CN116563598A - Network degree detection method and device, storage medium and electronic equipment - Google Patents

Abstract

The application discloses a network degree detection method, a network degree detection device, a storage medium and electronic equipment, wherein the method comprises the following steps: performing image processing on an image to be tested to obtain a plurality of target contour images corresponding to the to-be-tested attempt, wherein each target contour image comprises a target spinning contour; performing network node characteristic extraction processing on the to-be-tested image based on a preset semantic segmentation model to obtain a node prediction image corresponding to the to-be-tested image; and detecting the node prediction image by taking each target contour image as a mask image to obtain a network degree detection result of each target spinning. The detection method for obtaining the network degree index based on image processing is high in efficiency and more accurate in detection result.

Description

A network degree detection method, device, storage medium and electronic equipment

technical field

The invention relates to the field of chemical fibers, in particular to a network degree detection method, device, storage medium and electronic equipment.

Background technique

Chemical fiber products are closely related to our life, our daily wear and living environment cannot do without chemical fiber products. my country's chemical fiber industry has a large stock and a long history, and the market potential for industrial intelligent transformation and industrial upgrading is huge. With the implementation of the innovation-driven strategy of the global fiber industry, the promotion of the intersection and integration of the fiber field and emerging technologies has put forward new requirements for chemical fiber manufacturers. . Therefore, the quality inspection of chemical fiber products has a great influence on the subsequent production links. Network degree, as one of the key quality indicators of chemical fiber products, refers to the number of network nodes in chemical fiber filaments per unit length. For chemical fiber filaments, the degree of network per unit length needs to meet a certain range. Excessive network numbers will lead to relaxation. During processing, the silk cannot be fully untwisted, and the dyeing rate is different during dyeing, resulting in network spots on the surface of the fabric. If the number of networks is too small, the network points are easy to loosen and fluff during the weaving process. Therefore, the correct network degree is very important for the subsequent production process.

Now many chemical fiber spinning factories use the traditional artificial network index detection method. This method puts multiple spinning threads in parallel in water through the water bath method, and obtains the network degree detection index by visual counting, but it has many advantages in actual use. Big limitations. First of all, manual inspection relies on people's subjective evaluation, which is greatly unstable, unreliable and non-quantifiable due to the influence of people's mood, thinking and subjective and objective factors of lighting lamps, which brings great challenges to product quality evaluation. Lots of unstable and unreliable factors. Secondly, the human eye cannot meet the detection requirements of the network degree index when the product is produced at high speed.

Contents of the invention

In view of this, the present invention provides a network degree detection method, device, electronic equipment and storage medium, the main purpose of which is to solve the current problem that manual network degree detection is not accurate enough.

In order to solve the above problems, this application provides a network degree detection method, including:

Image processing is performed on the image to be tested to obtain a plurality of target contour images corresponding to the image to be tested, wherein each of the target contour images includes a target spinning profile;

Performing network node feature extraction processing on the image to be tested based on a preset semantic segmentation model to obtain a node prediction image corresponding to the image to be tested;

Using each target outline image as a mask image to detect the node prediction image to obtain the network degree detection result of each target spinning.

Optionally, performing image processing on the image to be tested to obtain a target contour image corresponding to the target spinning in the image to be tested, specifically including:

performing pre-processing on the image to be tested to obtain a pre-processed first image;

performing threshold processing on the first image to obtain a second image in which the spinning area is separated from the non-spinning area;

performing dilation and erosion processing on the second image to obtain a spinning profile image;

The contour detection is performed based on the spinning contour image, and the target contour image corresponding to the target spinning in the image to be tested is obtained.

Optionally, performing network node feature extraction processing on the image to be tested based on a preset semantic segmentation model to obtain a node prediction image corresponding to the image to be tested, the method further includes: training to obtain the semantic segmentation models, including:

Acquiring several historical images and tag images corresponding to each of the historical images;

performing grouping processing on each of the historical images and each of the label images to obtain a first data set comprising several first historical images and label images corresponding to each first historical image; a second data set of label images corresponding to each second historical image;

Each of the first data sets is used as a training sample, and a preset semantic segmentation method is used for model training to obtain the current semantic segmentation model and current model parameters, and the current semantic segmentation model and current model parameters are performed based on the second data set. Detect to obtain the semantic segmentation model.

