CN112686186A - High-altitude parabolic recognition method based on deep learning and related components thereof - Google Patents
High-altitude parabolic recognition method based on deep learning and related components thereof Download PDFInfo
- Publication number
- CN112686186A CN112686186A CN202110006895.XA CN202110006895A CN112686186A CN 112686186 A CN112686186 A CN 112686186A CN 202110006895 A CN202110006895 A CN 202110006895A CN 112686186 A CN112686186 A CN 112686186A
- Authority
- CN
- China
- Prior art keywords
- altitude parabolic
- altitude
- sample data
- picture
- target area
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 62
- 238000013135 deep learning Methods 0.000 title claims abstract description 27
- 238000012549 training Methods 0.000 claims abstract description 39
- 238000012544 monitoring process Methods 0.000 claims abstract description 23
- 230000008569 process Effects 0.000 claims abstract description 11
- 238000012545 processing Methods 0.000 claims description 14
- 238000004590 computer program Methods 0.000 claims description 11
- 230000009467 reduction Effects 0.000 claims description 8
- 238000004519 manufacturing process Methods 0.000 claims description 6
- 238000005516 engineering process Methods 0.000 claims description 5
- 239000011159 matrix material Substances 0.000 claims description 5
- 238000011176 pooling Methods 0.000 claims description 4
- 230000003213 activating effect Effects 0.000 claims description 3
- 238000007906 compression Methods 0.000 claims description 3
- 230000006835 compression Effects 0.000 claims description 3
- 230000006870 function Effects 0.000 claims description 3
- 238000009499 grossing Methods 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 238000001514 detection method Methods 0.000 description 6
- 230000009471 action Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 208000027418 Wounds and injury Diseases 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 235000013361 beverage Nutrition 0.000 description 1
- 238000012790 confirmation Methods 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 238000013144 data compression Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 208000014674 injury Diseases 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000011410 subtraction method Methods 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a high-altitude parabolic recognition method based on deep learning and related components thereof, wherein the method comprises the following steps: acquiring high-altitude parabolic sample data; training a ResNet50 model added with an attention SE module by using high-altitude parabolic sample data to obtain a high-altitude parabolic recognition model; acquiring a surveillance video sequence, acquiring a target area picture in the surveillance video sequence by using a frame difference method, and inputting the target area picture into a high-altitude parabolic identification model to obtain the high-altitude parabolic event probability of a target area; and comparing the probability of the high-altitude parabolic event with a preset threshold, and if the probability of the high-altitude parabolic event is greater than the threshold, judging that the high-altitude parabolic event is the high-altitude parabolic event. The method obtains the probability of the high-altitude parabolic events of the target area in the monitoring video through the high-altitude parabolic recognition model to judge whether the target area has the high-altitude parabolic events, does not need manual examination in the whole process, is more intelligent and real-time, directly obtains the target area pictures from the monitoring video, and is suitable for various environments.
Description
Technical Field
The invention relates to the technical field of public safety, in particular to a high-altitude parabolic recognition method based on deep learning and a related component thereof.
Background
With the rapid development of economic society and the acceleration of urbanization process, the number of modern urban population is increasing, and the safety problem of public places is also followed. In order to prevent the occurrence of safety accidents, a large number of monitoring videos are usually installed in public places for monitoring and preventing emergencies, so that the safety of the public places is guaranteed, and the long-term security of the society is maintained. However, in public places, high altitude parabolas pose a great threat to life safety of the masses, due to their burstiness and contingency. The high-altitude parabolic dish has huge impact force, and a small apple, a beverage bottle and an egg can cause serious injury to people and things. Therefore, the device can detect the high altitude parabola in time and send out early warning, and has great significance for the life and property safety of people.
The existing high-altitude parabolic detection and identification method mainly comprises the following steps:
the method comprises the following steps: preliminarily detecting by a frame difference method to obtain a moving target area, summarizing the drop characteristics of the parabola, judging whether the moving target area accords with the parabola law, and pushing the moving target area to manual examination;
the second method comprises the following steps: by detecting and tracking the moving target, the moving track of the target is analyzed based on a regular mode, and whether the target is a high-altitude parabola or not is judged.
