Dynamic identification picture and short video output method and device and storage medium

Abstract

The embodiment of the invention discloses a dynamically recognized picture and short video output method and device and a storage medium, and relates to the technical field of graph intelligent recognition in a specific scene, and the method comprises the steps: collecting image data from an input video stream in real time, carrying out the analysis and processing of the image data through employing a gray-scale binary difference method, and carrying out the recognition of a short video; obtaining an annotation target and a short video; the short video comprises the annotation target; in a non-real-time stage, carrying out deep analysis and identification on the labeled target by adopting a YOLO algorithm, and if the identification is successful, storing and outputting the labeled target and the short video; the labeling target is a static picture. By implementing the embodiment of the invention, the dynamic object can be identified and labeled, the target short video and the static picture are finally output, and the motion track and behavior of the target can be more comprehensively reflected, so that the accuracy and reliability of identification are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent graphic recognition in specific scenarios, and in particular to a method, device and storage medium for outputting dynamically recognized pictures and short videos. Background Art

[0002] With the advancement of science and technology, image recognition technology has been widely used in many fields, including but not limited to food safety, road speeding monitoring and other fields. Traditional image recognition methods mainly rely on static images, and identify and classify targets by calculating and saving single-frame image data in real time. However, in the case of tens of thousands of large-scale cameras for supervision or law enforcement, the recognition cost is high and static images cannot fully reflect the authenticity and dynamic changes of objects. Especially in the field of law enforcement, static images and short videos of objects can be directly provided to the person being executed.

[0003] Existing image recognition and output technologies usually rely on static images, while short videos can better reflect the dynamics of objects and improve recognition accuracy and efficiency. The problem with existing technologies is that accurate and real-time recognition of static images requires high computing power, and there is no dynamic video output, which is not authentic enough. Summary of the invention

[0004] In view of the defects in the prior art, the present invention provides a method, device and storage medium for outputting dynamically recognized pictures and short videos.

[0005] To achieve the above objectives, in a first aspect, an embodiment of the present invention provides a method for outputting dynamically recognized pictures and short videos, comprising:

[0006] Collect image data from the input video stream in real time, analyze and process the image data using a grayscale binary difference method, and obtain a marked target and a short video; the short video includes the marked target;

[0007] In the non-real-time stage, the YOLO algorithm is used to perform in-depth analysis and identification of the marked target. If the identification is successful, the marked target and the short video are saved and output; the marked target is a static picture.

[0008] Furthermore, the labeled target and short video are obtained, specifically:

[0009] Receive a video stream input from the outside, and collect image data from the video stream in real time based on a preset collection interval threshold;

[0010] Convert the image data into a grayscale image, and perform binarization processing on the grayscale image to obtain a binary image;

[0011] Performing frame difference calculation based on the binary image to obtain a difference image;

[0012] Based on the differential image, initially screening out suspected target objects;

[0013] Re-identify the suspected target object to obtain a possible target;

[0014] The possible targets are marked to obtain marked targets, and a short video containing the marked targets is temporarily stored.

[0015] Furthermore, the method of initially screening out suspected target objects based on the differential image is specifically as follows:

[0016] extracting a changed portion from the differential image;

[0017] If the changed portion is not located at the designated position, and the pixel size of the changed portion is within the pixel interval, it is determined that the suspected target object is found.

[0018] Furthermore, the YOLO algorithm is used to conduct in-depth analysis and identification of the labeled target, specifically:

[0019] Setting business rules; the business rules include that in the kitchen rat infestation identification scenario, if the identified target is a rat, it means that the identification is successful, otherwise the identification fails;

[0020] In the non-real-time stage, an accurate recognition algorithm model is formed based on the business rules, and the YOLO algorithm is used to recognize the marked target through the accurate recognition algorithm model;

[0021] If the recognition is successful, the marked object is formally marked, and the marked object and the short video are formally saved and output.

[0022] Furthermore, if the YOLO algorithm fails in recognition, the annotation of the marked target is discarded and the short video is deleted.