Optionally, each of the first data sets is used as a training sample, and a preset semantic segmentation method is used for model training to obtain the current semantic segmentation model and current model parameters, and the current semantic segmentation is performed based on the second data set. Model and current model parameter are detected, obtain described semantic segmentation model, specifically comprise the following steps:

Step 1: Based on each of the first data sets as a training sample, a preset semantic segmentation method is used for model training to obtain a current semantic segmentation model;

Step 2: Determine whether the current model accuracy value meets the preset accuracy value. When the current model accuracy value is less than the preset accuracy value, perform step 3; if the current model accuracy value is greater than or equal to the preset accuracy value, repeat step 1. Obtain the updated current semantic segmentation model and the updated current model parameters;

Step 3: Detect the current semantic segmentation model and current model parameters based on the second data set to obtain the semantic segmentation model.

Optionally, the acquiring the label image corresponding to each of the historical images specifically includes:

Perform image labeling processing on each of the historical images, and obtain a label file corresponding to each of the historical images, and the label file includes node area information and label category information of the historical image;

Based on the node area information and the label category information corresponding to each of the historical images, perform label image creation on the label file corresponding to each of the historical images, and obtain a label image corresponding to each of the historical images.

Optionally, the node prediction image is detected based on each target outline image as a mask image to obtain the network degree detection result of each target spinning, which specifically includes:

Using each target outline image as a mask image, detecting the node prediction image, and extracting the node image corresponding to the target spinning;

Based on the node image corresponding to the target spinning, calculate and obtain the network degree of the target spinning, and obtain the detection result of the network degree.

Optionally, the calculating and obtaining the network degree of the target spinning based on the node image corresponding to the target spinning, to obtain the detection result of the network degree, specifically includes:

Obtaining the number of nodes corresponding to the target spinning based on the node image corresponding to the target spinning;

Based on the number of pixels of the node image and the actual distance corresponding to each pixel, calculate and obtain the actual length of the target spinning;

Based on the number of nodes corresponding to the target spinning and the actual length of the target spinning, calculate and obtain the network degree of the target spinning, and obtain the detection result of the network degree.

In order to solve the above problems, the present application provides a network degree detection device, including:

Image processing module: used to perform image processing on the image to be tested to obtain a number of target contour images corresponding to the image to be tested, wherein each of the target contour images includes the contour of the target spinning;

Feature extraction module: for performing network node feature extraction processing on the image to be tested based on a preset semantic segmentation model, to obtain a node prediction image corresponding to the image to be tested;

A detection module: used to use each target outline image as a mask image to detect the node prediction image, and obtain a network degree detection result of each target spinning.

In order to solve the above problems, the present application provides a storage medium, the storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above network degree detection method are realized.

In order to solve the above problems, the present application provides an electronic device, which at least includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program on the memory, the above-mentioned network degree detection method is implemented. step.

In this application, image processing is performed on the image of the spinning to be tested to obtain the target contour image corresponding to the target spinning, based on the preset semantic segmentation model, feature extraction is performed on the image to be tested to obtain a node prediction image, based on the target contour image as a mask image pair The node prediction image is detected to obtain the detection result of the network degree of the target spinning, and the detection method based on the image processing to obtain the network degree index has high efficiency and the detection result is more accurate.

The above description is only an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention, it can be implemented according to the contents of the description, and in order to make the above and other purposes, features and advantages of the present invention more obvious and understandable , the specific embodiments of the present invention are enumerated below.

