Video analysis method, device and computer equipment

Abstract

The present invention discloses a video analysis method, device and computer equipment, which are executed by nodes corresponding to a target area, and the method includes: obtaining data information collected by at least one camera device in the target area, the data information including identification information, time information, and video information; based on the identification information and time information, selecting a target algorithm for identifying the video information from a preset model algorithm set; performing image extraction on the video information according to a preset frequency to obtain image information of a preset number of frames; using the target algorithm to identify each frame of image information to obtain an identification result corresponding to each frame of image information; and determining whether there is abnormal information in the video information according to the identification result. In this way, the waste of network resources is avoided, the problem of insufficient computing resources is solved, and the recognition accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and in particular to a video analysis method, device and computer equipment. Background Art

[0002] With the increasing application of deep learning and computer vision technologies in smart cities, smart transportation, intelligent monitoring and security, the demand for computing resources has also shown an exponential growth. Existing video analysis is completed by centralized processing in a large computing server shared by multiple communities to analyze video streams. This process consumes a lot of network resources and poses a huge challenge to the computing resources of the server. The network conditions and computing power of devices in the community vary, and the formats and specifications of the transmitted data are also different. Centralized cluster video analysis can no longer meet the rapid response requirements of community safety and monitoring tasks. Summary of the invention

[0003] Therefore, in order to solve the deficiencies of the prior art, the embodiments of the present invention provide a video analysis method, apparatus and computer device.

[0004] According to a first aspect, an embodiment of the present invention discloses a video analysis method, which is performed by a node corresponding to a target area, and the method includes:

[0005] Acquire data information collected by at least one camera device in the target area, the data information including identification information, time information, and video information;

[0006] Based on the identification information and the time information, a target algorithm for identifying the video information is selected from a preset model algorithm set;

[0007] Extracting images from the video information according to a preset frequency to obtain image information of a preset number of frames;

[0008] Using the target algorithm to identify each frame of image information, obtaining the identification result corresponding to each frame of image information;

[0009] Based on the recognition result, determine whether there is abnormal information in the video information.

[0010] Optionally, determining whether there is abnormal information in the video information according to the recognition result specifically includes:

[0011] If it is determined that there is abnormal information in the video information based on at least two recognition results, prompt information is generated.

[0012] Optionally, the method further includes: if each recognition result is inconsistent, determining that there is no abnormal information in the video information.

[0013] Optionally, based on the identification information and the time information, a target algorithm for identifying the video information is selected from a preset model algorithm set, specifically including:

[0014] Determine a first algorithm subset from a preset model algorithm set according to the identification information, wherein the first algorithm subset includes at least one candidate algorithm;

[0015] According to the time information, a target algorithm is determined from the first algorithm subset. Optionally,

[0016] Optionally, before identifying each frame of image information using a target algorithm to obtain a recognition result corresponding to each frame of image information, the method further includes:

[0017] According to the target algorithm, each frame of image information in the video information is preprocessed to obtain each frame of image information after preprocessing, which is used to input into the target algorithm for recognition.

[0018] Optionally, before using the target algorithm to identify each image information to obtain a preset number of identification results corresponding to the image information, the method further includes:

[0019] Train the target algorithm based on pre-acquired labeled samples;

[0020] When the target algorithm reaches the preset accuracy, the training is stopped so that the trained target algorithm can be used to recognize the image information.

[0021] Optionally, when the camera device includes multiple cameras, the method further includes:

[0022] Using the target algorithm, each frame of image information in the video information of each camera device is recognized respectively to obtain a recognition result corresponding to the image information;

[0023] If the recognition result corresponding to the image information of at least one camera device includes the object to be recognized, then record the identification information and time information of the image information where the object to be recognized is located;

[0024] Sorting all the image information containing the object to be identified according to the time information of all the image information containing the object to be identified, and obtaining a sorting result;

[0025] According to the identification information in the sorting result, the moving path of the object to be identified is determined.

[0026] According to a second aspect, an embodiment of the present invention further discloses a video analysis device, which is arranged in a node corresponding to a target area, and the device includes:

[0027] An acquisition module, used to acquire data information collected by at least one camera device in the target area, the data information including identification information, time information, and video information;

[0028] A selection module, for selecting a target algorithm for identifying video information from a preset model algorithm set based on identification information and time information;

[0029] An extraction module is used to extract images from video information according to a preset frequency to obtain image information of a preset number of frames;

[0030] The recognition module is used to recognize each frame of image information by using a target algorithm to obtain a recognition result corresponding to each frame of image information;

[0031] The determination module is used to determine whether there is abnormal information in the video information according to the recognition result.

