Video monitoring system and device based on AI analysis and storage medium

Abstract

The invention relates to the technical field of video monitoring, in particular to a video monitoring system and device based on AI analysis and a storage medium, and the method comprises the steps: obtaining user related information in response to a detected trigger operation for a login control on a login page on a system control panel; in response to determining that the user-related information is found in a user information database, entering a user background corresponding to the user-related information; in response to the detected triggering operation for the determined control on the background of the user, obtaining search information input on the background of the user; searching the target video in the video database based on the search information, and displaying the video index corresponding to the target video on the user background display page, so that an advanced AI algorithm is adopted, targets and behaviors in a monitoring picture can be accurately identified and analyzed in a complex environment, the reliability of a monitoring system is improved, and the user experience is improved. Multiple intelligent applications are integrated, accurate prevention and control can be achieved, and the security level is improved.

Description

Technical Field

[0001] The present application relates to the field of video surveillance technology, and in particular to video surveillance systems, devices, and storage media based on AI analysis. Background Art

[0002] Video surveillance systems play a vital role in modern society and are widely used in public safety, traffic management, commercial security, etc. Traditional video surveillance systems mainly rely on manual monitoring and simple motion detection technology.

[0003] Traditional video surveillance systems require a large number of manual monitoring personnel to view the surveillance images in real time, which not only increases labor costs, but also easily leads to missed reports and false alarms due to human negligence. Traditional motion detection technology mainly relies on pixel changes to determine whether an object is moving. This method is prone to false alarms in situations such as light changes and weather changes, and cannot identify specific object types and behaviors. With the popularization of high-definition cameras, the amount of video data has increased dramatically. Traditional systems face tremendous pressure in data storage and processing, making it difficult to achieve real-time analysis and rapid response.

[0004] At present, there are many AI-based video surveillance systems on the market, and these systems have their own characteristics in technical implementation. Front-end AI intelligent analysis is to embed relevant AI algorithms into smart cameras, perform intelligent analysis on the collected video images and output the results. This intelligent analysis module is only effective for one camera. If each point needs to be intelligently analyzed, then each camera needs to be embedded with an AI algorithm. Back-end AI intelligent analysis is to use the back-end system platform to perform intelligent analysis on all video resources collected by the monitoring equipment. The difference between embedded and software platforms is only that the computing platforms are different, but the functions implemented are the same. Back-end AI intelligent analysis can analyze video streams and pictures, and intelligently identify and analyze unsafe behaviors of people, unsafe states of objects, and unsafe factors in the environment. Among them, the cost of front-end AI intelligent analysis is high, and the real-time performance of back-end AI intelligent analysis is poor. Summary of the invention

[0005] The main purpose of this application is to provide a video surveillance system, device and storage medium based on AI analysis, which is used to solve the technical problems of high cost of front-end AI intelligent analysis and poor real-time performance of back-end AI intelligent analysis.

[0006] To achieve the above-mentioned object, the present invention provides a video surveillance system based on AI analysis, comprising: in response to detecting a triggering operation on a login control on a login page on a system control panel, obtaining user related information; In response to determining that the user-related information is found in the user information database, entering a user backend corresponding to the user-related information; In response to detecting a trigger operation on a certain control on the user background, obtaining search information input on the user background; The target video is retrieved from the video database based on the search information, and the video index corresponding to the target video is displayed on the user background display page.

[0007] Optionally, the system further comprises: Performing video preprocessing on the video data received from the terminal device and transmitted through the network to obtain preprocessed video; Inputting the preprocessed video into a pre-trained real-time processing model to obtain a corresponding detection result; According to the corresponding detection result, the video data and the corresponding detection result are stored in the video database.

[0008] Optionally, performing video preprocessing on the video data received from the terminal device and transmitted through the network to obtain the preprocessed video includes: Performing frame extraction processing on the video data to obtain first video data; Scaling the first video data to obtain second video data; The second video data is subjected to denoising processing to obtain the preprocessed video.

[0009] Optionally, the inputting the preprocessed video into a pre-trained real-time processing model to obtain a corresponding detection result includes: Using convolutional neural networks and a single-stage detection method to construct the pre-trained real-time processing model; Inputting each image in the preprocessed video into the convolutional neural network in sequence to extract image features of each image; Performing classification and recognition processing on the image based on the image features to obtain a category probability distribution of each category corresponding to each object in the image; Based on the category probability distribution of each classification, the category corresponding to each object in the image is filtered to obtain the corresponding detection result.