However, the process of high-altitude parabolic detection by the first method cannot be intelligentized and real-time, and the event is pushed to relevant personnel for auditing, so that the speed of detection and identification is influenced, the subjectivity of manual detection is excessively relied on, and the long-time work is not facilitated; and the high-altitude parabolic detection is performed by the second method, so that the identification precision is lacked, and the method is limited by the surrounding environment.
Disclosure of Invention
The invention aims to provide a high-altitude parabolic recognition method based on deep learning and related components thereof, and aims to solve the problems that the existing high-altitude parabolic detection process cannot be intelligentized and real-time and is limited by the surrounding environment.
In a first aspect, an embodiment of the present invention provides a high-altitude parabolic recognition method based on deep learning, which includes:
manufacturing corresponding high-altitude parabolic sample data according to a pre-acquired high-altitude parabolic picture;
training a ResNet50 model added with an attention SE module by using the high-altitude parabolic sample data to obtain a high-altitude parabolic recognition model;
acquiring a surveillance video sequence, acquiring a target area picture in the surveillance video sequence by using a frame difference method, and inputting the target area picture into the high-altitude parabolic identification model to acquire the high-altitude parabolic event probability of a target area;
and comparing the high-altitude parabolic event probability with a preset threshold, and if the high-altitude parabolic event probability is greater than the threshold, judging that the high-altitude parabolic event is the high-altitude parabolic event.
In a second aspect, an embodiment of the present invention provides a high-altitude parabolic recognition system based on deep learning, including:
the sample acquisition unit is used for manufacturing corresponding high-altitude parabolic sample data according to a pre-acquired high-altitude parabolic picture;
the model training unit is used for training a ResNet50 model added with an attention SE module by using the high-altitude parabolic sample data to obtain a high-altitude parabolic recognition model;
the high-altitude parabolic probability obtaining unit is used for obtaining a monitoring video sequence, obtaining a target area picture in the monitoring video sequence by using a frame difference method, and inputting the target area picture into the high-altitude parabolic recognition model to obtain a high-altitude parabolic event probability of a target area;
and the high-altitude parabolic event confirming unit is used for comparing the high-altitude parabolic event probability with a preset threshold value, and if the high-altitude parabolic event probability is greater than the threshold value, judging that the high-altitude parabolic event is the high-altitude parabolic event.
In a third aspect, an embodiment of the present invention further provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor, when executing the computer program, implements the high-altitude parabolic recognition method based on deep learning as described above.
In a fourth aspect, the present invention further provides a computer-readable storage medium, where a computer program is stored, and when executed by a processor, the computer program implements the deep learning-based high-altitude parabola identification method as described above.
The embodiment of the invention provides a high-altitude parabolic recognition method based on deep learning and related components thereof, wherein the method comprises the following steps: manufacturing corresponding high-altitude parabolic sample data according to a pre-acquired high-altitude parabolic picture; training a ResNet50 model added with an attention SE module by using the high-altitude parabolic sample data to obtain a high-altitude parabolic recognition model; acquiring a surveillance video sequence, acquiring a target area picture in the surveillance video sequence by using a frame difference method, and inputting the target area picture into the high-altitude parabolic identification model to acquire the high-altitude parabolic event probability of a target area; and comparing the high-altitude parabolic event probability with a preset threshold, and if the high-altitude parabolic event probability is greater than the threshold, judging that the high-altitude parabolic event is the high-altitude parabolic event. According to the embodiment of the invention, the high-altitude parabolic event probability of the target area in the monitoring video is obtained through the high-altitude parabolic identification model so as to judge whether the target area has the high-altitude parabolic event, the whole process does not need to be audited manually, and is more intelligent and real-time, and the target area picture is directly obtained from the monitoring video, so that the method and the device are suitable for various environments.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a schematic flow chart of a high-altitude parabolic recognition method based on deep learning according to an embodiment of the present invention;
fig. 2 is a schematic block diagram of a high-altitude parabolic recognition system based on deep learning according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It is to be understood that the terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in the specification of the present invention and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.