[0032] According to the third aspect, an embodiment of the present invention further discloses a computer device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor performs the steps of the video analysis method of the first aspect or any optional implementation manner of the first aspect.

[0033] According to a fourth aspect, an embodiment of the present invention further discloses a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the video analysis method of the first aspect or any optional embodiment of the first aspect.

[0034] The technical solution of the present invention has the following advantages:

[0035] The video analysis method provided by the present invention is executed by a node corresponding to a target area, and the method includes: obtaining data information collected by at least one camera device in the target area, wherein the data information includes identification information, time information and video information; selecting a corresponding target algorithm from a preset model algorithm set to identify the video information according to the identification information and the time information; performing image extraction on the video information according to a preset frequency to obtain image information of a preset number of frames, using the target algorithm to identify each piece of image information, and determining whether there is abnormal information in the corresponding video information according to each identification result.

[0036] By acquiring data information collected by at least one camera device in the target area through the node corresponding to the target area, and identifying and judging the data information, it avoids the waste of network resources caused by sending the data information in the target area to the central cluster server, and also solves the problem of allocation and shortage of computing resources; further, according to the identification information and time information in the data information, the target algorithm that best matches the video information can be accurately selected from the preset model algorithm set; thereby, the image information of a preset number of frames in the video information is identified by the target algorithm, and the recognition result of the preset number of frames is obtained, thereby avoiding the recognition of all the image information in the video information, and reducing the waste of computing resources while ensuring accuracy; according to the recognition result of the preset number of frames, it is determined whether there is abnormal information in the video information, thereby improving the recognition accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0038] Figure 1 is a flowchart of a specific example of a video analysis method in an embodiment of the present invention;

[0039] Figure 2 is a flowchart of a specific example of a video analysis method in an embodiment of the present invention;

[0040] Figure 3 is a flowchart of a specific example of a video analysis method in an embodiment of the present invention;

[0041] Figure 4 is a flowchart of a specific example of a video analysis method in an embodiment of the present invention;

[0042] Figure 5 is a schematic diagram of a specific example of a video analysis method in an embodiment of the present invention;

[0043] Figure 6 is a principle block diagram of a specific example of a video analysis device in an embodiment of the present invention;

[0044] Figure 7 FIG. 4 is a specific example diagram of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0047] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, it can also be the internal connection of two components, it can be a wireless connection, or it can be a wired connection. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0048] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0049] In order to solve the problem of network resource waste and insufficient computing power in the central cluster when processing video analysis in each community, a node (also called an edge computing node) is set up in each community to complete the recognition and analysis of the videos of the corresponding community within the node.

[0050] In view of the technical problems mentioned in the background technology, the present application embodiment provides a video analysis method, which is specifically referred to as Figure 1 To solve the problem in the background technology, a node is set in each community to complete the video recognition and analysis in the corresponding community. The node has a built-in recognition algorithm for the data information collected by all the cameras in the community, so that the data information can be directly recognized and analyzed according to the recognition algorithm in the node.

[0051] Figure 1A video analysis method provided by an embodiment of the present invention is performed by a node corresponding to a target area, that is, the node set in the corresponding community (target area) mentioned above, and the method includes the following steps:

[0052] Step 101: Acquire data information collected by at least one camera device in a target area.

[0053] The data information includes identification information, time information, and video information.

[0054] Exemplarily, the target area may be a community, square, supermarket, shopping mall, etc., where video analysis is required. At least one camera device is provided in the target area to collect data information in the corresponding target area, wherein the data information includes identification information, time information, and video information. The identification information is the area location information collected by the camera device, or may be the installation location information of the camera device. The time information is the time information of the data information collected by the camera device. The embodiment of the present invention does not limit the type of the target area, and those skilled in the art may determine it according to actual needs.

[0055] Step 102: Select a target algorithm for identifying video information from a preset model algorithm set based on the identification information and the time information.

[0056] Exemplarily, based on the identification information and time information of the data information, the most appropriate model algorithm is selected from a preset model algorithm set to identify the corresponding video information.

