be closed normally, it will bring road environment and safety hazards.

In the above steps, the order of the priority of the confidence level judgment from high to low is: the fourth confidence level, the first confidence level, the second confidence level, the third confidence level, and the fifth confidence level.

In an optional embodiment, it may be first determined whether the fourth confidence level is greater than the fifth preset value. If the fourth confidence level is greater than the fifth preset value, it means that the camera of the muck truck is abnormal. The data of other confidence levels are inaccurate. Therefore, it can be directly determined that the status of the cargo box is abnormal with the camera, which improves the accuracy of determining the status of the cargo box.

Exemplarily, it may be determined that the first confidence level is greater than the second preset value, the second confidence level is less than the third preset value, the third confidence level is less than the fourth preset value, and the fourth confidence level is less than the fifth preset value. After the value, the fifth confidence level is less than the sixth preset value, and the output container state is that the top cover of the container represented by the first confidence level is closed.

The preset amount in the above steps can be set by the user according to the capacity of the cargo box. The second preset value, the third preset value, the fourth preset value, the fifth preset value, and the sixth preset value may be the same or different, which are not limited herein.

Through the above embodiments of the present disclosure, after the image of the cargo box is obtained, the first network model is used to process the image of the cargo box to obtain the confidence level of the status of the cargo box, and then the status of the cargo box is determined based on the confidence level of the status of the cargo box, thereby realizing the detection When the cargo box is in the state, it can not only detect whether the top cover of the shipping box is effectively closed, but also detect whether the cargo box is loaded with goods, and detect the capacity of the loaded goods in the cargo box. Only using the camera to obtain the image of the cargo box can reduce the detection cost. . For example: after obtaining the image of the cargo box, the first network model can be used to analyze the image of the cargo box to obtain the confidence level of the closed top cover of the cargo box, the confidence level of the unloaded goods in the cargo box, and the quantity of goods loaded in the cargo box. The confidence level of meeting the preset amount, the confidence level of the camera being abnormal, and the confidence level that the amount of goods loaded in the container does not meet the preset amount can be determined according to the confidence level of various states of the container. The state of the shipping box can not only detect Whether the top cover of the cargo box is effectively closed, it can also detect whether the cargo box is loaded with goods, and when the obtained image of the cargo box does not show the status of the cargo box, the abnormality of the camera can be detected, so as to effectively detect the status of the cargo box. At the same time, the present disclosure Compared with the conventional method of using radar to realize detection, the cost is reduced, the performance is stable and the scope of application is wide, which solves the problem that the method of detecting the state of the cargo box in the related art cannot detect the inside of the cargo box. Whether or not to load the goods technically. Optionally, using the first network model to process the image of the container to obtain the confidence of the state of the container, further comprising: extracting the confidence of the state of the multi-frame container corresponding to the images of the multi-frame container; Enter the second network model to obtain the confidence level of the container state.

Further, in order to avoid unstable detection results, false detections and missed detections in the confidence results obtained based on single-frame image detection, the present disclosure determines the status of the cargo box based on the confidence results obtained from multi-frame image detection. The selection of the number of frames for multi-frame processing is a trade-off between processing efficiency and result quality, which can improve the accuracy of container status detection while maintaining the stability of the detection results. Specifically, the state of the container is continuously detected for 30s, and the multi-frame images within this time period are extracted and entered into the first network model to obtain the corresponding multi-frame container state confidence, and the obtained multi-frame container state confidence is input into the second The network model is trained to obtain the confidence level of the final cargo box state, and the above-mentioned second network model may be a support vector machine.

The original cargo box image in the above steps can be acquired through a camera, can also be acquired through a network, or can be acquired from a file stored locally, which is not limited here.

The training samples in the above steps are also the parameters of the initial model. After training, it can be considered that the model system has been established. The initial model may be an untrained model or a model that has been trained at least once before.

It should be noted that the first network model has taken into account the factors of ordinary images and infrared images in the training process. Therefore, the first network model is used to perform recognition based on different types of images to improve the recognition accuracy.

Exemplarily, when the first network model is trained based on ordinary images and infrared images, respectively, when it is detected that the image of the cargo box is an ordinary image, the first network model can identify the image of the cargo box based on the ordinary image. ; When the detected image of the cargo box is an infrared image, the first network model can recognize the infrared image based on the training demerits of the infrared image. By training two different types of cargo box images, the accuracy of the image recognition of the cargo box can be improved. As well as extending the compatibility of the first network model.