[0023] In a second aspect, an embodiment of the present invention provides a dynamically recognized picture and short video output device, including:

[0024] An acquisition unit, used for acquiring a video stream input from the outside;

[0025] A primary recognition unit, used to collect image data from the video stream, and analyze and process the image data in real time by using a grayscale binary difference method to obtain a marked target and a short video; the short video includes the marked target;

[0026] The secondary recognition unit is used to use the YOLO algorithm to perform in-depth analysis and recognition of the marked target in a non-real-time stage. If the recognition is successful, the marked target and the short video are saved and output; the marked target is a static picture.

[0027] Furthermore, the primary identification unit is specifically used for:

[0028] Collecting image data from the video stream in real time based on a preset collection interval threshold;

[0029] Converting the image data into a grayscale image, and performing binarization processing on the grayscale image to obtain a binary image;

[0030] Performing frame difference calculation based on the binary image to obtain a difference image;

[0031] Based on the differential image, initially screening out suspected target objects;

[0032] Re-identify the suspected target object to obtain a possible target;

[0033] The possible targets are marked to obtain marked targets, and a short video containing the marked targets is temporarily stored.

[0034] Furthermore, the secondary identification unit is specifically used for:

[0035] Setting business rules; the business rules include that in the kitchen rat infestation identification scenario, if the identified target is a rat, it means that the identification is successful, otherwise the identification fails;

[0036] In the non-real-time stage, an accurate recognition algorithm model is formed based on the business rules, and the YOLO algorithm is used to recognize the marked target through the accurate recognition algorithm model;

[0037] If the recognition is successful, the marked object is formally marked, and the marked object and the short video are formally saved and output.

[0038] In a third aspect, an embodiment of the present invention provides another dynamically recognized image and short video output device, comprising a processor, an input device, an output device and a memory, wherein the processor, input device, output device and memory are interconnected, wherein the memory is used to store a computer program, and the computer program includes program instructions, and the processor is configured to call the program instructions to execute the method of the first aspect above.

[0039] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the processor executes the method of the first aspect.

[0040] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0041] 1. The use of secondary algorithm recognition improves recognition accuracy while saving costs

[0042] In the scenario where real-time identification of target objects is required, the grayscale binary difference method is used to initially screen out suspected target objects, and then annotate them to obtain annotated targets. Later, in the scenario where non-real-time identification of target objects is required, the YOLO algorithm is used to identify the annotated targets again to ensure the accuracy of identification. The different algorithm identification steps in real-time and non-real-time scenarios fully utilize system resources to achieve cost savings when finding the balance between accuracy and computing power cost. Compared with traditional identification methods, this method can also quickly eliminate most of the irrelevant interference data, such as timestamps, floaters, and moving human bodies.

[0043] 2. Introduce short videos to save dynamic recognition results

[0044] After two recognitions, the system starts saving a 5-second short video related to the target object, which helps record the dynamic performance of the target and provides additional information support for subsequent recognition. Compared with the recognition method that only relies on static images, the short video can more comprehensively reflect the target's movement trajectory and behavior, thereby improving the accuracy and reliability of recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for the specific embodiments or the description of the prior art. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn according to the actual scale.

[0046] Figure 1 A schematic diagram of a process flow of a method for outputting dynamically recognized pictures and short videos provided by an embodiment of the present invention;

[0047] Figure 2 To obtain the labeled target and short video process diagram;

[0048] Figure 3 A schematic diagram of the structure of a dynamic recognition image and short video output device provided by an embodiment of the present invention;

[0049] Figure 4 Another structural schematic diagram of a dynamically recognized picture and short video output device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The following embodiments of the technical solution of the present invention are described in detail in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and are therefore only used as examples, and cannot be used to limit the protection scope of the present invention.

[0051] It should be noted that, unless otherwise specified, the technical terms or scientific terms used in this application should have the common meanings understood by those skilled in the art to which the present invention belongs.