[0057] Among them, the preset model algorithm set is the model algorithm set and stored in the node. When storing the algorithm, each model algorithm can be compressed and calibrated before storage, thereby reducing the occupation of the node memory by the model algorithm.

[0058] In a specific embodiment, the compression of the model algorithm can use a quantization algorithm, inter-layer fusion and operator optimization method to compress and speed up the trained model. The compressed and accelerated model algorithm may reduce the recognition accuracy, and the compressed and accelerated model algorithm needs to be calibrated. The calibration method can use a small amount of image data from multiple scenes to train and calibrate the compressed and accelerated model algorithm.

[0059] Optionally, the method for training and calibrating the compressed and accelerated model algorithm can be implemented by the following steps:

[0060] Train the target algorithm based on pre-acquired labeled samples;

[0061] When the target algorithm reaches the preset accuracy, the training is stopped so that the trained target algorithm can be used to recognize the image information.

[0062] Alternatively, if Figure 2 As shown, the specific process of selecting the target algorithm can be achieved through the following steps:

[0063] Step 1021: Determine a first algorithm subset from a preset model algorithm set according to the identification information.

[0064] The first algorithm subset includes at least one candidate algorithm.

[0065] Step 1022: Determine a target algorithm from the first algorithm subset according to the time information.

[0066] Exemplarily, among all the preset model algorithms, each model algorithm can explain the input data of the model algorithm, the file address of the storage of the model algorithm, the training accuracy, etc. by writing a configuration file, so that it can be directly called during the use of the model algorithm.

[0067] Each camera device collects different types of data information, and the content of the data information that needs to be identified is also different, so each camera device corresponds to a different model algorithm. Therefore, choosing a suitable target algorithm is an important guarantee for accurate data information identification.

[0068] In a specific embodiment, when selecting a model algorithm, a model algorithm set corresponding to the camera device is selected according to the identification information of the camera device. For example, if the camera device is set in an elevator, the selected model algorithm is an algorithm related to whether the elevator is faulty and whether someone is occupying the elevator.

[0069] Within the time span of the data information collected by the camera device, different time periods need to focus on judging and identifying different abnormal problems. In the set of selected model algorithms, the model algorithm that needs to be identified in each time period is selected as the target algorithm. For example, during the morning rush hour, judging whether someone occupies the elevator is the point that needs to be focused on. In other time periods, it is possible to cross-check whether someone occupies the elevator or whether the elevator is faulty. Of course, if the computing power is sufficient, the morning rush hour time period can simultaneously judge whether the elevator is occupied and whether the elevator is faulty. The embodiment of the present invention does not limit the specific selection type of the model algorithm, and those skilled in the art can determine it according to actual needs.

[0070] Step 103: extracting images from the video information according to a preset frequency to obtain image information of a preset number of frames.

[0071] Exemplarily, in the process of identifying video information, the video information is framed at a preset frequency, and the image information after frame extraction is identified. This not only reduces the computing pressure of the node computing power, but also ensures the accuracy of recognition. The preset frequency can be pre-set in the configuration file of the model algorithm, or can be pre-set in the time information of the camera device. The embodiment of the present invention does not limit the size of the preset frequency and the configuration method of the preset frequency, and those skilled in the art can determine it according to actual needs.

[0072] Step 104, using the target algorithm to identify each frame of image information, and obtaining a recognition result corresponding to each frame of image information.

[0073] Step 105: Determine whether there is abnormal information in the video information based on the recognition result.

[0074] Exemplarily, after identifying each frame of image information obtained by frame extraction according to the target algorithm, a corresponding identification result is obtained. It is determined whether there is abnormal information in this event segment based on all the identification results.

[0075] In a specific embodiment, Figure 3 As shown, the process of judging whether there is abnormal information according to the recognition result is implemented by the following steps:

[0076] Step 1051: If it is determined that there is abnormal information in the video information based on at least two recognition results, a prompt message is generated.

[0077] Step 1052: If each recognition result is inconsistent, it is determined that there is no abnormal information in the video information.

[0078] Exemplarily, according to each recognition result, within the time sequence of the corresponding time period, if all the recognition results are consistent, then the abnormal information is output. If there is at least one recognition result that is inconsistent, then no abnormal information is output.