It should be further understood that the term "and/or" as used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.
Referring to fig. 1, fig. 1 is a schematic flow chart of a high-altitude parabolic recognition method based on deep learning according to an embodiment of the present invention, where the method includes steps S101 to S104:
s101, manufacturing corresponding high-altitude parabolic sample data according to a pre-acquired high-altitude parabolic picture;
in the step, high-altitude parabolic pictures under different scenes are obtained in advance, and corresponding high-altitude parabolic data samples are made. And acquiring a plurality of high-altitude parabolic pictures in each scene to increase the richness of high-altitude parabolic sample data.
In a specific embodiment, the step S101 includes:
acquiring a high-altitude parabolic picture in a monitoring video to serve as the high-altitude parabolic sample data;
and/or capturing a high-altitude parabolic picture from the Internet through a crawler technology, and taking the high-altitude parabolic picture as the high-altitude parabolic sample data.
In this embodiment, the high-altitude parabolic sample data may be obtained from a monitoring video or from the internet. When a high-altitude parabolic object is monitored in a monitoring video, intercepting the section of the monitoring video, and intercepting high-altitude parabolic images of the high-altitude parabolic object at different positions in the section of the monitoring video to serve as high-altitude parabolic object sample data; and directly capturing related high-altitude parabolic pictures from the Internet by a crawler technology to serve as the high-altitude parabolic sample data. The Web crawler technology, i.e., Web crawler, is a program or script that automatically captures Web information according to a certain rule.
In a specific embodiment, the obtaining a high altitude parabolic picture in a surveillance video as the high altitude parabolic sample data includes:
acquiring a high-altitude parabolic off-line video of a monitoring camera under different variable conditions recorded in a test process;
acquiring a target area in the high-altitude parabolic off-line video by adopting a frame difference method, taking an original image of coordinates corresponding to the target area as a foreground, and taking an area with the overlapping degree smaller than 0.3 in the target area as a background;
and training the depth residual error network by utilizing the foreground and the background to obtain a two-classification depth residual error network model.
In this embodiment, surveillance videos shot by multiple surveillance cameras at different angles are obtained, high-altitude parabolic offline videos with high-altitude parabolic events are screened out, then a target area in the high-altitude parabolic offline videos is obtained by using a frame difference method, a foreground of the target area is set as an original image of coordinates corresponding to the target area, a background is set as an area with an overlap degree smaller than 0.3 in the target area, and finally a depth residual error network is trained by using the foreground and the background to obtain a final two-classification depth residual error network model. In this embodiment, the variable may be an environmental variable, such as weather, a cell, or a parabolic floor; but also physical variables such as object size of high altitude parabola, camera position, etc.
In a specific embodiment, the training the depth residual error network by using the foreground and the background to obtain the two-class depth residual error network model includes:
performing data enhancement processing on the foreground and the background, and training a depth residual error network by using the processed foreground and the processed background to obtain a two-classification depth residual error network model; the data enhancement processing comprises picture turning, small-angle rotation, picture brightness changing and picture blurring enhancement.
In this embodiment, before the deep residual error network is trained, data enhancement processing needs to be performed on the foreground and the background, and then the processed foreground and background are used for training to obtain a two-classification deep residual error network model. The adopted data enhancement processing method comprises the following steps: turning the picture, rotating the picture at a small angle, changing the brightness of the picture and enhancing the blurring degree of the picture.
S102, training a ResNet50 model added with an attention SE module by using the high-altitude parabolic sample data to obtain a high-altitude parabolic recognition model;
in the step, a ResNet50 model is selected as a main training model, an SE module with an attention mechanism is added into the ResNet50 model to form an SE-ResNet50 network structure, and the SE-ResNet50 network structure is trained by using the high-altitude parabolic sample data to obtain a high-altitude parabolic recognition model.