Description of drawings

Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiment. The drawings are only for the purpose of illustrating a preferred embodiment and are not to be considered as limiting the invention. Also throughout the drawings, the same reference numerals are used to designate the same parts. In the attached picture:

Fig. 1 is the flowchart of a kind of network degree detection method of the embodiment of the present application;

FIG. 2 is a flowchart of a network degree detection method according to another embodiment of the present application;

FIG. 3 is a structural block diagram of a network degree detection device according to yet another embodiment of the present application;

Figure 4(a) the image to be tested in this application;

Figure 4(b) is the spinning profile image obtained after corrosion and expansion treatment in this application;

Fig. 4 (c) In this application, the node prediction image obtained after the network node feature extraction process is performed on the test image by the semantic segmentation model;

Figure 4(d) the target contour image corresponding to the target spinning in this application;

Figure 4(e) The node image corresponding to the target spinning in this application;

Figure 4(f) is the image obtained by marking the circumscribed rectangle of the node outline on the image to be tested.

Detailed ways

Various aspects and features of the present application are described herein with reference to the accompanying drawings.

It should be understood that various modifications may be made to the embodiments applied for herein. Accordingly, the above description should not be viewed as limiting, but only as exemplifications of embodiments. Those skilled in the art will envision other modifications within the scope and spirit of the application.

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments of the application and, together with the general description of the application given above and the detailed description of the embodiments given below, serve to explain the embodiments of the application. principle.

These and other characteristics of the present application will become apparent from the following description of preferred forms of embodiment given as non-limiting examples with reference to the accompanying drawings.

It should also be understood that, while the application has been described with reference to a few specific examples, those skilled in the art will be able to implement certain other equivalents of the application.

The above and other aspects, features and advantages of the present application will become more apparent in view of the following detailed description when taken in conjunction with the accompanying drawings.

Specific embodiments of the present application are hereinafter described with reference to the accompanying drawings; however, it should be understood that the applied embodiments are merely examples of the present application, which can be implemented in various ways. Well-known and/or repetitive functions and constructions are not described in detail to avoid obscuring the application with unnecessary or redundant detail. Therefore, specific structural and functional details disclosed herein are not intended to be limiting, but merely as a basis for the claims and as a representative basis for teaching one skilled in the art to variously employ the present invention in virtually any suitable detailed structure. Apply.

This specification may use the phrases "in one embodiment," "in another embodiment," "in yet another embodiment," or "in other embodiments," which may refer to the same or one or more of the different embodiments.

The embodiment of the present application provides a network degree detection method, as shown in Figure 1, including:

Step S101: Perform image processing on the image to be tested to obtain several target contour images corresponding to the image to be tested, wherein each of the target contour images includes the target spinning contour;

In the specific implementation process of this step, firstly, pre-processing is performed on the image to be tested, mean value filtering is performed and prominent noise points are smoothed, the size of the image is adjusted to obtain the first image, and threshold processing is performed on the first image , the threshold processing can adopt the ostu threshold detection method, through the threshold detection, the spinning area in the image is separated from the non-spinning area to obtain a second image, and finally the second image is corroded and expanded, and the In the second image, the thread area and the background area are separated to obtain the spinning contour image, and finally the spinning contour image is subjected to contour detection processing to obtain the contour corresponding to each spinning, so as to obtain the target spinning contour image. The target contour image corresponding to the wire.

Step S102: Perform network node feature extraction processing on the image to be tested based on a preset semantic segmentation model, and obtain a node prediction image corresponding to the image to be tested;

In the specific implementation process of this step, the network node feature extraction is performed on the image to be tested based on the preset semantic segmentation model and model parameters. In this step, the semantic segmentation model first loads the model parameters that have been trained, and the image to be tested The image is input into the trained semantic segmentation model, and the node prediction image containing network node features is outputted. The node prediction image marks the node area as white and other areas as black.

Step S103: using each target outline image as a mask image, detecting the node prediction image, and obtaining a network degree detection result of each target spinning.