[0079] For example, when judging whether there is any unauthorized use of the elevator, if it is detected for 3 consecutive seconds that a person or a car is blocking the elevator door so that the elevator door cannot be closed, abnormal information of the unauthorized use of the elevator is output, rendered on the video information interface, and a target box or text of the abnormal information is displayed; if within 3 seconds, there is a time period in the middle during which the elevator is not unauthorized use, abnormal information of the unauthorized use of the elevator is not output, and the elevator is operating normally at this time, and no abnormal information exists.

[0080] On the basis of the above-mentioned embodiments, the embodiments of the present invention further provide another video analysis method. In this embodiment, the contents introduced in the above-mentioned embodiments will not be repeated. In this embodiment, considering that there are multiple cameras, the corresponding model algorithm will be determined in the node according to the data information of each camera. After the model algorithm is determined, all the video information in all the data information will be identified in parallel.

[0081] Before inputting the corresponding image information into the model algorithm, since each model algorithm has different requirements for input data, the image information needs to be preprocessed according to the requirements of the input data.

[0082] Optionally, each frame of image information in the video information is preprocessed according to the target algorithm to obtain each frame of image information after preprocessing, which is input into the target algorithm for recognition.

[0083] Exemplarily, the image information obtained by frame extraction can be processed by a preprocessing plug-in. According to the input data requirements of the model algorithm, operations such as scaling, maintaining aspect ratio, cropping, etc. can be performed on the image information. The preprocessing process varies depending on the model algorithm. The embodiment of the present invention does not limit the preprocessing method, and those skilled in the art can determine it according to actual needs.

[0084] In another optional embodiment, when there are multiple camera devices and it is necessary to track the target object in the data information collected by the camera devices, see Figure 4 As shown, the moving path of the target object is determined in the following manner.

[0085] Step 401 , using a target algorithm, recognize each frame of image information in the video information of each camera device, and obtain a recognition result corresponding to the image information.

[0086] Step 402: If the recognition result corresponding to the image information of at least one camera device includes the object to be recognized, the identification information and time information of the image information where the object to be recognized is located are recorded.

[0087] Step 403: sort all the image information containing the objects to be identified according to the time information of all the image information containing the objects to be identified, and obtain a sorting result.

[0088] Step 405: Determine the moving path of the object to be identified according to the identification information in the sorting result.

[0089] For example, when tracking a target object and determining its moving path, it is necessary to identify the video information in all the camera devices using the target recognition algorithm, and save the identification information and time information corresponding to the video information of the identified target object. All the time information is sorted in chronological order, and the moving path of the target object can be determined according to the identification information corresponding to the sorting result, so that the target object can be tracked across the camera devices.

[0090] The video analysis method is supplemented with a specific embodiment. Figure 5 As shown, multiple video streams are accessed to the video information in the data information collected by multiple cameras in the above embodiment; the round-robin rule configuration is a pre-set configuration file, which specifies which model algorithm is used to identify the video information collected by each camera in each time period; the round-robin configuration analysis is to parse the configuration file configured by the round-robin rule, and match the input multiple video streams with the model algorithm; after the model algorithm is matched, the matched model algorithm is loaded from the node memory, wherein the model algorithm stored in the node memory needs to be compressed, integrated and retrained to reduce memory usage and improve the accuracy of the model algorithm, and a configuration file is written for each model algorithm input data, training accuracy, address of the model algorithm and other information, thereby improving the calling efficiency of the model algorithm.

[0091] After the model algorithm is matched, multiple video streams are calculated in parallel according to the matched model algorithm. The calculation process can be: single-frame decoding of the video stream, and preprocessing of the decoded image according to the matched model algorithm, so that the image data meets the data standard of the model algorithm, further increasing the recognition accuracy; the preprocessed image data is input into the model algorithm to obtain a structured recognition result, and the recognition result is transmitted to the node backend through the kafka middleware for parsing. Within the time range, the multi-pin structured result sequence is integrated, and the abnormal information is output according to the actual situation of the video stream.

[0092] By executing this method, data information collected by at least one camera device in the target area is obtained through the node corresponding to the target area, and the data information is identified and judged, thereby avoiding the waste of network resources caused by sending the data information in the target area to the central cluster server, and also solving the congestion and shortage of computing resources; according to the identification information and time information in the data information, the target algorithm that best matches the video information can be accurately selected from the preset model algorithm set; the image information of a preset number of frames in the video information is identified by the target algorithm, and the recognition result of the preset number of frames is obtained, thereby avoiding the recognition of all the image information in the video information, while ensuring accuracy, reducing the waste of computing resources; according to the recognition result of the preset number of frames, it is determined whether there is abnormal information in the video information, thereby improving the recognition accuracy.