In one embodiment, the step S102 includes:
adding the SE module between the convolutional layers of the ResNet50 model to obtain an SE-ResNet50 model;
and training the SE-ResNet50 model by utilizing the high-altitude parabolic sample data based on a preset training strategy to obtain a high-altitude parabolic recognition model.
In this embodiment, the SE module is added between convolutional layers of the ResNet50 model, so that the high-altitude parabolic sample data must be processed by the SE module when being input to the ResNet50 model, and the SE-ResNet50 model is trained according to a preset training strategy. The SE module is used for controlling the importance of the channels through the weight coefficients, so that the network pays attention to the important channels, and an attention mechanism is realized. The training strategy includes one or more of a stepped down learning rate, L2 regularization, label smoothing.
In a specific embodiment, the training the SE-ResNet50 model with the high altitude parabolic sample data based on a preset training strategy to obtain a high altitude parabolic recognition model includes:
inputting the high-altitude parabolic sample data into a global pooling layer of the SE module for compression, inputting the compressed high-altitude parabolic sample data into a first full-connection layer for feature dimension reduction processing, and acquiring channel weight of the first full-connection layer;
activating the high-altitude parabolic sample data subjected to feature dimensionality reduction by a ReLu function, inputting the high-altitude parabolic sample data into a second full-connection layer to obtain the high-altitude parabolic sample data of the original dimensionality, and acquiring channel weight of the second full-connection layer;
and multiplying the first full-connection layer channel weight and the second full-connection layer channel weight by the two-dimensional matrix corresponding to the high-altitude parabolic sample data to obtain a final high-altitude parabolic identification model.
In this embodiment, the high-altitude parabolic sample data is input into the SE-ResNet50 model, compressed sequentially through a global pooling layer of an SE module, feature dimension reduction processing is performed in a first full-connected layer, a dimension is converted into an original dimension in a second full-connected layer, and finally, a two-dimensional matrix corresponding to the high-altitude parabolic sample data is multiplied by the first full-connected layer channel weight and the second full-connected layer channel weight to obtain a final high-altitude parabolic identification model.
S103, acquiring a surveillance video sequence, acquiring a target area picture in the surveillance video sequence by using a frame difference method, and inputting the target area picture into the high-altitude parabolic identification model to obtain the high-altitude parabolic event probability of the target area.
In the step, frame-by-frame subtraction is performed on the obtained surveillance video sequence by using a frame subtraction method to obtain a target area picture containing a target area in the surveillance video sequence, and then the target area picture is input into the high-altitude parabolic identification model to perform event prediction to obtain the high-altitude parabolic event probability of the target area. And if the current monitoring video sequence does not contain the target area, deleting the current monitoring video sequence and acquiring the next monitoring video sequence.
S104, comparing the high altitude parabolic event probability with a preset threshold, and if the high altitude parabolic event probability is larger than the threshold, judging that the high altitude parabolic event is the high altitude parabolic event.
In this step, the high altitude parabolic event probability obtained through the training of the high altitude parabolic recognition model is compared with a preset threshold, if the high altitude parabolic event probability is greater than the threshold, the high altitude parabolic event is determined, and if the high altitude parabolic event probability is less than the threshold, the non-high altitude parabolic event is determined. Wherein the threshold may be set to 0.3, and an event with a high altitude parabolic event probability greater than 0.3 is considered as a high altitude parabolic event. Of course, the threshold value can also be set to other values, and can be flexibly set according to different environments.