The first network model in the above steps may be ResNet (Residual Neural Network, residual network).

Optionally, processing the original cargo box images to obtain multiple sets of training samples, including: performing perturbation processing on the original cargo box images to obtain multiple sets of original cargo box images, wherein the perturbation processing is used to expand the original cargo box images and enhancement; trimming multiple sets of original cargo box images to obtain multiple sets of target area images; preprocessing multiple sets of target area images to obtain multiple sets of training samples.

The real scene environment information is complex and changeable. In contrast, the amount of training sample data collected is limited. In order to still adapt to the changing environmental scene in actual detection, the present disclosure expands and enhances the original image samples, and enriches the training samples to make them suitable for complex environments. Stronger anti-interference ability.

In an optional embodiment, rotation, translation, scaling, noise addition, blurring, illumination change, and channel change may be performed on the original cargo box image to obtain multiple sets of original cargo container images, and then multiple sets of original cargo container images can be processed through ROI. The box images are cropped to obtain multiple sets of target area images; after the multiple sets of target area images are scaled to a fixed size, the scaled multiple sets of target area images are normalized to obtain multiple sets of training samples.

In another optional embodiment, after preprocessing multiple groups of target area images, perturbation processing may be performed again to obtain more abundant training samples, so that the anti-interference ability of complex environments is stronger.

Optionally, the perturbation processing includes at least one of the following: rotation, translation, scaling, noise addition, blurring, illumination change, channel change.

In an optional embodiment, 0 to 15 left and right rotation operations may be randomly selected for the original cargo box image.

In another optional embodiment, a translation operation may be performed on the original cargo box image by moving up, down, left, and right by 0 to 50 pixels.

In another optional embodiment, an image scaling operation of 0.8 to 1.2 may be randomly performed on the original cargo box image.

In another optional embodiment, Gaussian or salt and pepper noise with a mean value of 0 to 4.0 may be randomly added to the original cargo box image.

In another optional embodiment, a Gaussian blur with a template of 0 to 9 may be randomly added to the original cargo box image.

In another alternative embodiment, the original container image may be randomly multiplied by a lighting transformation of 0.8 to 1.2 per pixel.

In another optional embodiment, channels of the original cargo box image may be converted, for example, RGB channels are converted to BRG channels.

The first preset value in the above steps can be set by the user based on requirements.

In the above steps, the first camera can be an RGB camera, and the second camera can be an IR camera; the RGB camera can acquire images in a light state; the IR camera is an infrared camera, and can acquire images in a dark state.

In an optional embodiment, the sensor can detect whether the current light intensity is greater than the first preset value; if the current light intensity is greater than the first preset value, it means that the current cargo box is in a lighting state, and an RGB camera can be used Obtain the image of the cargo box; if the current light intensity is less than or equal to the first preset value, it means that the current cargo box is in a dark state, and the IR camera can be used to obtain the image of the cargo box.

Since both infrared images and ordinary images are used in the training of the first network model, in order to avoid the difference in imaging between infrared images and ordinary images, a batch normalization layer is added to the initial layer of the first network model to normalize the data. At the same time, the training of the batch normalization layer will gradually eliminate the influence caused by the two image qualities, so that the two images are mapped to the same sample space, which can improve the recognition accuracy of the first network model of the state of the container and the recognition speed.

The trained first network model can output the confidence level of the state of the container, including: a first confidence level for representing the closure of the top cover of the container, a second confidence level for representing the unloaded goods in the container, and a second confidence level for representing the container The third confidence level for the amount of goods loaded in the container meets the preset amount, the fourth confidence level for indicating an abnormality of the camera, and the fifth confidence level for indicating that the amount of goods loaded in the container does not meet the preset amount.

A preferred embodiment of the present disclosure will be described in detail below with reference to FIG. 6 . As shown in Figure 6, the method may include the following steps:

Step S601, acquiring the image of the muck truck cargo box;

Optionally, a camera may be selected according to the light intensity to obtain images of the cargo box of the muck truck. If the current light intensity is greater than the first preset value, it means that the current cargo box is in a light state, and the RGB camera can be used to obtain the image of the cargo box; if the current light intensity is less than or equal to the first preset value, it means that the current cargo box is in a dark state Next, the image of the cargo box can be obtained by using the IR camera.