[0052] like Figure 1 As shown, the embodiment of the present invention provides a method for outputting dynamically recognized pictures and short videos, which may include the following steps:

[0053] S1, collect image data from the input video stream in real time, analyze and process the image data using grayscale binary difference method, and obtain the marked target and short video; the short video includes the marked target. The specific steps of obtaining the marked target and short video are as follows Figure 2 As shown, including:

[0054] S11, receiving a video stream input from the outside, and collecting image data from the video stream in real time based on a preset collection interval threshold;

[0055] Specifically, one or more video streams input from the outside are received, and image data is collected from the video streams in real time based on a preset collection interval threshold; the interval threshold is to collect picture frames from one or more data streams every 10S, and collect picture frames again after 2S of the last collection.

[0056] S12, converting the image data into a grayscale image, and performing binarization processing on the grayscale image to obtain a binary image;

[0057] Specifically, the image data is converted into a grayscale image; the grayscale image is an image with only brightness; a grayscale binarization threshold T=30 is set, pixels in the grayscale image whose pixel values ​​are greater than the grayscale binarization threshold T are set to white, and pixels whose pixel values ​​are less than or equal to the grayscale binarization threshold T are set to black, thereby obtaining a binary image.

[0058] S13, performing frame difference calculation based on the binary image to obtain a difference image;

[0059] Specifically, the binary image includes a binary image 1 and a binary image 2; the binary image 1 is a binary image converted from a picture frame collected from one or more data streams every 10S; the binary image 2 is a binary image converted from a picture frame collected again 2S after the last collection;

[0060] The binary image 1 and the binary image 2 are calculated by the frame difference method, that is, D(t0)=|B(t0)-B(t2)|, to obtain a difference image; the D(t0) represents the difference image; the B(t0) represents the binary image 1; the B(t2) represents the binary image 2; the frame difference method is to calculate the absolute difference between the two frames of the binary image 1 and the binary image 2.

[0061] S14, initially screening out suspected target objects based on the differential image;

[0062] Specifically, a changed portion is extracted from the differential image; then the designated position is set to be 10% above and 10% below the screen, and the pixel interval is [400, 6400]; if the changed portion is not located in the designated position, and the pixel size of the changed portion is within the pixel interval, it is determined that a suspected target object is found.

[0063] S15, identifying the suspected target object again to obtain a possible target;

[0064] S16, marking the possible targets to obtain marked targets, and temporarily storing the short video containing the marked targets.

[0065] Specifically, the possible target is marked with a red frame to obtain a marked target; the size of the red frame is 10 pixels larger than the side length of the possible target pixel, and the total area is increased by more than 10%; after obtaining the marked target, a saving mechanism will be triggered to temporarily store the marked target in a 5S short video to record the dynamic performance of the marked target in the image.

[0066] It should be noted that the above steps can be understood as:

[0067] (1) Real-time collection

[0068] This stage is the starting point of the entire process. The system continuously collects image data from all input video streams, collecting 2 pictures 2 seconds apart every 10 seconds for each video stream. This step ensures that the system can respond to new information in a timely manner.

[0069] (2) Initial identification and screening

[0070] This is a key step in the recognition process. The system uses the grayscale binary difference method to compare two pictures. It can quickly and efficiently identify the changed parts in the picture and preliminarily screen and discover the target.

[0071] By setting the grayscale binarization threshold (T=30), the grayscale image I(t) is converted into a binary image B(t), and then the difference image D(t0) of two frames with an interval of 2 seconds is calculated. The picture at t0 is used as the background frame to quickly determine the activity area of ​​the potential moving target.

[0072] In order to reduce the interference of changes in the date and time stamps in the image on the recognition results, when the changed part is not located in the upper 10% and lower 10% of the screen, and the pixel size of the changed part is between 20*20=400 points and 80*80=6400 points, the system determines that a suspected target object is found.

[0073] The above values ​​can quickly eliminate most of the irrelevant interference data, such as timestamps, floaters, moving human bodies, etc.

[0074] (3) Positioning target: data labeling + temporary storage of short videos

[0075] After the initial screening and determination of the target object, the system will try to identify possible targets and mark them with red frames. The size of the red frame is 10 pixels larger than the side length of the target pixel, and the total area is increased by more than 10%, ensuring that the subsequent positioning of the target and the second YOLO algorithm recognition will not be interfered with by the red frame.