[0093] The above is an embodiment of the video analysis method provided by the present application. The following is an introduction to other embodiments of the video analysis method provided by the present application. Please refer to the following for details.

[0094] The embodiment of the present invention also discloses a video analysis device, such as Figure 6 As shown, the device comprises:

[0095] The acquisition module 601 is used to acquire data information collected by at least one camera device in the target area.

[0096] The data information includes identification information, time information, and video information.

[0097] The selection module 602 is used to select a target algorithm for identifying video information from a preset model algorithm set based on the identification information and the time information.

[0098] The extraction module 603 is used to extract images from the video information according to a preset frequency to obtain image information of a preset number of frames.

[0099] The recognition module 604 is used to recognize each frame of image information by using a target algorithm to obtain a recognition result corresponding to each frame of image information.

[0100] The determination module 605 is used to determine whether there is abnormal information in the video information according to the recognition result.

[0101] As an optional implementation manner of the present invention, the determination module 605 specifically includes:

[0102] The first determination submodule is used to generate prompt information if it is determined that there is abnormal information in the video information based on at least two recognition results.

[0103] As an optional implementation manner of the present invention, the determination module 605 further includes:

[0104] The second determination submodule is used to determine that there is no abnormal information in the video information if each recognition result is inconsistent.

[0105] As an optional implementation of the present invention, the selection module 602 specifically includes:

[0106] A first selection submodule, configured to determine a first algorithm subset from a preset model algorithm set according to the identification information, wherein the first algorithm subset includes at least one candidate algorithm;

[0107] The second selection submodule is used to determine a target algorithm from the first algorithm subset according to the time information.

[0108] As an optional implementation of the present invention, before the identification module 604, the device further includes:

[0109] The preprocessing module is used to preprocess each frame of image information in the video information according to the target algorithm, and obtain each frame of image information after preprocessing for input into the target algorithm for recognition.

[0110] As an optional implementation of the present invention, before the identification module 604, the device further includes:

[0111] A training submodule, used to train the target algorithm based on pre-acquired labeled samples;

[0112] The correction submodule is used to stop training when the target algorithm reaches a preset accuracy, so as to use the trained target algorithm to recognize image information.

[0113] As an optional embodiment of the present invention, when the camera device includes multiple devices, the device also includes:

[0114] The recognition submodule is used to use the target algorithm to recognize each frame of image information in the video information of each camera device, and obtain a recognition result corresponding to the image information;

[0115] a recording submodule, configured to record identification information and time information of the image information where the object to be identified is located if the identification result corresponding to the image information of at least one camera device includes the object to be identified;

[0116] A sorting submodule, used to sort all the image information containing the objects to be identified according to the time information of all the image information containing the objects to be identified, and obtain the sorting result;

[0117] The determination submodule is used to determine the moving path of the object to be identified according to the identification information in the sorting result.

[0118] The functions performed by the various components in the video analysis device provided in the embodiment of the present invention have been described in detail in any of the above method embodiments, and therefore will not be repeated here.

[0119] By executing the device, data information collected by at least one camera device in the target area is obtained through the node corresponding to the target area, and the data information is identified and judged, thereby avoiding the waste of network resources caused by sending the data information in the target area to the central cluster server, and also solving the congestion and shortage of computing resources; according to the identification information and time information in the data information, the target algorithm that best matches the video information can be accurately selected from the preset model algorithm set; the image information of a preset number of frames in the video information is identified by the target algorithm, and the recognition result of the preset number of frames is obtained, thereby avoiding the recognition of all the image information in the video information, while ensuring accuracy, reducing the waste of computing resources; according to the recognition result of the preset number of frames, it is determined whether there is abnormal information in the video information, thereby improving the recognition accuracy.

[0120] The embodiment of the present invention also provides a computer device, such as Figure 7 As shown, the computer device may include a processor 701 and a memory 702, wherein the processor 701 and the memory 702 may be connected via a bus or other means. Figure 7 The example of the video analysis method is shown in FIG.

[0121] The processor 701 may be a central processing unit (CPU). The processor 701 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, or a combination of the above chips.