Step S602, crop the image of the cargo box to obtain the image of the target area;

Optionally, the cargo box image can be cropped by ROI.

Step S603, preprocessing the target area image;

Optionally, the preprocessing can be scaling processing and normalization processing; the target area image can be scaled to a fixed size: 128*128; the pixels of the target area image can be subtracted from the mean value of 128, and then divided by the variance of 256.

Step S604, sending the preprocessed image of the target area into the CNN network to obtain the confidence level of the state of the container.

Optionally, the confidence level of the state of the container includes one of the following: confidence level of the top cover of the container being closed, confidence level of no goods loaded in the container, confidence level of the amount of goods loaded in the container meeting a preset value, confidence level of camera abnormality, and confidence level of the cargo box. The amount of goods loaded in the box does not meet the preset confidence level.

It should be noted that if the top cover of the cargo box is closed, it means that the cover of the muck truck is completely closed; no cargo is loaded in the cargo box, the amount of goods loaded in the cargo box meets the preset value, and the amount of cargo loaded in the cargo box does not meet the preset amount, indicating that the slag truck is completely closed. The top cover of the dirt truck is open; if the camera is abnormal, it means that the camera is blocked or the camera is deflected, resulting in effective detection.

Step S605, based on the confidence of the state of the container and the artificially designed logic rule, determine the state of the container of the muck truck.

Optionally, the state of the container of the muck truck may be determined based on the confidence level of the state of the container and the signal of the cover plate of the container of the muck truck.

Exemplarily, within 30s of sending the signal from the cargo box cover of the muck truck, the confidence level of the state of the container of the multi-frame corresponding to the multi-frame images collected within 30s can be input to the support vector machine and output the confidence level of the final state of the container. According to the above The confidence level of the cargo box status determines whether the cargo box status of the muck truck is not loaded with cargo in the cargo box or whether the amount of cargo loaded in the cargo box meets the preset value.

Optionally, the state of the container of the muck truck may be determined based on the confidence level of the state of the container and the transport state of the muck truck.

A preferred embodiment of the first network model training of the present disclosure will be described in detail below with reference to FIG. 7 . As shown in Figure 7, the method may include the following steps:

Step S701, obtaining an image of the original cargo box of the muck truck;

The original cargo box image of the muck truck in the above steps can be obtained through a camera, can also be obtained through a network, and can also be obtained from a file stored locally, which is not limited here.

Optionally, if the original cargo box image of the muck truck is obtained through a camera, a camera can be selected according to the light intensity to obtain the original cargo box image of the muck truck. If the current light intensity is greater than the first preset value, it means that the current cargo box is in the light state, and the RGB camera can be used to obtain the original image of the cargo box; if the current light intensity is less than or equal to the first preset value, it means that the current cargo box is in darkness In this state, the IR camera can be used to obtain the original image of the cargo box.

Step S702, performing perturbation processing on the original cargo box image to obtain multiple sets of original cargo box images;

Optionally, rotation, translation, scaling, noise addition, blurring, illumination change, and channel change may be performed on the original cargo box image to obtain multiple sets of original cargo box images.

Optionally, 0 to 15 left and right rotation operations may be randomly selected for the original cargo box image.

Optionally, the original cargo box image can be moved up, down, left, and right by 0 to 50 pixels in a translation operation.

Optionally, an image scaling operation of 0.8 to 1.2 may be randomly performed on the original cargo box image.

Optionally, Gaussian or salt and pepper noise with a mean value of 0 to 4.0 can be randomly added to the original cargo box image.

Optionally, a Gaussian blur with templates ranging from 0 to 9 can be randomly added to the original cargo box image.

Optionally, a lighting transformation of 0.8 to 1.2 can be randomly multiplied per pixel on the original crate image.

Optionally, the channels of the original crate image can be transformed, eg, RGB channels to BRG channels.

Step S703, performing clipping processing on multiple sets of original cargo box images to obtain multiple sets of target area images;

Optionally, multiple sets of original cargo box images may be cropped through the ROI to obtain multiple sets of target area images.

Step S704, preprocessing multiple groups of target area images to obtain multiple groups of training samples;

Among them, the preprocessing includes scaling and normalization.

Optionally, multiple sets of target area images can be scaled to a fixed size: 128*128, and then the pixels of the scaled multiple sets of target area images are subtracted from the mean value of 128, and then divided by the variance of 256 to obtain multiple sets of training samples. .