In the specific implementation process of this step, based on each of the target contour images as a mask image, the node prediction image is detected, and the node image corresponding to the target spinning is obtained by extraction, and based on the node image corresponding to the target spinning , filter the nodes according to the length and width of the nodes in the node image corresponding to the target spinning, and obtain the number of nodes corresponding to the target spinning, based on the number of pixels in the node image and the actual distance corresponding to each pixel, calculate and obtain The actual length of the target spinning is calculated to obtain the network degree of the target spinning based on the number of nodes corresponding to the target spinning and the actual length of the target spinning, and the detection result of the network degree is obtained.

In this application, image processing is performed on the image of the spinning to be tested to obtain the target contour image corresponding to the target spinning, based on the preset semantic segmentation model, feature extraction is performed on the image to be tested to obtain a node prediction image, based on the target contour image as a mask image pair The node prediction image is detected to obtain the detection result of the network degree of the target spinning, and the detection method based on the image processing to obtain the network degree index has high efficiency and the detection result is more accurate.

Another embodiment of the present application provides a network degree detection method, as shown in Figure 2, including:

Step S201: acquiring several historical images;

In the specific implementation process of this step, industrial cameras can be used to collect historical spinning images, and the historical spinning can be placed in the water tank through the network degree detection sampling equipment. When the historical spinning is dispersed in the water tank and has a better shape , to obtain several historical spinning images by taking photos with a camera. In order to enrich the sample set, cut the several historical spinning images and unify the size of the pictures to a specified size. For example: take 100 historical spinning images with a camera. The 100 historical spinning images are cropped to obtain 276 sample images, and the size of the sample images is uniformly converted to an image with a size of 512*512 pixels to obtain 276 historical images. In this embodiment, there is no limit to the number of historical spinning images captured by the camera and the number of sample images after cropping, and the number of historical spinning images captured by the camera and sample images after cropping can be adjusted according to actual needs.

Step S202: Based on each of the historical images, obtain a label image corresponding to each of the historical images;

In the specific implementation process of this step, first perform image labeling processing on each of the historical images, and obtain a label file corresponding to each of the historical images, and the label file includes node area information and label category information of the historical image Can use labelme software to carry out image marking process to described historical image, manually mark the node area in each image, the category of the marked node area is recorded as network node area, obtain the corresponding label file with each described historical image, The label file is a json file, which records the point set information of the marked node area in the image and its corresponding label category information, for example: 276 historical images as mentioned above, you can use the labelme software for the described 276 label files are obtained after image labeling processing of the historical images; then, based on the node area information and the label category information corresponding to each of the historical images, label images are produced for the label files corresponding to each of the historical images, A label image corresponding to each of the historical images is obtained. Specifically, according to the node area information and label category information in the label file, the python program is used to draw the label image, the pixel of the node area in the label file is assigned a value of 255, and the area other than the node area is assigned a value of 0, A label image corresponding to the label file is obtained. In this embodiment, there is no limit to the pixel assignment size, and the assignment in this scheme is to make the node area more prominent.

Step S203: Based on each of the historical images and each of the label images, train and obtain a semantic segmentation model;

In the specific implementation process of this step, first: perform grouping processing on each of the historical images and each of the label images, and obtain a first data set including several first historical images and label images corresponding to each of the first historical images , and a second data set including several second historical images and label images corresponding to each second historical image. For example: the above-mentioned 276 historical images and the label images corresponding to the 276 historical images are grouped, and the 221 first historical images and the label images corresponding to the 221 first historical images are randomly selected. The first data set is formed, and the remaining 55 second historical images and label images corresponding to the 55 second historical images form the second data set.

Secondly, each of the first data sets is used as a training sample, and the preset semantic segmentation method is used for model training to obtain the current semantic segmentation model and current model parameters, and the current semantic segmentation model and the current model are analyzed based on the second data set. Parameters are detected to obtain the semantic segmentation model. Specific steps are as follows:

Step 1: Based on each of the first data sets as a training sample, a preset semantic segmentation method is used for model training to obtain a current semantic segmentation model;

Specifically, each of the first data sets is used as a training sample, and the U-Net deep semantic segmentation method is used for training, and the BCEWithLogitsLoss loss function is used to obtain the current semantic segmentation model after training.