[0010] Optionally, inputting each image in the preprocessed video into the convolutional neural network in sequence to extract image features of each image includes: Adjusting a fixed size of each image in the preprocessed video, wherein the preprocessed video includes a plurality of images; The fixed-size image is input into the convolutional layer in the YOLOv8 network to extract image features of the image.

[0011] Optionally, performing classification and recognition processing on the image based on the image features to obtain a category probability distribution of each category corresponding to each object in the image includes: Dividing the fixed-size image into a plurality of grid cells, wherein each of the grid cells includes predicting a plurality of bounding boxes; Predicting a set of bounding box parameters for each of the plurality of grid cells, wherein the bounding box parameters include a center position, a width, a height, and a confidence score of the bounding box; Based on the bounding box parameters, marking an anchor point in each of the plurality of grid units to obtain a plurality of anchor points and corresponding bounding boxes, wherein each anchor point in the plurality of anchor points corresponds to a bounding box; Predicting the offset and object category of the bounding box corresponding to each anchor point; Based on the offset and the object category, a category probability distribution corresponding to each bounding box is determined, and the category probability distribution corresponding to each bounding box is used as the category probability distribution of each classification.

[0012] Optionally, filtering the category corresponding to each object in the image based on the category probability distribution of each classification to obtain the corresponding detection result includes: Based on non-maximum suppression, the category corresponding to each object in the image is filtered to obtain the corresponding detection result, wherein the detection result includes a set of bounding boxes, wherein each bounding box corresponds to an object and its category and confidence score.

[0013] Optionally, the pre-trained real-time processing model is trained by YOLOv8 using a standard target detection loss function, wherein the target detection loss function includes a bounding box regression loss, a classification loss, and a confidence loss.

[0014] In addition, to achieve the above-mentioned purpose, the present invention also provides a video surveillance device based on AI analysis, characterized in that the video surveillance device based on AI analysis includes: a memory, a processor, and a video surveillance program based on AI analysis stored in the memory and executable on the processor, and when the video surveillance program based on AI analysis is executed by the processor, the steps of the video surveillance system based on AI analysis as described in any one of the above are implemented.

[0015] In addition, to achieve the above-mentioned purpose, the present invention also provides a computer-readable storage medium, characterized in that a video monitoring program based on AI analysis is stored on the computer-readable storage medium, and when the video monitoring program based on AI analysis is executed by a processor, the steps of the video monitoring system based on AI analysis as described in any one of the above are implemented.

[0016] Compared with the prior art, the present invention has the following beneficial effects: In response to detecting a trigger operation for a login control on a login page on a system control panel, user-related information is obtained; in response to determining that the user-related information is found in a user information database, a user background corresponding to the user-related information is entered; in response to detecting a trigger operation for a control determined on the user background, search information entered on the user background is obtained; based on the search information, a target video is retrieved from a video database, and a video index corresponding to the target video is displayed on the user background display page. The embodiments of the present disclosure implement the use of advanced AI algorithms, which can accurately identify and analyze targets and behaviors in monitoring images in complex environments, reduce false alarms and missed alarms, improve the reliability of the monitoring system, integrate a variety of intelligent applications, and can accurately prevent and control and improve security levels. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A schematic diagram of an embodiment of a video surveillance system based on AI analysis in an embodiment of the present invention; Figure 2 A schematic diagram of an embodiment of a video database in a video surveillance system based on AI analysis in an embodiment of the present invention; Figure 3 Schematic diagram of an embodiment of a video surveillance device based on AI analysis in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The main purpose of the present invention is to provide a video surveillance system, equipment and storage medium based on AI analysis; to realize the use of advanced AI algorithms, which can accurately identify and analyze targets and behaviors in the monitoring screen in a complex environment, reduce false alarms and missed alarms, and improve the reliability of the monitoring system; to integrate a variety of intelligent applications, such as face recognition, behavior analysis, anomaly detection, etc., which can accurately prevent and control and improve the security level; to adopt unified standards and protocols, support the access and data sharing of multi-brand equipment, solve the problems of system integration and standardization, and eliminate the phenomenon of data islands.

[0019] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0020] First, the present invention uses deep learning technology to train an AI model that can run on an embedded system. The model can achieve high-frame rate human behavior detection, object retrieval, and vehicle status detection. The model can run completely offline in the terminal device without relying on the network.

[0021] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 , is a schematic diagram of an embodiment of a video surveillance system based on AI analysis in an embodiment of the present invention, including: Step 101, in response to detecting a triggering operation on a login control on a login page on a system control panel, obtaining user related information.