Step S705, constructing a first network model, and sending multiple sets of training samples into the first network model for training.

Optionally, the constructed first network model may be ResNet-10.

Optionally, the confidence level output by the first network model after training has five categories: the confidence level of the top cover of the cargo box being closed, the confidence level of the cargo not loaded in the cargo box, and the amount of the cargo loaded in the cargo box meeting the preset amount. The confidence level, the camera anomaly confidence level, and the amount of goods loaded in the container do not meet the preset confidence level.

It should be noted that the closure of the top cover of the cargo box can mean that the cargo box can be normally closed after loading goods; the top cover of the cargo box can be opened when no cargo is loaded in the cargo box, and the cargo box is empty; If the amount of goods meets the preset amount, it may be that the amount of goods loaded in the container is too large, so that the top cover of the container cannot be closed normally. Check the condition inside the cargo box.

Example 2

According to an embodiment of the present disclosure, a device for detecting the state of a cargo box is also provided. The device can execute the method for detecting the state of a cargo box in the above-mentioned embodiment. The specific implementation manner and preferred application scenario are the same as those of the above-mentioned embodiment. Do repeat.

FIG. 8 is a schematic diagram of a device for detecting the state of a cargo box according to an embodiment of the present disclosure. As shown in FIG. 8 , the device includes:

An acquisition component 82 is configured to acquire images of the cargo box.

The processing component 84 is configured to process the image of the container by using the first network model to obtain the confidence level of the state of the container.

The determining component 86 is configured to determine the status of the cargo box based on the confidence level of the status of the cargo box, wherein the status of the cargo box includes one of the following: the top cover of the cargo box is closed, no cargo is loaded in the cargo box, and the amount of cargo loaded in the cargo box is sufficient The preset amount, the camera is abnormal, and the amount of goods loaded in the container does not meet the preset amount.

It should be noted here that the above-mentioned acquiring component 82, processing component 84 and determining component 86 may be run in a terminal as a part of the device, and the functions implemented by the above-mentioned components may be executed by a processor in the terminal, and the terminal may also be a smart phone. (such as Android mobile phones, iOS mobile phones, etc.), tablet computers, PDAs, and terminal devices such as Mobile Internet Devices (MID) and PAD.

It should be noted that, the above-mentioned obtaining component 82, processing component 84, and determining component 86 may be configured to execute steps S102 to S106 in the embodiments of the present disclosure. The examples and application scenarios implemented by the above components and corresponding steps are the same, but are not limited to the contents disclosed in the above embodiments.

Optionally, the processing component includes: a cropping component, configured to crop an image of the cargo box to obtain an image of the target area; a preprocessing component, configured to preprocess the image of the target area to obtain a processed image; an input component, configured as The processed image is input into the first network model to obtain the confidence level of the state of the container.

It should be noted here that the above-mentioned trimming component, pre-processing component and input component may run in the terminal as a part of the apparatus, and the functions implemented by the above-mentioned components may be executed by a processor in the terminal.

It should be noted that the above-mentioned trimming component, preprocessing component and input component are the same as the examples and application scenarios implemented by the corresponding steps, but are not limited to the contents disclosed in the above-mentioned embodiments.

Optionally, the confidence level of the state of the container in the processing module includes a first confidence level for characterizing the closure of the top cover of the container, a second confidence level for characterizing the unloaded goods in the container, and a second confidence level for characterizing the loaded goods in the container. The third confidence level that the quantity of goods satisfies the preset quantity, the fourth confidence level that is used to represent the abnormality of the camera, and the fifth confidence level that is used to represent that the quantity of goods loaded in the cargo box does not meet the preset quantity.

Optionally, the processing component is further configured to extract the multi-frame container state confidence level corresponding to the multi-frame container images; the processing component is further configured to input the multi-frame container state confidence level into the second network model, and obtain the container state confidence level. .

It should be noted that the examples and application scenarios implemented by the foregoing processing components and corresponding steps are the same, but are not limited to the contents disclosed in the foregoing embodiments.

Optionally, the determining component includes: a first extracting component, configured to extract the status of the container corresponding to the highest value among the first confidence level, the second confidence level, the third confidence level, the fourth confidence level, and the fifth confidence level as cargo. or, a second extracting component, configured to extract the container state whose first confidence level, the second confidence level, the third confidence level, the fourth confidence level and the fifth confidence level are greater than the preset threshold as the container state; Or, the third extraction component is configured to extract the highest value among the first confidence level, the second confidence level, the third confidence level, the fourth confidence level and the fifth confidence level, and the status of the container corresponding to the value greater than the preset threshold is box status.