[0076] At the same time, the system will immediately start the saving mechanism to temporarily save a 5-second short video containing the marked target. The short video recording the dynamic performance of the marked target in the image will provide additional information support in the subsequent recognition and use process.

[0077] S2, in the non-real-time stage, the YOLO algorithm is used to perform in-depth analysis and identification of the labeled target.

[0078] Specifically, business rules are set; the business rules include that in the back kitchen rodent infestation identification scenario, if the identified target is a rat, it means that the identification is successful, otherwise the identification fails; in the non-real-time stage, a precise identification algorithm model is formed based on the business rules, and the YOLO algorithm is used to identify the marked target through the precise identification algorithm model.

[0079] Preferably, the business rules also include but are not limited to the greenhouse planting scenario. If the identified target is thrips or white lice, it means the identification is successful, otherwise the identification fails.

[0080] That is, during the daytime and other periods of idle computing power in the system, a second YOLO algorithm recognition is performed on the area that has been marked in the initial identification and screening, which can eliminate other interference and improve the accuracy of target object recognition.

[0081] S3, determine whether the recognition is successful, if the recognition is successful, execute step S4, otherwise execute step S5.

[0082] S4, if the recognition is successful, the marked object is formally marked, and the marked object and the short video are formally saved and output.

[0083] If the YOLO algorithm successfully recognizes an object, the system will give it a formal label, officially save the related short video, and output the labeled target and short video. The labeled target is a static picture. This step ensures that all valuable images and corresponding short videos are properly saved for future retrieval.

[0084] In addition, after manual review and annotation, the formal recognition results can also be used to reversely tune the deep learning model YOLO algorithm to form a specific and accurate recognition model for this scenario.

[0085] S5: If recognition fails, the annotation of the annotation target is discarded and the short video is deleted.

[0086] For those object targets that the YOLO algorithm cannot identify, the system will delete the relevant images and short videos to avoid the accumulation of useless information and maintain the efficient operation of the system.

[0087] The beneficial effects of the present invention are embodied in:

[0088] 1. Grayscale binary difference method is used for initial screening, effectively improving recognition efficiency

[0089] For the video stream, two pictures are collected every 10 seconds, each 2 seconds apart, and the grayscale binary difference method is used to detect changes. When the changed part is not located in the upper 10% and lower 10% of the screen, and the pixel size of the changed part is between 20*20=400 points and 80*80=6400 points, the system determines that a suspected target object has been found. Compared with traditional recognition methods, this method can quickly exclude most of the irrelevant interference data, such as timestamps, mosquitoes, and moving human bodies.

[0090] 2. Suspected target objects are marked to ensure the speed and accuracy of subsequent identification

[0091] After the initial screening and determination of the target object, the system will mark it with a red frame, the size of which is more than 10 points larger than the target pixel, and the total area is increased by more than 10%. This means that the subsequent YOLO algorithm recognition does not need to re-recognize the entire image, but only needs to recognize the information within the annotation, ensuring speed and accuracy.

[0092] 3. Non-real-time YOLO algorithm recognition effectively improves the accuracy of recognition results

[0093] Based on the business rules set for the scenario of identifying rat infestation in the back kitchen, an accurate recognition algorithm model is formed. Through this model, the YOLO algorithm is used to identify the marked targets, making the results more accurate.

[0094] 4. YOLO algorithm model for identifying rat infestation in the kitchen based on deep learning

[0095] After manually reviewing and labeling the recognition result images over a period of time, the YOLO algorithm of the deep learning model was reversely tuned to form a specific and accurate recognition model for the scenario of identifying rat infestation in the kitchen at night.

[0096] 5. The realization of short video output results effectively reflects the target's movement trajectory and behavior

[0097] After the initial identification and screening, the system temporarily stores a 5-second short video related to the target object, and finally saves the short video after the YOLO algorithm successfully recognizes it. This helps to record the dynamic performance of the target and provides additional information support for subsequent identification. Compared with the recognition method that only relies on static images, short videos can more comprehensively reflect the movement trajectory and behavior of the target, thereby improving the accuracy and reliability of recognition.