[0022] It is understandable that the execution subject of the present invention may be a video surveillance system based on AI analysis, or may be a terminal or a server, which is not specifically limited here. The embodiment of the present invention is described by taking a server as the execution subject as an example.

[0023] The system control panel can be an NVR (Network Video Recorder) control panel, which is an interface for managing and operating NVR devices; NVR devices are devices used in digital video surveillance systems, and their main functions are to receive, store, manage and play video streams from network cameras. The login control can be a control for triggering user login and entering the user background; the triggering operation includes click operation, sliding operation, voice control, etc.; the user-related information includes the user name and the corresponding password.

[0024] Step 102, in response to determining that the user-related information is found in the user information database, entering the user backend corresponding to the user-related information.

[0025] The user information database can be a database for storing user-related information; the videos that can be obtained by the user background corresponding to each user-related information are different; for example, the user-related information of a staff member on a train can be used to find videos on the train and on the track; while the user-related information of a train manager in a certain area can be used to find videos on all trains in the area and on all railway tracks in the area. The user background includes an information retrieval area and a video index display area; the information retrieval area is used for users to input search information in the area; the video index display area is used to display video indexes.

[0026] Step 103: in response to detecting a trigger operation on a certain control on the user background, obtaining search information input on the user background.

[0027] The above-mentioned determination control is on the information retrieval area of ​​the above-mentioned user background; the above-mentioned search information may include alarm information, dangerous goods, passenger behavior, item management, identity recognition, environmental monitoring, start time and end time, etc. Among them, the above-mentioned alarm information includes fire alarm and emergency evacuation; the above-mentioned dangerous goods include explosives, knives and chemicals; the above-mentioned passenger behavior includes people gathering and fainting; the above-mentioned item management includes mobile phones, wallets, backpacks, suitcases and water cups; the above-mentioned identity recognition includes wanted criminals, ticket evaders and VIP passengers; the above-mentioned environmental monitoring includes environmental parameters such as temperature, humidity, light in the car; the above-mentioned start time and end time are the time range for retrieving videos.

[0028] Step 104: Retrieve the target video from the video database based on the search information, and display the video index corresponding to the target video on the user background display page.

[0029] The execution entity may retrieve the target video in the video database based on the search information, and display the video index corresponding to the target video on the user background display page.

[0030] As an example, alarms or information push can be issued for certain content of concern, such as abnormal human behavior and abnormal vehicle posture. Equipment operators can query specific events, items, behaviors, etc. through the user interface, such as: Query lost items: Users can enter lost item keywords, such as "wallet", through the user background to query video content with the object in the picture. Event query: Users can select the corresponding event, such as "people gathering" in the user background to query video content with the object in the picture.

[0031] As an example, the above-mentioned execution entities can automatically identify abnormal behaviors of passengers, such as running, fighting, pushing, and other behaviors that may threaten public safety, and issue alarms in time through deep learning algorithms. In the carriage, by detecting the density of the crowd, safety accidents caused by crowds can be predicted and prevented. By analyzing the passengers' boarding and alighting behaviors, seat selection preferences, etc., passenger behavior data can be provided to train operators to optimize service processes and resource allocation. Using emotion recognition technology, the emotional state of passengers can be analyzed to provide information to train attendants so that they can better provide services and support to passengers. Facial recognition and identity verification are used to verify the identity of passengers, check whether there are wanted criminals, ticket evaders, or identify VIP passengers to provide customized services. At the security checkpoint or in the carriage, scanning luggage and personal belongings through deep learning algorithms can automatically identify and alarm contraband to improve safety. Attaching RFID tags or performing image recognition to passengers' luggage can track the location of luggage in real time to reduce the loss and mistaking of luggage. Combined with the camera outside the car, the driver can be assisted in driving safely by identifying whether there are obstacles, snow or other dangerous conditions on the track. Vision sensors monitor the temperature, humidity, light and other environmental parameters in the carriage, automatically adjust air conditioning, lighting and other equipment to provide passengers with a more comfortable riding environment. Detect the sanitary conditions in the carriage, such as overflowing trash cans and stains on the ground, and promptly remind cleaning staff to clean up. Through face recognition technology, passengers' identity information can be quickly verified, ticket checking efficiency can be improved, and ticket fraud can be prevented. Combined with passengers' historical travel records and preferences, deep learning algorithms can be used to provide passengers with personalized recommendations for dining, entertainment and other services. Deep learning algorithms can be used to detect smoke in the carriage, timely discover fire hazards and initiate emergency plans. In an emergency, visual sensors and deep learning algorithms can be used to analyze crowd flow and provide passengers with the best evacuation routes and instructions.