It should be noted here that the above-mentioned first extracting component, second extracting component or third extracting component may run in a terminal as a part of the apparatus, and the functions implemented by the above-mentioned components may be executed by a processor in the terminal.

It should be noted that the examples and application scenarios implemented by the first extracting component, the second extracting component or the third extracting component and the corresponding steps are the same, but are not limited to the contents disclosed in the above embodiments.

Optionally, the determining component is further configured to determine the state of the container based on the confidence of the state of the container and a preset rule, wherein the preset rule is used to represent the priority of the preset rule used to represent the confidence of the state of the container.

It should be noted that the examples and application scenarios implemented by the above determining components and corresponding steps are the same, but are not limited to the contents disclosed in the above embodiments.

Optionally, the determining component further includes: a first judging component, configured to judge whether the fourth confidence level is greater than the fifth preset value, and if the fourth confidence level is greater than the fifth preset value, determine that the state of the cargo box is an abnormality of the camera; The second judging component is configured to judge whether the first confidence level is greater than the second preset value if the fourth confidence level is less than or equal to the fifth preset value, and determine whether the first confidence level is greater than the second preset value. The state is that the top cover of the cargo box is closed; the third judging component is configured to judge whether the second confidence level is greater than the third preset value if the first confidence level is less than or equal to the second preset value, and if the second confidence level is greater than the third If the second confidence level is less than or equal to the third preset value, it is determined whether the third confidence level is greater than the fourth preset value. , if the third confidence level is greater than the fourth preset value, it is determined that the state of the cargo box is that the amount of goods loaded in the cargo box meets the preset amount; the fifth judgment component is configured so that if the third confidence level is less than or equal to the fourth preset value value, then determine whether the fifth confidence level is greater than the sixth preset value, and if the fifth confidence level is greater than the sixth preset value, it is determined that the state of the container is that the amount of goods loaded in the container does not meet the preset amount.

It should be noted here that the above-mentioned first judging component, second judging component, third judging component, fourth judging component and fifth judging component may run in the terminal as a part of the device, and may be processed by a processor in the terminal. Execute the function implemented by the above components.

It should be noted that the examples and application scenarios implemented by the first judgment component, the second judgment component, the third judgment component, the fourth judgment component, and the fifth judgment component are the same as those implemented by the corresponding steps, but are not limited to those in the above embodiments. public content.

Optionally, the preprocessing in the preprocessing unit includes at least one of the following: scaling processing and normalization processing.

Optionally, the acquiring component is further configured to acquire the original cargo box image; the processing component is further configured to process the original cargo box image to obtain multiple sets of training samples; the device further includes: a training component configured to use multiple sets of training samples to The initial model is trained to obtain the first network model.

Optionally, the processing component further includes: a perturbation processing component configured to perform perturbation processing on the original cargo box images to obtain multiple sets of original cargo box images; the trimming component is further configured to perform clipping processing on multiple sets of original cargo box images to obtain multiple groups of target area images; the preprocessing component is further configured to preprocess multiple groups of target area images to obtain multiple groups of training samples.

It should be noted here that the above-mentioned disturbance processing component may run in the terminal as a part of the apparatus, and the functions implemented by the above-mentioned component may be executed by a processor in the terminal.

It should be noted that the above disturbance processing component, trimming component and preprocessing component have the same examples and application scenarios as the corresponding steps, but are not limited to the contents disclosed in the above embodiments.

Optionally, the disturbance processing in the disturbance processing component includes at least one of the following: rotation, translation, scaling, noise addition, blurring, illumination change, and channel change.

Optionally, the acquisition component further includes: a detection component configured to detect whether the current light intensity is greater than the first preset value; the first acquisition component configured to use the first camera when the current light intensity is greater than the first preset value The image of the cargo box is acquired; the second acquisition component is configured to use the second camera to acquire the image of the cargo box when the current light intensity is less than or equal to the first preset value.

It should be noted here that the above-mentioned detection component, the first acquisition component and the second acquisition component may run in the terminal as a part of the apparatus, and the functions implemented by the above components may be executed by a processor in the terminal.