[0098] Based on the same inventive concept, an embodiment of the present invention provides a device for outputting dynamically recognized pictures and short videos. Figure 3 As shown, the device comprises:

[0099] An acquisition unit 10 is used to acquire a video stream input from the outside;

[0100] A primary recognition unit 11 is used to collect image data from the video stream, and analyze and process the image data in real time by using a grayscale binary difference method to obtain a marked target and a short video; the short video includes the marked target;

[0101] The secondary recognition unit 12 is used to use the YOLO algorithm to perform in-depth analysis and recognition of the marked target in a non-real-time stage, and if the recognition is successful, save the marked target and the short video.

[0102] The primary identification unit 11 is specifically used for:

[0103] Collecting image data from the video stream in real time based on a preset collection interval threshold;

[0104] Convert the image data into a grayscale image, and perform binarization processing on the grayscale image to obtain a binary image;

[0105] Performing frame difference calculation based on the binary image to obtain a difference image;

[0106] Based on the differential image, initially screening out suspected target objects;

[0107] Re-identify the suspected target object to obtain a possible target;

[0108] The possible targets are marked to obtain marked targets, and a short video containing the marked targets is temporarily stored.

[0109] Specifically, the secondary identification unit 12 is specifically used for:

[0110] Setting business rules; the business rules include that in the kitchen rat infestation identification scenario, if the identified target is a rat, it means that the identification is successful, otherwise the identification fails;

[0111] In the non-real-time stage, an accurate recognition algorithm model is formed based on the business rules, and the YOLO algorithm is used to recognize the marked target through the accurate recognition algorithm model;

[0112] If the recognition is successful, the marked object is formally marked, and the marked object and the short video are formally saved and output.

[0113] Furthermore, if the recognition fails, the annotation of the annotated target is discarded and the short video is deleted.

[0114] Alternatively, if Figure 4 As shown, in another preferred embodiment of the present invention, a dynamically recognized picture and short video output device may include: one or more processors 101, one or more input devices 102, one or more output devices 103 and a memory 104, wherein the processor 101, the input device 102, the output device 103 and the memory 104 are interconnected via a bus 105. The memory 104 is used to store a computer program, wherein the computer program includes program instructions, and the processor 101 is configured to call the program instructions to execute the method of the embodiment of the method for outputting a dynamically recognized picture and short video.

[0115] It should be understood that in the embodiment of the present invention, the processor 101 may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0116] The input device 102 may include a keyboard, etc., and the output device 103 may include a display (LCD, etc.), a speaker, etc.

[0117] The memory 104 may include a read-only memory and a random access memory, and provides instructions and data to the processor 101. A portion of the memory 104 may also include a non-volatile random access memory. For example, the memory 104 may also store information on the device type.

[0118] In a specific implementation, the processor 101, input device 102, and output device 103 described in the embodiment of the present invention can execute the implementation method described in the embodiment of a dynamically recognized picture and short video output method provided in the embodiment of the present invention, which will not be repeated here.

[0119] It should be noted that, for the specific working process of a dynamic recognition image and short video output device, please refer to the aforementioned method embodiment part, which will not be repeated here.

[0120] Furthermore, corresponding to the first embodiment of a method for dynamically identifying pictures and short video output, an embodiment of the present invention also provides a readable storage medium storing a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, they implement: the above-mentioned method for dynamically identifying pictures and short video output.

[0121] The computer-readable storage medium may be an internal storage unit of the client described in the foregoing embodiment, such as a hard disk or memory of the system. The computer-readable storage medium may also be an external storage device of the system, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the system. Further, the computer-readable storage medium may also include both an internal storage unit of the system and an external storage device. The computer-readable storage medium is used to store the computer program and other programs and data required by the system. The computer-readable storage medium may also be used to temporarily store data that has been output or is to be output.

[0122] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0123] In the several embodiments provided in the present application, it should be understood that the disclosed units and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, or it can be an electrical, mechanical or other form of connection.