Step 2: Determine whether the current model accuracy value meets the preset accuracy value. When the current model accuracy value is less than or equal to the preset accuracy value, perform step 3; if the current model accuracy value is greater than the preset accuracy value, repeat steps 1. Obtain the updated current semantic segmentation model and the updated current model parameters;

In the specific implementation process of this step, the preset precision value can be set to 0.01. When the current model precision value is less than or equal to 0.01, perform step 3 to obtain the semantic segmentation model; repeat the steps when the current model precision value is greater than 0.01 First, the updated current semantic segmentation model is obtained after training, and the preset accuracy value can be adjusted according to actual needs.

Step 3: Detect the current semantic segmentation model and current model parameters based on the second data set to obtain the semantic segmentation model.

Based on the second data set, the semantic segmentation model obtained after cyclic iteration training is detected, the output result obtained by inputting each of the second historical images into the semantic segmentation model, and the label image corresponding to the second historical image By performing comparison and detection, the similarity between the obtained output image and the label image corresponding to the second image meets the requirements, and the semantic segmentation model is obtained.

Step S204: Perform image processing on the image to be tested to obtain several target contour images corresponding to the test image, wherein each of the target contour images includes the target spinning contour;

In the specific implementation process of this step, first, the image of the spinning sample to be tested is collected by an industrial camera, and the size of the collected image is adjusted. In this scheme, in order to facilitate the input of the semantic segmentation network, the size of the image to be tested is adjusted. is a multiple of 32, the image to be tested is obtained, and the image to be tested is shown in Figure 4(a). Then carry out pre-processing to the image to be tested, perform mean value filtering and smooth the outstanding noise points, adjust the size of the picture to obtain the first image, and perform threshold processing on the first image, the threshold processing can adopt ostu Threshold value detection method, through threshold value detection, the spinning area in the image is separated from the non-spinning area to obtain a second image, and finally the second image is corroded and expanded, and the silk thread area in the second image is Separate from the background area to obtain the spinning profile image, as shown in Figure 4(b), and finally perform profile detection processing on the spinning profile image to obtain the corresponding profile of each spinning , to obtain the target contour image corresponding to the target spinning, the target contour image is shown in FIG. 4( d ).

Step S205: performing network node feature extraction processing on the image to be tested based on the semantic segmentation model, and obtaining a node prediction image corresponding to the image to be tested;

In the specific implementation process of this step, the network node feature extraction is performed on the image to be tested based on the semantic segmentation model and model parameters. In this step, the semantic segmentation model first loads the model parameters that have been trained, and the image to be tested is Input into the trained semantic segmentation model, the output node area is marked as white, and other areas are marked as black, and the node features after filtering and expansion processing are more obvious. The node prediction image, the node prediction image is shown in Figure 4 (c) shown.

Step S206: using each target outline image as a mask image, detecting the node prediction image, and extracting the node image corresponding to the target spinning;

In the specific implementation process of this step, the node prediction image is detected based on each of the target contour images as a mask image, and the contour of the node in the node prediction image corresponding to the target contour image is extracted to obtain the target spinning correspondence The node image corresponding to the target spinning is shown in Figure 4(e).

Step S207: Based on the node image corresponding to the target spinning, calculate and obtain the network degree of the target spinning, and obtain the detection result of the network degree.