It should be noted that the examples and application scenarios implemented by the detection component, the first acquisition component, and the second acquisition component are the same as those implemented by the corresponding steps, but are not limited to the contents disclosed in the foregoing embodiments.

Optionally, the apparatus further includes an adding component configured to add a normalization layer to the initial layer of the initial model, wherein the normalization layer is used for normalizing multiple groups of training samples.

It should be noted here that the above-mentioned adding component may run in the terminal as a part of the apparatus, and the functions implemented by the above-mentioned component may be executed by a processor in the terminal.

It should be noted that the examples and application scenarios implemented by the above-mentioned adding components and corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned embodiments.

Example 3

According to an embodiment of the present disclosure, a computer-readable storage medium is also provided, where the computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the cargo box in the above-mentioned Embodiment 1 Status detection method.

Example 4

According to an embodiment of the present disclosure, a processor is also provided, and the processor is used for running a program, wherein the method for detecting the state of a cargo box in the above-mentioned Embodiment 1 is executed when the program is running.

The above-mentioned serial numbers of the embodiments of the present disclosure are only for description, and do not represent the advantages or disadvantages of the embodiments.

In the above-mentioned embodiments of the present disclosure, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.

In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are only illustrative, for example, the division of units may be a logical function division, and there may be other division methods in actual implementation, for example, multiple units or components may be combined or integrated into Another system, or some features can be ignored, or not implemented. On the other hand, the shown or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, indirect coupling or communication connection of units or modules, and may be in electrical or other forms.

The units described as separate components may or may not be physically separated, and components shown as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution in this embodiment.

In addition, each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above-mentioned integrated units may be implemented in the form of hardware, or may be implemented in the form of software functional units.

The integrated unit, if implemented in the form of a software functional unit and sold or used as an independent product, may be stored in a computer-readable storage medium. Based on such understanding, the technical solutions of the present disclosure essentially or the parts that contribute to the prior art, or all or part of the technical solutions can be embodied in the form of software products, and the computer software products are stored in a storage medium , including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present disclosure. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes .

The above are only the preferred embodiments of the present disclosure. It should be pointed out that for those skilled in the art, without departing from the principles of the present disclosure, several improvements and modifications can be made. It should be regarded as the protection scope of the present disclosure.

Industrial Applicability

After obtaining the image of the cargo box, the first network model is used to process the image of the cargo box to obtain the confidence of the state of the cargo box, and then the state of the cargo box is determined based on the confidence of the state of the cargo box. It can detect whether the top cover of the shipping box is effectively closed, and can also detect whether the cargo box is loaded with goods, and detect the capacity of the goods loaded in the cargo box, and can detect the abnormality of the camera when the obtained image of the cargo box does not show the status of the cargo box , so as to effectively detect the status of the cargo box, compared with the solution of using radar to realize detection in the prior art, the cost is reduced, and the method of detecting the status of the cargo box in the related art can not detect whether the cargo box is loaded with goods. question.

Claims

A method for detecting the state of a cargo box, comprising:

Get the image of the cargo box;

Using the first network model to process the image of the container to obtain the confidence level of the state of the container;

Based on the confidence level of the state of the cargo box, the state of the cargo box is determined, wherein the state of the cargo box includes one of the following: the top cover of the cargo box is closed, no cargo is loaded in the cargo box, and the amount of cargo loaded in the cargo box meets a preset value volume, abnormal camera, and the amount of goods loaded in the container does not meet the preset amount.
The method according to claim 1, wherein, using the first network model to process the image of the cargo box to obtain the confidence level of the state of the cargo box, comprising:

Cropping the image of the cargo box to obtain an image of the target area;

Preprocessing the target area image to obtain a processed image;

The processed image is input into the first network model to obtain the confidence level of the state of the cargo box.
3. The method of claim 2, wherein the container state confidence level comprises: a first confidence level characterizing the closure of the container top lid, a second confidence level characterizing the unloaded cargo in the container a confidence level, a third confidence level used to characterize that the amount of goods loaded in the container satisfies the preset amount, a fourth confidence level used to characterize the abnormality of the camera, and a confidence level used to characterize the container The fifth confidence level that the quantity of goods loaded inside does not meet the preset quantity.
The method according to claim 1, wherein the first network model is used to process the image of the cargo box to obtain the confidence level of the state of the cargo box, further comprising:

extracting the multi-frame container state confidence levels corresponding to the multi-frame container images;