[0124] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. A method for outputting dynamically recognized pictures and short videos, characterized in that: include: Collect image data from the input video stream in real time, analyze and process the image data using grayscale binary difference method to obtain marked targets and short videos; The short video includes the marked target; In the non-real-time stage, the YOLO algorithm is used to perform in-depth analysis and identification of the marked target. If the identification is successful, the marked target and the short video are saved and output; the marked target is a static picture.

2. The method for outputting dynamically recognized pictures and short videos according to claim 1, characterized in that: Get the marked target and short video, specifically: Receive a video stream input from the outside, and collect image data from the video stream in real time based on a preset collection interval threshold; Convert the image data into a grayscale image, and perform binarization processing on the grayscale image to obtain a binary image; Performing frame difference calculation based on the binary image to obtain a difference image; Based on the differential image, initially screening out suspected target objects; Re-identify the suspected target object to obtain a possible target; The possible targets are marked to obtain marked targets, and a short video containing the marked targets is temporarily stored.

3. The method for outputting dynamically recognized pictures and short videos according to claim 2, characterized in that: The method of initially screening out suspected target objects based on the differential image is specifically as follows: extracting a changed portion from the differential image; If the changed portion is not located at the designated position, and the pixel size of the changed portion is within the pixel interval, it is determined that the suspected target object is found.

4. The method for outputting dynamically recognized pictures and short videos according to claim 1, characterized in that: The YOLO algorithm is used to conduct in-depth analysis and identification of the labeled target, specifically: Setting business rules; the business rules include that in the kitchen rat infestation identification scenario, if the identified target is a rat, it means that the identification is successful, otherwise the identification fails; In the non-real-time stage, an accurate recognition algorithm model is formed based on the business rules, and the YOLO algorithm is used to recognize the marked target through the accurate recognition algorithm model; If the recognition is successful, the marked object is formally marked, and the marked object and the short video are formally saved and output.

5. The method for outputting dynamically recognized pictures and short videos according to claim 4, characterized in that: If the YOLO algorithm fails to recognize, the annotation of the marked target is discarded and the short video is deleted.

6. A dynamic recognition image and short video output device, characterized in that: include: An acquisition unit, used for acquiring a video stream input from the outside; A primary recognition unit, used to collect image data from the video stream, and analyze and process the image data in real time by using a grayscale binary difference method to obtain a marked target and a short video; the short video includes the marked target; The secondary recognition unit is used to use the YOLO algorithm to perform in-depth analysis and recognition of the marked target in a non-real-time stage. If the recognition is successful, the marked target and the short video are saved and output; the marked target is a static picture.

7. A dynamic recognition image and short video output device according to claim 6, characterized in that: The primary identification unit is specifically used for: Collecting image data from the video stream in real time based on a preset collection interval threshold; Convert the image data into a grayscale image, and perform binarization processing on the grayscale image to obtain a binary image; Performing frame difference calculation based on the binary image to obtain a difference image; Based on the differential image, initially screening out suspected target objects; Re-identify the suspected target object to obtain a possible target; The possible targets are marked to obtain marked targets, and a short video containing the marked targets is temporarily stored.

8. The dynamic recognition image and short video output device according to claim 6, characterized in that: The secondary identification unit is specifically used for: Setting business rules; the business rules include that in the kitchen rat infestation identification scenario, if the identified target is a rat, it means that the identification is successful, otherwise the identification fails; In the non-real-time stage, an accurate recognition algorithm model is formed based on the business rules, and the YOLO algorithm is used to recognize the marked target through the accurate recognition algorithm model; If the recognition is successful, the marked object is formally marked, and the marked object and the short video are formally saved and output.

9. A dynamic recognition image and short video output device, characterized in that: The method comprises a processor, an input device, an output device and a memory, wherein the processor, the input device, the output device and the memory are interconnected, wherein the memory is used to store a computer program, the computer program comprises program instructions, and the processor is configured to call the program instructions to execute the method according to any one of claims 1 to 5.

10. A computer-readable storage medium storing a computer program, characterized in that: The computer program comprises program instructions which, when executed by a processor, cause the processor to perform the method according to any one of claims 1 to 5.

Images