In the specific implementation process of this step, first, based on the node image corresponding to the target spinning, the number of nodes corresponding to the target spinning is obtained; specifically, based on the node image of the target spinning, according to the length and width of the node The nodes are screened, and the interfering nodes that do not meet the node characteristics are eliminated, and the circumscribed rectangle is taken for each node outline to obtain the number of nodes corresponding to the target spinning, and the screened nodes are marked on the image to be tested, as shown in Figure 4(f). Then, based on the number of pixels of the node image and the actual distance corresponding to each pixel, the actual length of the target spinning is calculated; for example, an image with a size of 512*512 pixels as described above is equivalent to 512* For an image composed of 512 pixels, the actual spinning length is approximately equal to the product of 512 pixels and the actual length represented by each pixel. For example, when the actual distance corresponding to each pixel point calculated based on the camera parameters is 0.025 cm, the actual spinning length acquired at this time is 512*0.025=12.8 cm. Finally, based on the number of nodes corresponding to the target spinning and the actual length of the target spinning, calculate and obtain the network degree of the target spinning, and obtain the detection result of the network degree. Specifically, the number of nodes corresponding to the target spinning is divided by the actual length of the target spinning to calculate the network degree of the target spinning. Obtain the detection result of the network degree.

In this application, the target contour image corresponding to the target spinning is obtained by image processing on the image of the spinning to be tested, and the node prediction image is obtained by constructing a semantic segmentation model to extract the features of the image to be tested, and based on the target contour image as a mask image for all The network degree detection result of the target spinning is obtained by detecting the predicted image of the above nodes. The detection method based on the image processing to obtain the network degree index has high efficiency and more accurate detection results.

Another embodiment of the present application provides a network degree detection device, as shown in FIG. 3 , including:

Image processing module 1: used to perform image processing on the image to be tested to obtain several target contour images corresponding to the image to be tested, wherein each of the target contour images includes the contour of the target spinning;

Feature extraction module 2: used to perform network node feature extraction processing on the image to be tested based on a preset semantic segmentation model, and obtain a node prediction image corresponding to the image to be tested;

Detection module 3: used to use each target outline image as a mask image to detect the node prediction image, and obtain the network degree detection result of each target spinning.

In the specific implementation process, the image processing module is specifically used to: perform pre-processing on the image to be tested to obtain the pre-processed first image; perform threshold processing on the first image to obtain the spinning area A second image separated from the non-spinning area; performing expansion and corrosion processing on the second image to obtain a spinning profile image; performing profile detection based on the spinning profile image to obtain a target corresponding to the target spinning in the image to be tested Contour image.

In the specific implementation process, the network degree detection device also includes: a model training module, and the model training module is specifically used to: acquire a number of historical images and label images corresponding to each of the historical images; and performing grouping processing with each of the label images to obtain a first data set comprising several first historical images and label images corresponding to each first historical image; and comprising a plurality of second historical images and corresponding to each second historical image The second data set of the label image; Each of the first data sets is used as a training sample, and the preset semantic segmentation method is used for model training to obtain the current semantic segmentation model and current model parameters, and the second data set is used for the described The current semantic segmentation model and current model parameters are detected to obtain the semantic segmentation model.

In the specific implementation process, the model training module is also used to: use each of the first data sets as training samples, and use the preset semantic segmentation method to perform model training to obtain the current semantic segmentation model and current model parameters, based on the first The second data set detects the current semantic segmentation model and the current model parameters, and obtains the semantic segmentation model, which specifically includes: Step 1: Based on each of the first data sets as a training sample, a preset semantic segmentation method is used to perform Model training to obtain the current semantic segmentation model; Step 2: Determine whether the current model accuracy value meets the preset accuracy value, and when the current model accuracy value is less than the preset accuracy value, perform Step 3; when the current model accuracy value is greater than or equal to the preset In the case of the accuracy value, repeat step 1 to obtain the updated current semantic segmentation model and updated current model parameters; step 3: detect the current semantic segmentation model and current model parameters based on the second data set, and obtain The semantic segmentation model.

In the specific implementation process, the model training module is also used to: perform image labeling processing on each of the historical images, and obtain a label file corresponding to each of the historical images, and the label file includes nodes of the historical images area information and label category information; based on the node area information and the label category information corresponding to each of the historical images, perform label image creation on the label files corresponding to each of the historical images, and obtain the corresponding to each of the historical images label image.