Referring to fig. 2, fig. 2 is a schematic block diagram of a deep learning-based high-altitude parabolic recognition system according to an embodiment of the present invention, where the deep learning-based high-altitude parabolic recognition system includes:
a sample obtaining unit 201, configured to produce corresponding high-altitude parabolic sample data according to a high-altitude parabolic picture obtained in advance;
the model training unit 202 is used for training a ResNet50 model added with an attention SE module by using the high-altitude parabolic sample data to obtain a high-altitude parabolic recognition model;
a high-altitude parabolic probability obtaining unit 203, configured to obtain a surveillance video sequence, obtain a target area picture in the surveillance video sequence by using a frame difference method, and input the target area picture into the high-altitude parabolic identification model to obtain a high-altitude parabolic event probability of a target area;
a high-altitude parabolic event confirmation unit 204, configured to compare the high-altitude parabolic event probability with a preset threshold, and if the high-altitude parabolic event probability is greater than the threshold, determine that the high-altitude parabolic event is a high-altitude parabolic event.
In one embodiment, the sample acquiring unit 201 includes:
the first picture acquisition unit is used for acquiring a high-altitude parabolic picture in a monitoring video to serve as the high-altitude parabolic sample data;
and the second picture acquisition unit is used for and/or capturing a high-altitude parabolic picture from the Internet through a crawler technology and taking the high-altitude parabolic picture as the high-altitude parabolic sample data.
In one embodiment, the first picture taking unit includes:
the offline video acquisition unit is used for acquiring a high-altitude parabolic offline video of the monitoring camera under different variable conditions recorded in the test process;
the foreground and background acquisition unit is used for acquiring a target area in the high-altitude parabolic off-line video by adopting a frame difference method, taking an original image of coordinates corresponding to the target area as a foreground, and taking an area with the overlapping degree smaller than 0.3 in the target area as a background;
and the depth residual error network model obtaining unit is used for training the depth residual error network by utilizing the foreground and the background so as to obtain a two-classification depth residual error network model.
In an embodiment, the depth residual network model obtaining unit includes:
the enhancement processing unit is used for carrying out data enhancement processing on the foreground and the background and utilizing the processed foreground and the processed background to train the depth residual error network so as to obtain a two-classification depth residual error network model; the data enhancement processing comprises picture turning, small-angle rotation, picture brightness changing and picture blurring enhancement.
In one embodiment, the model training unit 202 includes:
the SE module adding unit is used for adding the SE module between the convolutional layers of the ResNet50 model to obtain an SE-ResNet50 model;
and the SE-ResNet50 model training unit is used for training the SE-ResNet50 model by utilizing the high-altitude parabolic sample data based on a preset training strategy so as to obtain a high-altitude parabolic recognition model.
In one embodiment, the SE-ResNet50 model training unit includes:
the data compression and dimension reduction processing unit is used for inputting the high-altitude parabolic sample data into a global pooling layer of the SE module for compression, inputting the compressed high-altitude parabolic sample data into a first full-connection layer for feature dimension reduction processing, and acquiring channel weight of the first full-connection layer;
the original dimension data acquisition unit is used for activating the high-altitude parabolic sample data subjected to feature dimension reduction processing by a ReLu function and inputting the high-altitude parabolic sample data into a second full-connection layer to obtain the high-altitude parabolic sample data of the original dimension and acquire channel weight of the second full-connection layer;
and the weight matrix multiplication unit is used for multiplying the first full-connection layer channel weight and the second full-connection layer channel weight by the two-dimensional matrix corresponding to the high-altitude parabolic sample data to obtain a final high-altitude parabolic identification model.
In an embodiment, the training strategy units comprise one or more of a stepwise decreasing learning rate, L2 regularization, label smoothing.
The embodiment of the present invention further provides a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the processor implements the high altitude parabola identification method based on deep learning as described above.
An embodiment of the present invention further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the method for identifying a high altitude parabola based on deep learning as described above is implemented.
The embodiments are described in a progressive manner in the specification, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description. It should be noted that, for those skilled in the art, it is possible to make various improvements and modifications to the present invention without departing from the principle of the present invention, and those improvements and modifications also fall within the scope of the claims of the present invention.