[0122] The memory 702 is a non-transitory computer-readable storage medium that can be used to store non-transitory software programs, non-transitory computer executable programs and modules, such as program instructions / modules corresponding to the video analysis method in the embodiment of the present invention. The processor 701 executes various functional applications and data processing of the processor by running the non-transitory software programs, instructions and modules stored in the memory 702, that is, implementing the video analysis method in the above method embodiment.

[0123] The memory 702 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required by at least one function; the data storage area may store data created by the processor 701, etc. In addition, the memory 702 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 702 may optionally include a memory remotely arranged relative to the processor 701, and these remote memories may be connected to the processor 701 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0124] One or more modules are stored in the memory 702, and when executed by the processor 701, the execution is as follows: Figure 1 The video analysis method in the illustrated embodiment.

[0125] For details of the above computer equipment, please refer to Figure 1 The corresponding related descriptions and effects in the illustrated embodiments can be understood and will not be repeated here.

[0126] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the storage medium can be a disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive (HDD) or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above-mentioned types of memory.

[0127] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A video analysis method, It is characterized in that Executed by a node corresponding to the target area, the method includes: Acquire data information collected by at least one camera device in the target area, wherein the data information includes identification information, time information, and video information; Based on the identification information and the time information, selecting a target algorithm for identifying the video information from a preset model algorithm set; Extracting images from the video information according to a preset frequency to obtain a preset number of frames of image information; Using the target algorithm to identify the image information of each frame, and obtaining a recognition result corresponding to the image information of each frame; According to the recognition result, it is determined whether there is abnormal information in the video information.

2. The method according to claim 1, It is characterized in that The determining, according to the recognition result, whether there is abnormal information in the video information specifically includes: If it is determined that abnormal information exists in the video information based on at least two of the recognition results, prompt information is generated.

3. The method according to claim 2, It is characterized in that The method further comprises: If each recognition result is inconsistent, it is determined that there is no abnormal information in the video information.

4. The method according to any one of claims 1 to 3, It is characterized in that The selecting, based on the identification information and the time information, a target algorithm for identifying the video information from a preset model algorithm set specifically includes: Determine a first algorithm subset from the preset model algorithm set according to the identification information, wherein the first algorithm subset includes at least one candidate algorithm; The target algorithm is determined from the first algorithm subset according to the time information.

5. The method according to any one of claims 1 to 3, It is characterized in that Before using the target algorithm to identify each frame of the image information and obtaining the identification result corresponding to each frame of the image information, the method further includes: According to the target algorithm, each frame of image information in the video information is preprocessed to obtain each frame of image information after preprocessing, which is used to input into the target algorithm for recognition.

6. The method according to any one of claims 1 to 3, It is characterized in that Before using the target algorithm to identify each of the image information to obtain a preset number of identification results corresponding to the image information, the method further includes: Training the target algorithm based on pre-acquired labeled samples; When the target algorithm reaches a preset accuracy, the training is stopped to use the trained target algorithm to recognize the image information.

7. The method according to any one of claims 1 to 3, It is characterized in that When the camera device includes multiple cameras, the method further includes: Using the target algorithm, each frame of image information in the video information of each camera device is respectively identified to obtain an identification result corresponding to the image information; If the recognition result corresponding to the image information of the at least one camera device includes an object to be recognized, recording identification information and time information of the image information where the object to be recognized is located; Sorting all the image information containing the object to be identified according to the time information of all the image information containing the object to be identified, and obtaining a sorting result; The moving path of the object to be identified is determined according to the identification information in the sorting result.

8. A video analysis device, It is characterized in that The device comprises: An acquisition module, used to acquire data information collected by at least one camera device in a target area, wherein the data information includes identification information, time information, and video information; A selection module, configured to select a target algorithm for identifying the video information from a preset model algorithm set based on the identification information and the time information; An extraction module, used to extract images from the video information according to a preset frequency to obtain image information of a preset number of frames; A recognition module, used to recognize each frame of the image information using the target algorithm to obtain a recognition result corresponding to each frame of the image information; A determination module is used to determine whether there is abnormal information in the video information according to the recognition result.

9. A computer device, It is characterized in that include: at least one processor; And a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor performs the steps of the video analysis method as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the steps of the video analysis method according to any one of claims 1 to 7 are implemented.

Images