In the specific implementation process, the detection module 3 is specifically used to: use each of the target contour images as a mask image, detect the node prediction image, and extract the node image corresponding to the target spinning; The node image corresponding to the spinning is calculated to obtain the network degree of the target spinning, and the detection result of the network degree is obtained.

In the specific implementation process, the detection module 3 is also used to: obtain the number of nodes corresponding to the target spinning based on the node image corresponding to the target spinning; the number of pixels based on the node image corresponds to each pixel The actual distance is calculated to obtain the actual length of the target spinning; based on the number of nodes corresponding to the target spinning and the actual length of the target spinning, the network degree of the target spinning is calculated to obtain the network degree of detection results.

Another embodiment of the present application provides a storage medium, the storage medium stores a computer program, and when the computer program is executed by a processor, the following method steps are implemented:

Step 1. Perform image processing on the image to be tested to obtain a number of target contour images corresponding to the image to be tested, wherein each of the target contour images includes the target spinning contour;

Step 2, performing network node feature extraction processing on the image to be tested based on a preset semantic segmentation model, and obtaining a node prediction image corresponding to the image to be tested;

Step 3: Using each target outline image as a mask image, detect the node prediction image, and obtain the network degree detection result of each target spinning.

For the specific implementation process of the steps of the above method, reference may be made to the above embodiments of any network degree detection method, and this embodiment will not be repeated here.

In this application, image processing is performed on the image of the spinning to be tested to obtain the target contour image corresponding to the target spinning, based on the preset semantic segmentation model, feature extraction is performed on the image to be tested to obtain a node prediction image, based on the target contour image as a mask image pair The node prediction image is detected to obtain the detection result of the network degree of the target spinning, and the detection method based on the image processing to obtain the network degree index has high efficiency and the detection result is more accurate.

Another embodiment of the present application provides an electronic device, including at least a memory and a processor, where a computer program is stored on the memory, and the processor implements the following method steps when executing the computer program on the memory:

Step 2, performing network node feature extraction processing on the image to be tested based on a preset semantic segmentation model, and obtaining a node prediction image corresponding to the image to be tested;

Step 3: Using each target outline image as a mask image, detect the node prediction image, and obtain the network degree detection result of each target spinning.

For the specific implementation process of the steps of the above method, reference may be made to the above embodiments of any network degree detection method, and this embodiment will not be repeated here.

In this application, image processing is performed on the image of the spinning to be tested to obtain the target contour image corresponding to the target spinning, based on the preset semantic segmentation model, feature extraction is performed on the image to be tested to obtain a node prediction image, based on the target contour image as a mask image pair The node prediction image is detected to obtain the detection result of the network degree of the target spinning, and the detection method based on the image processing to obtain the network degree index has high efficiency and the detection result is more accurate.

For the specific implementation process of the steps of the above method, reference may be made to the above embodiments of any network degree detection method, and this embodiment will not be repeated here.

The above embodiments are only exemplary embodiments of the present application, and are not intended to limit the present application, and the protection scope of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent replacements to the present application within the spirit and protection scope of the present application, and such modifications or equivalent replacements shall also be deemed to fall within the protection scope of the present application.

Claims (10)

1. A network degree detection method, comprising:

performing image processing on an image to be tested to obtain a plurality of target contour images corresponding to the to-be-tested attempt, wherein each target contour image comprises a target spinning contour;

performing network node characteristic extraction processing on the to-be-tested image based on a preset semantic segmentation model to obtain a node prediction image corresponding to the to-be-tested image;

and detecting the node prediction image by taking each target contour image as a mask image to obtain a network degree detection result of each target spinning.

2. The method of claim 1, wherein the image processing is performed on the image to be tested to obtain a target contour image corresponding to the target spinning in the image to be tested, and the method specifically comprises:

pre-processing the to-be-detected image to obtain a first image after pre-processing;

performing threshold processing on the first image to obtain a second image with a spinning area separated from a non-spinning area;

performing expansion corrosion treatment on the second image to obtain a spinning contour image;

and performing contour detection based on the spinning contour image to obtain a target contour image corresponding to target spinning in the image to be tested.