It is further noted that, in the present specification, relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
Claims (10)
1. A high-altitude parabolic recognition method based on deep learning is characterized by comprising the following steps:
manufacturing corresponding high-altitude parabolic sample data according to a pre-acquired high-altitude parabolic picture;
training a ResNet50 model added with an attention SE module by using the high-altitude parabolic sample data to obtain a high-altitude parabolic recognition model;
acquiring a surveillance video sequence, acquiring a target area picture in the surveillance video sequence by using a frame difference method, and inputting the target area picture into the high-altitude parabolic identification model to acquire the high-altitude parabolic event probability of a target area;
and comparing the high-altitude parabolic event probability with a preset threshold, and if the high-altitude parabolic event probability is greater than the threshold, judging that the high-altitude parabolic event is the high-altitude parabolic event.
2. The high-altitude parabolic recognition method based on deep learning of claim 1, wherein the making of corresponding high-altitude parabolic sample data according to a pre-acquired high-altitude parabolic picture comprises:
acquiring a high-altitude parabolic picture in a monitoring video to serve as the high-altitude parabolic sample data;
and/or capturing a high-altitude parabolic picture from the Internet through a crawler technology, and taking the high-altitude parabolic picture as the high-altitude parabolic sample data.
3. The high-altitude parabolic recognition method based on deep learning of claim 2, wherein the obtaining of the high-altitude parabolic picture in the surveillance video as the high-altitude parabolic sample data comprises:
acquiring a high-altitude parabolic off-line video of a monitoring camera under different variable conditions recorded in a test process;
acquiring a target area in the high-altitude parabolic off-line video by adopting a frame difference method, taking an original image of coordinates corresponding to the target area as a foreground, and taking an area with an overlapping degree smaller than a preset threshold value in the target area as a background;
and training the depth residual error network by utilizing the foreground and the background to obtain a two-classification depth residual error network model.
4. The high-altitude parabolic recognition method based on deep learning of claim 3, wherein the training of the depth residual error network with the foreground and the background to obtain a two-class depth residual error network model comprises:
performing data enhancement processing on the foreground and the background, and training a depth residual error network by using the processed foreground and the processed background to obtain a two-classification depth residual error network model; the data enhancement processing comprises picture turning, small-angle rotation, picture brightness changing and picture blurring enhancement.
5. The high-altitude parabolic recognition method based on deep learning of claim 1, wherein the training of a ResNet50 model with an attention SE module by using the high-altitude parabolic sample data to obtain the high-altitude parabolic recognition model comprises:
adding the SE module between the convolutional layers of the ResNet50 model to obtain an SE-ResNet50 model;
and training the SE-ResNet50 model by utilizing the high-altitude parabolic sample data based on a preset training strategy to obtain a high-altitude parabolic recognition model.
6. The high-altitude parabolic recognition method based on deep learning of claim 5, wherein the training of the SE-ResNet50 model with the high-altitude parabolic sample data based on a preset training strategy to obtain a high-altitude parabolic recognition model comprises:
inputting the high-altitude parabolic sample data into a global pooling layer of the SE module for compression, inputting the compressed high-altitude parabolic sample data into a first full-connection layer for feature dimension reduction processing, and acquiring channel weight of the first full-connection layer;
activating the high-altitude parabolic sample data subjected to feature dimensionality reduction by a ReLu function, inputting the high-altitude parabolic sample data into a second full-connection layer to obtain the high-altitude parabolic sample data of the original dimensionality, and acquiring channel weight of the second full-connection layer;
and multiplying the first full-connection layer channel weight and the second full-connection layer channel weight by the two-dimensional matrix corresponding to the high-altitude parabolic sample data to obtain a final high-altitude parabolic identification model.
7. The high altitude parabolic recognition method based on deep learning of claim 5, wherein the training strategy comprises one or more of stepped down learning rate, L2 regularization, label smoothing.