[0032] The device uses this model to process video data transmitted from the camera through the network in real time, classify and mark the video content, and store it in the corresponding database.

[0033] Continue to refer Figure 2 , is a schematic diagram of an embodiment of a video database in a video surveillance system based on AI analysis in an embodiment of the present invention, including: Step 201, performing video preprocessing on the video data received from the terminal device and transmitted through the network to obtain preprocessed video.

[0034] The terminal device may be a camera; the video preprocessing may include frame extraction, scaling / cropping and denoising / enhancement.

[0035] Due to the scattered construction of video platforms in different units, numerous equipment brands and different standards, data sharing between systems is insufficient and data islands are serious. The present invention adopts unified standards and protocols, supports the access and data sharing of multi-brand equipment, solves the system integration and standardization problems, and eliminates the data island phenomenon.

[0036] Optionally, frame extraction processing is performed on the video data to obtain first video data; scaling processing is performed on the first video data to obtain second video data; and denoising processing is performed on the second video data to obtain the pre-processed video.

[0037] The above-mentioned frame extraction is the process of selecting a specific frame image from a video stream. In video processing and analysis, it is sometimes necessary to extract key frames to represent the main content of the video, or to extract specific frames according to a specific time point or event. For example, in video surveillance, key frames can be extracted for quick browsing and event analysis; in video editing, specific frames can be extracted for special effects processing or image synthesis. Scaling is the process of changing the size of an image or video frame. An image can be enlarged to see more details, or it can be reduced to save storage space or adapt to a specific display size. Cropping is the process of selecting a specific area of ​​an image or video frame and removing other parts. It can be used to remove unwanted backgrounds, highlight specific objects, or adjust the composition of an image. Denoising is the process of removing noise from an image or video. Noise may come from the image acquisition device, the transmission process, or other factors, which will reduce the quality and clarity of the image. Enhancement is the process of improving the quality and visual effects of an image or video. Images can be made clearer, more vivid, and more attractive by enhancing contrast, brightness, color saturation, etc.

[0038] Step 202: input the preprocessed video into a pre-trained real-time processing model to obtain a corresponding detection result.

[0039] The above-mentioned real-time processing model can be a pre-trained AI model that can run on an embedded system. The model can achieve high frame rate human behavior detection, object retrieval and vehicle status detection. The model can run completely offline in the terminal device without relying on the network. The above-mentioned detection results can be a set of bounding boxes, each box corresponding to an object and its category and confidence score. These results can be used directly to mark objects in the image or perform further analysis. The above-mentioned real-time processing model can be a YOLOv8 network.

[0040] Optionally, a convolutional neural network and a single-stage detection method are used to construct the pre-trained real-time processing model; each image in the pre-processed video is input into the convolutional neural network in turn to extract image features of each image; the image is classified and recognized based on the image features to obtain a category probability distribution of each category corresponding to each object in the image; based on the category probability distribution of each category, the category corresponding to each object in the image is filtered to obtain the corresponding detection result.

[0041] YOLOv8's object recognition principle is based on convolutional neural networks (CNN) and one-stage detection methods; YOLOv8 uses a single neural network to generate bounding boxes and category probabilities directly from the input image. Unlike the traditional region proposal network (R-CNN) method, YOLOv8 does not need to generate candidate regions first, and then perform classification and regression processing, but merges these steps into one network, thereby speeding up the processing. YOLOv8 uses a deep convolutional neural network to extract features of the input image. The convolutional layers in the network extract low-level to high-level features such as edges, textures, shapes, etc. layer by layer. These features will be passed to the subsequent detection layer for further analysis.

[0042] Optionally, the fixed size of each image in the preprocessed video is adjusted, wherein the preprocessed video includes multiple images; the fixed-size image is input into a convolutional layer in a YOLOv8 network to extract image features of the image.

[0043] The fixed size mentioned above can be 640 pixels x 640 pixels; the image is first resized to a fixed size and then input into the YOLOv8 network. The network processes the entire image at once and extracts the features of the image.