3. The method according to claim 1, wherein, in performing network node feature extraction processing on the image to be tested based on a preset semantic segmentation model, a node prediction image corresponding to the image to be tested is obtained, the method further comprises: training to obtain the semantic segmentation model specifically comprises the following steps:

acquiring a plurality of historical images and tag images corresponding to the historical images;

grouping each history image and each label image to obtain a first data set containing a plurality of first history images and the label images corresponding to each first history image; and a second dataset comprising a plurality of second history images and a label image corresponding to each second history image;

and taking each first data set as a training sample, carrying out model training by adopting a preset semantic segmentation method to obtain a current semantic segmentation model and current model parameters, and detecting the current semantic segmentation model and the current model parameters based on a second data set to obtain the semantic segmentation model.

4. The method of claim 3, wherein the model training is performed by using each of the first data sets as a training sample and using a preset semantic segmentation method to obtain a current semantic segmentation model and current model parameters, and the current semantic segmentation model and the current model parameters are detected based on a second data set to obtain the semantic segmentation model, and the method specifically comprises the following steps:

step one: based on each first data set as a training sample, performing model training by adopting a preset semantic segmentation method to obtain a current semantic segmentation model;

step two: judging whether the precision value of the current model meets a preset precision value or not, and executing the third step when the precision value of the current model is smaller than the preset precision value; repeatedly executing the first step under the condition that the precision value of the current model is larger than or equal to the preset precision value, and obtaining an updated current semantic segmentation model and updated current model parameters;

step three: and detecting the current semantic segmentation model and current model parameters based on a second data set to obtain the semantic segmentation model.

5. The method of claim 3, wherein the acquiring the label image corresponding to each of the history images specifically comprises:

performing image marking processing on each historical image to obtain a tag file corresponding to each historical image, wherein the tag file comprises node area information and marking category information of the historical image;

and carrying out label image making on the label file corresponding to each history image based on the node area information and the label category information corresponding to each history image, and obtaining a label image corresponding to each history image.

6. The method according to claim 1, wherein the detecting the node prediction image based on each of the target contour images as a mask image to obtain a network degree detection result of each of the target spinning specifically includes:

detecting the node prediction images by taking each target contour image as a mask image, and extracting to obtain node images corresponding to target spinning;

and calculating and obtaining the network degree of the target spinning based on the node image corresponding to the target spinning, and obtaining a detection result of the network degree.

7. The method of claim 5, wherein calculating the network degree of the target spinning based on the node image corresponding to the target spinning to obtain a detection result of the network degree specifically comprises:

obtaining the node number corresponding to the target spinning based on the node image corresponding to the target spinning;

calculating and obtaining the actual length of the target spinning based on the number of the pixel points of the node image and the actual distance corresponding to each pixel point;

and calculating and obtaining the network degree of the target spinning based on the number of nodes corresponding to the target spinning and the actual length of the target spinning, and obtaining a detection result of the network degree.

8. A network degree detection device, characterized by comprising:

an image processing module: the method comprises the steps of performing image processing on an image to be tested to obtain a plurality of target contour images corresponding to the to-be-tested attempt, wherein each target contour image comprises a target spinning contour;

and the feature extraction module is used for: the method comprises the steps of carrying out network node characteristic extraction processing on the to-be-tested image based on a preset semantic segmentation model to obtain a node prediction image corresponding to the to-be-tested image;

and a detection module: and the node prediction images are used for detecting the node prediction images by taking the target contour images as mask images, so as to obtain network degree detection results of the target spinning.

9. A storage medium storing a computer program which, when executed by a processor, implements the steps of the network detection method according to any one of the preceding claims 1-7.

10. An electronic device comprising at least a memory, a processor, the memory having stored thereon a computer program, the processor, when executing the computer program on the memory, implementing the steps of the network detection method according to any of the preceding claims 1-7.