8. A high-altitude parabolic recognition system based on deep learning is characterized by comprising:
the sample acquisition unit is used for manufacturing corresponding high-altitude parabolic sample data according to a pre-acquired high-altitude parabolic picture;
the model training unit is used for training a ResNet50 model added with an attention SE module by using the high-altitude parabolic sample data to obtain a high-altitude parabolic recognition model;
the high-altitude parabolic probability obtaining unit is used for obtaining a monitoring video sequence, obtaining a target area picture in the monitoring video sequence by using a frame difference method, and inputting the target area picture into the high-altitude parabolic recognition model to obtain a high-altitude parabolic event probability of a target area;
and the high-altitude parabolic event confirming unit is used for comparing the high-altitude parabolic event probability with a preset threshold value, and if the high-altitude parabolic event probability is greater than the threshold value, judging that the high-altitude parabolic event is the high-altitude parabolic event.
9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing a deep learning based high altitude parabolic recognition method according to any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, characterized in that a computer program is stored thereon, which, when being executed by a processor, implements a deep learning based high altitude parabolic recognition method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110006895.XA CN112686186A (en) | 2021-01-05 | 2021-01-05 | High-altitude parabolic recognition method based on deep learning and related components thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110006895.XA CN112686186A (en) | 2021-01-05 | 2021-01-05 | High-altitude parabolic recognition method based on deep learning and related components thereof |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112686186A true CN112686186A (en) | 2021-04-20 |
Family
ID=75457279
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110006895.XA Pending CN112686186A (en) | 2021-01-05 | 2021-01-05 | High-altitude parabolic recognition method based on deep learning and related components thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112686186A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110796087A (en) * | 2019-10-30 | 2020-02-14 | 江西赣鄱云新型智慧城市技术研究有限公司 | Method and system for quickly generating high-altitude parabolic training sample |
| CN113139478A (en) * | 2021-04-27 | 2021-07-20 | 广东博智林机器人有限公司 | High-altitude parabolic detection method and device, electronic equipment and storage medium |
| CN115147450A (en) * | 2022-09-05 | 2022-10-04 | 中印云端(深圳)科技有限公司 | Moving target detection method and detection device based on motion frame difference image |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110543879A (en) * | 2019-08-20 | 2019-12-06 | 高新兴科技集团股份有限公司 | SSD target detection method based on SE module and computer storage medium |
| CN110929577A (en) * | 2019-10-23 | 2020-03-27 | 桂林电子科技大学 | Improved target identification method based on YOLOv3 lightweight framework |
| CN111401177A (en) * | 2020-03-09 | 2020-07-10 | 山东大学 | End-to-end behavior recognition method and system based on adaptive space-time attention mechanism |
| CN111476131A (en) * | 2020-03-30 | 2020-07-31 | 北京微播易科技股份有限公司 | Video processing method and device |
| CN111507199A (en) * | 2020-03-25 | 2020-08-07 | 杭州电子科技大学 | Method and device for detecting mask wearing behavior |
| CN111723654A (en) * | 2020-05-12 | 2020-09-29 | 中国电子系统技术有限公司 | High-altitude parabolic detection method and device based on background modeling, YOLOv3 and self-optimization |
-
2021
- 2021-01-05 CN CN202110006895.XA patent/CN112686186A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110543879A (en) * | 2019-08-20 | 2019-12-06 | 高新兴科技集团股份有限公司 | SSD target detection method based on SE module and computer storage medium |
| CN110929577A (en) * | 2019-10-23 | 2020-03-27 | 桂林电子科技大学 | Improved target identification method based on YOLOv3 lightweight framework |
| CN111401177A (en) * | 2020-03-09 | 2020-07-10 | 山东大学 | End-to-end behavior recognition method and system based on adaptive space-time attention mechanism |
| CN111507199A (en) * | 2020-03-25 | 2020-08-07 | 杭州电子科技大学 | Method and device for detecting mask wearing behavior |