[0044] Optionally, the fixed-size image is divided into a plurality of grid units, wherein each of the grids includes a plurality of predicted bounding boxes; a set of bounding box parameters is predicted for each of the plurality of grid units, wherein the bounding box parameters include a center position, width, height, and confidence score of the bounding box; based on the bounding box parameters, anchor points are marked in each of the plurality of grid units to obtain a plurality of anchor points and corresponding bounding boxes, wherein each of the plurality of anchor points corresponds to a bounding box; an offset and an object category of the bounding box corresponding to each anchor point are predicted; based on the offset and the object category, a category probability distribution corresponding to each bounding box is determined, and the category probability distribution corresponding to each bounding box is used as the category probability distribution of each classification.

[0045] The image is divided into multiple grid cells (e.g. 13x13, 26x26, etc.), each of which is responsible for predicting multiple bounding boxes. YOLOv8 uses an anchor mechanism to handle objects of different sizes and shapes. Each anchor represents a possible object candidate, and the network predicts the offset and object category of the bounding box for each anchor. For each grid cell, YOLOv8 predicts a set of bounding box parameters, including the center position (x, y), width, height, and confidence score of the box. The confidence score indicates the possibility of an object in the box. With these parameters, the network generates the bounding box of the object. While predicting the bounding box, YOLOv8 also predicts the category of the object in each box. This is achieved through the classification head in the network, which outputs a category probability distribution for each bounding box.

[0046] Optionally, based on non-maximum suppression, the category corresponding to each object in the image is filtered to obtain the corresponding detection result, wherein the detection result includes a set of bounding boxes, wherein each bounding box corresponds to an object and its category and confidence score.

[0047] YOLOv8 uses non-maximum suppression to filter out redundant detection boxes. For the same object, there may be multiple overlapping bounding boxes. NMS will retain the box with the highest confidence and remove other boxes with higher overlap (IoU) to obtain the final detection result. YOLOv8 outputs a set of bounding boxes, each of which corresponds to an object and its category and confidence score. These results can be used directly to mark objects in the image or for further analysis.

[0048] Optionally, the pre-trained real-time processing model is trained by YOLOv8 using a standard target detection loss function, wherein the target detection loss function includes a bounding box regression loss, a classification loss, and a confidence loss.

[0049] YOLOv8 uses standard target detection loss functions for training, including bounding box regression loss, classification loss, and confidence loss. The network continuously updates weights through backpropagation and optimization algorithms to improve detection accuracy.

[0050] Step 203: store the video data and the corresponding detection result in the video database according to the corresponding detection result.

[0051] The device uses this model to process video data transmitted from the camera through the network in real time, classify and mark the video content, and store it in the corresponding database.

[0052] Traditional motion detection technology mainly relies on pixel changes to determine whether an object is moving. This method is prone to false alarms in situations such as light changes and weather changes, and cannot identify specific object types and behaviors. The use of advanced AI algorithms can accurately identify and analyze targets and behaviors in monitoring images in complex environments, reduce false alarms and missed alarms, and improve the reliability of monitoring systems. With the popularity of high-definition cameras, the amount of video data has increased dramatically. Traditional systems face tremendous pressure in data storage and processing, and it is difficult to achieve real-time analysis and rapid response. The existing video surveillance system lacks intelligent applications, cannot accurately prevent and control, and cannot meet modern security needs. It integrates a variety of intelligent applications, such as face recognition, behavior analysis, anomaly detection, etc., which can accurately prevent and control and improve security levels.

[0053] above Figure 1 The video surveillance system based on AI analysis in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The video surveillance device based on AI analysis in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0054] Figure 3 : is a structural diagram of a video surveillance device based on AI analysis provided by an embodiment of the present invention. The video surveillance device 300 based on AI analysis may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 310 (for example, one or more processors) and a memory 320, and one or more storage media 330 (for example, one or more mass storage devices) storing application programs 333 or data 332. Among them, the memory 320 and the storage medium 330 can be short-term storage or permanent storage. The program stored in the storage medium 330 may include one or more modules (not shown in the figure), and each module may include a series of computer program operations in the video surveillance device 300 based on AI analysis. Furthermore, the processor 310 may be configured to communicate with the storage medium 330 to execute a series of computer program operations in the storage medium 330 on the video surveillance device 300 based on AI analysis.

[0055] The video surveillance device 300 based on AI analysis may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input and output interfaces 360, and / or one or more operating systems 331, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. It will be appreciated by those skilled in the art that Figure 3The structure of the video surveillance device based on AI analysis shown does not constitute a limitation on the video surveillance device based on AI analysis, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0056] The present invention also provides a video surveillance device based on AI analysis, comprising: a memory and at least one processor, wherein a computer program is stored in the memory, and the memory and the at least one processor are interconnected through a line; the at least one processor calls the computer program in the memory so that the video surveillance device based on AI analysis executes the steps in the above-mentioned video surveillance system based on AI analysis.