The multi-frame container state confidence level is input into the second network model to obtain the container state confidence level.
The method of claim 3, wherein determining the state of the container based on the confidence level of the state of the container comprises:

extracting the container state corresponding to the highest value among the first confidence level, the second confidence level, the third confidence level, the fourth confidence level and the fifth confidence level as the container state; or,

extracting the container state whose first confidence level, the second confidence level, the third confidence level, the fourth confidence level and the fifth confidence level are greater than a preset threshold as the container state; or,

The highest value among the first confidence level, the second confidence level, the third confidence level, the fourth confidence level, and the fifth confidence level is extracted, and the state of the container that is greater than the preset threshold value is: The state of the box.
The method according to claim 3, wherein determining the status of the cargo box based on the confidence level of the status of the cargo box comprises: determining the status of the cargo box based on the confidence level of the status of the cargo box and a preset rule, wherein the The preset rule is used to represent the priority of the confidence level of the state of the container.
6. The method of claim 6, wherein determining the container state based on the container state confidence comprises:

judging whether the fourth confidence level is greater than a fifth preset value;

If the fourth confidence level is greater than the fifth preset value, it is determined that the state of the cargo box is an abnormality of the camera;

If the fourth confidence level is less than or equal to the fifth preset value, determine whether the first confidence level is greater than the second preset value;

If the first confidence level is greater than the second preset value, determining that the state of the cargo box is that the top cover of the cargo box is closed;

If the first confidence level is less than or equal to the second preset value, determine whether the second confidence level is greater than a third preset value;

If the second confidence level is greater than the third preset value, it is determined that the state of the cargo box is that no cargo is loaded in the cargo box;

If the second confidence level is less than or equal to the third preset value, determine whether the third confidence level is greater than a fourth preset value;

If the third confidence level is greater than the fourth preset value, determining that the state of the cargo box is that the amount of goods loaded in the cargo box meets the preset amount;

If the third confidence level is less than or equal to the fourth preset value, determine whether the fifth confidence level is greater than the sixth preset value;

If the fifth confidence level is greater than the sixth preset value, it is determined that the state of the cargo box is that the amount of goods loaded in the cargo box does not meet the preset amount.
The method of claim 2, wherein the preprocessing includes at least one of: scaling and normalization.
The method of claim 1, wherein the method further comprises:

Get the original crate image;

processing the original cargo box image to obtain multiple sets of training samples;

The initial model is trained by using the multiple sets of training samples to obtain the first network model.
The method according to claim 9, wherein the original cargo box image is processed to obtain multiple sets of training samples, including:

Performing perturbation processing on the original cargo box image to obtain multiple sets of original cargo box images;

Perform clipping processing on the multiple sets of original cargo box images to obtain multiple sets of target area images;

The multiple sets of target area images are preprocessed to obtain the multiple sets of training samples.
11. The method of claim 10, wherein the perturbation processing includes at least one of: rotation, translation, scaling, noise, blur, illumination change, and channel change.
The method of claim 1, wherein acquiring an image of the cargo box comprises:

Detecting whether the current light intensity is greater than the first preset value;

If the current light intensity is greater than the first preset value, use the first camera to acquire the image of the cargo box;

If the current light intensity is less than or equal to the first preset value, use the second camera to acquire the image of the cargo box.
The method according to claim 9, wherein, before using the multiple sets of training samples to train an initial model to obtain the first network model, the method further comprises:

A normalization layer is added to the initial layer of the initial model, wherein the normalization layer is used for normalizing the multiple groups of training samples.
A detection device for the state of a cargo box, comprising:

The acquisition module is used to acquire the image of the cargo box;

a processing module, configured to process the image of the cargo box by using the first network model to obtain the confidence level of the status of the cargo box;

The determining module is configured to determine the state of the cargo box based on the confidence level of the state of the cargo box, wherein the state of the cargo box includes one of the following: the top cover of the cargo box is closed, the cargo is not loaded in the cargo box, and the cargo is loaded in the cargo box. The amount of goods meets the preset amount, the camera is abnormal, and the amount of goods loaded in the container does not meet the preset amount.
A storage medium, the storage medium comprising a stored program, wherein when the program is executed, a device where the storage medium is located is controlled to execute the method for detecting the state of a cargo box according to any one of claims 1 to 13.
A processor, which is used for running a program, wherein the method for detecting the state of a cargo box according to any one of claims 1 to 13 is executed when the program is running.