| CN111476131A (en) * | 2020-03-30 | 2020-07-31 | 北京微播易科技股份有限公司 | Video processing method and device |
| CN111723654A (en) * | 2020-05-12 | 2020-09-29 | 中国电子系统技术有限公司 | High-altitude parabolic detection method and device based on background modeling, YOLOv3 and self-optimization |
Non-Patent Citations (3)
| Title |
|---|
| JIE HU ET AL: "Squeeze-and-Excitation_Networks", 《IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE》 * |
| 赵春晖等: "《视频图像运动目标分析》", 30 June 2011, 国防工业出版社 * |
| 陈敏杰: "基于深度学习的视频中人体行为识别算法研究", 《中国优秀博硕士学位论文全文数据库(硕士) 信息科技辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110796087A (en) * | 2019-10-30 | 2020-02-14 | 江西赣鄱云新型智慧城市技术研究有限公司 | Method and system for quickly generating high-altitude parabolic training sample |
| CN113139478A (en) * | 2021-04-27 | 2021-07-20 | 广东博智林机器人有限公司 | High-altitude parabolic detection method and device, electronic equipment and storage medium |
| CN115147450A (en) * | 2022-09-05 | 2022-10-04 | 中印云端(深圳)科技有限公司 | Moving target detection method and detection device based on motion frame difference image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111754498B (en) | Conveyor belt carrier roller detection method based on YOLOv3 | |
| CN112686186A (en) | High-altitude parabolic recognition method based on deep learning and related components thereof | |
| KR102195706B1 (en) | Method and Apparatus for Detecting Intruder | |
| CN112464846B (en) | Automatic recognition method for abnormal faults of freight train carriage of station | |
| CN113963301A (en) | Space-time feature fused video fire and smoke detection method and system | |
| CN116563762A (en) | Fire detection method, system, medium, equipment and terminal for oil and gas station | |
| CN112597928B (en) | Event detection method and related device | |
| CN117274881A (en) | Semi-supervised video fire detection method based on consistency regularization and distribution alignment | |
| EP2447912B1 (en) | Method and device for the detection of change in illumination for vision systems | |
| CN114359594B (en) | Scene matching method and device, electronic equipment and storage medium | |
| CN109684982B (en) | Flame detection method based on video analysis and combined with miscible target elimination | |
| CN116910491B (en) | Lightning monitoring and early warning system and method, electronic equipment and storage medium | |
| CN112967335A (en) | Bubble size monitoring method and device | |
| CN116665016B (en) | Single-frame infrared dim target detection method based on improved YOLOv5 | |
| CN116403162B (en) | Airport scene target behavior recognition method and system and electronic equipment | |
| US20230360402A1 (en) | Video-based public safety incident prediction system and method therefor | |
| CN113450385B (en) | Night work engineering machine vision tracking method, device and storage medium | |
| CN115797770A (en) | Continuous image target detection method, system and terminal considering relative movement of target | |
| CN111145558B (en) | Illegal behavior identification method based on high-point video monitoring | |
| CN115909144A (en) | Method and system for detecting abnormity of surveillance video based on counterstudy | |
| Padmaja et al. | Crowd abnormal behaviour detection using convolutional neural network and bidirectional LSTM | |
| CN114119531A (en) | Fire detection method and device applied to campus smart platform and computer equipment | |
| CN116994201B (en) | Method for tracing and monitoring high-altitude parabolic objects and computing equipment | |
| CN116156149B (en) | Detection method and device for detecting camera movement | |
| CN111144248B (en) | People counting method, system and medium based on ST-FHCD network model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information |
Address after: Room 801, building 2, Shenzhen new generation industrial park, 136 Zhongkang Road, Meidu community, Meilin street, Futian District, Shenzhen, Guangdong 518000 Applicant after: China Resources Digital Technology Co.,Ltd. Address before: Room 801, building 2, Shenzhen new generation industrial park, 136 Zhongkang Road, Meidu community, Meilin street, Futian District, Shenzhen, Guangdong 518000 Applicant before: Runlian software system (Shenzhen) Co.,Ltd. |
|
| CB02 | Change of applicant information |