[0057] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. A computer program is stored in the computer-readable storage medium. When the computer program runs on a computer, the computer executes the steps of a video surveillance system based on AI analysis.

[0058] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned system embodiments and will not be repeated here.

[0059] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several computer programs to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0060] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A video surveillance system based on AI analysis, characterized in that: include: In response to detecting a triggering operation on a login control on a login page on a system control panel, obtaining user related information; In response to determining that the user-related information is found in the user information database, entering a user backend corresponding to the user-related information; In response to detecting a trigger operation on a certain control on the user background, obtaining search information input on the user background; The target video is retrieved from the video database based on the search information, and the video index corresponding to the target video is displayed on the user background display page.

2. The video surveillance system based on AI analysis according to claim 1, characterized in that: The system further comprises: Performing video preprocessing on the video data received from the terminal device and transmitted through the network to obtain preprocessed video; Inputting the preprocessed video into a pre-trained real-time processing model to obtain a corresponding detection result; According to the corresponding detection result, the video data and the corresponding detection result are stored in the video database.

3. The video surveillance system based on AI analysis according to claim 2, characterized in that: The step of performing video preprocessing on the video data received from the terminal device and transmitted through the network to obtain the preprocessed video includes: Performing frame extraction processing on the video data to obtain first video data; Scaling the first video data to obtain second video data; The second video data is subjected to denoising processing to obtain the preprocessed video.

4. The video surveillance system based on AI analysis according to claim 2, characterized in that: The pre-processed video is input into a pre-trained real-time processing model to obtain a corresponding detection result, including: Using convolutional neural networks and a single-stage detection method to construct the pre-trained real-time processing model; Inputting each image in the preprocessed video into the convolutional neural network in sequence to extract image features of each image; Performing classification and recognition processing on the image based on the image features to obtain a category probability distribution of each category corresponding to each object in the image; Based on the category probability distribution of each classification, the category corresponding to each object in the image is filtered to obtain the corresponding detection result.

5. The video surveillance system based on AI analysis according to claim 4, characterized in that: The step of sequentially inputting each image in the preprocessed video into the convolutional neural network to extract image features of each image includes: Adjusting a fixed size of each image in the preprocessed video, wherein the preprocessed video includes a plurality of images; The fixed-size image is input into the convolutional layer in the YOLOv8 network to extract image features of the image.

6. The video surveillance system based on AI analysis according to claim 5, characterized in that: The performing classification and recognition processing on the image based on the image features to obtain the category probability distribution of each category corresponding to each object in the image includes: Dividing the fixed-size image into a plurality of grid cells, wherein each of the grid cells includes predicting a plurality of bounding boxes; Predicting a set of bounding box parameters for each of the plurality of grid cells, wherein the bounding box parameters include a center position, a width, a height, and a confidence score of the bounding box; Based on the bounding box parameters, marking an anchor point in each of the plurality of grid units to obtain a plurality of anchor points and corresponding bounding boxes, wherein each anchor point in the plurality of anchor points corresponds to a bounding box; Predicting the offset and object category of the bounding box corresponding to each anchor point; Based on the offset and the object category, a category probability distribution corresponding to each bounding box is determined, and the category probability distribution corresponding to each bounding box is used as the category probability distribution of each classification.

7. The video surveillance system based on AI analysis according to claim 6, characterized in that: The filtering of the category corresponding to each object in the image based on the category probability distribution of each classification to obtain the corresponding detection result includes: Based on non-maximum suppression, the category corresponding to each object in the image is filtered to obtain the corresponding detection result, wherein the detection result includes a set of bounding boxes, wherein each bounding box corresponds to an object and its category and confidence score.

8. The video surveillance system based on AI analysis according to claim 1, characterized in that: The pre-trained real-time processing model is trained by YOLOv8 using a standard target detection loss function, wherein the target detection loss function includes a bounding box regression loss, a classification loss, and a confidence loss.

9. A video surveillance device based on AI analysis, characterized in that: The video surveillance device based on AI analysis includes: a memory, a processor, and a video surveillance program based on AI analysis stored in the memory and executable on the processor. When the video surveillance program based on AI analysis is executed by the processor, the steps of the video surveillance system based on AI analysis as described in any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a video monitoring program based on AI analysis, and when the video monitoring program based on AI analysis is executed by a processor, the steps of the video monitoring system based on AI analysis as described in any one of claims 1 to 8 are implemented.

Images