Identification method for behavior analysis intelligent monitoring system

Abstract

The invention relates to an identification method for a behavior analysis intelligent monitoring system, and the method comprises the following steps: S1, data collection: collecting image data through a camera of the intelligent monitoring system; s2, data processing: processing and explaining image data acquired by a camera of the intelligent monitoring system through an algorithm of the intelligent monitoring system so as to subsequently identify and analyze specific behaviors; s3, behavior recognition: detecting, tracking and recognizing behaviors of persons and objects in the image through an intelligent monitoring system by using the data processed in the S2; and S4, behavior analysis and judgment: analyzing a specific behavior according to an identification result. According to the method, an LBP feature extraction algorithm and a Yolov5 algorithm are sequentially adopted to comprehensively judge the condition of a monitoring area of a monitoring system, then AI computing power is input according to the human body of a person, whether the behavior of the person is damaged or not is judged, and therefore the computing power input is reduced, hardware with low cost can be input, and electric energy consumption is reduced.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of intelligent monitoring systems, and particularly relates to a method for identifying a behavior analysis intelligent monitoring system. BACKGROUND

[0002] The method for identifying a behavior analysis intelligent monitoring system is usually simply dependent on a mature AI algorithm, so that the intelligent monitoring system is always in a high-speed running state, a large amount of computing power needs to be called and a large amount of electric power energy needs to be consumed, so that a large hardware cost needs to be invested, the use cost is also high, and the single AI algorithm also has the problem of misidentification. SUMMARY

[0003] This section aims to summarize some aspects of the embodiments of the application and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract and title of the specification to avoid obscuring the purpose of this section, the abstract and the title, and such simplifications or omissions cannot be used to limit the scope of the application.

[0004] In view of the following technical problems in the prior art: the method for identifying a behavior analysis intelligent monitoring system is usually simply dependent on a mature AI algorithm, so that the intelligent monitoring system is always in a high-speed running state, a large amount of computing power needs to be called and a large amount of electric power energy needs to be consumed, so that a large hardware cost needs to be invested, the use cost is also high, and the single AI algorithm also has the problem of misidentification.

[0005] To solve the above technical problems, the application provides the following technical scheme: a method for identifying a behavior analysis intelligent monitoring system, comprising the following steps:

[0006] S1: data acquisition, image data is acquired by a camera of the intelligent monitoring system;

[0007] S2: data processing, the image data acquired by the camera of the intelligent monitoring system is processed and interpreted by an algorithm of the intelligent monitoring system, so as to identify and analyze specific behaviors subsequently;

[0008] S3: behavior identification, the data processed by S2 is used to detect, track and identify the behaviors of personnel and objects in the image by the intelligent monitoring system;

[0009] S4: behavior analysis and judgment, according to the identification result, specific behaviors are analyzed, and if abnormal behaviors or behaviors threatening public safety are found, the intelligent monitoring system will issue a warning;

[0010] S5: Data storage, backup and storage of important data in S1-S4 process for subsequent viewing and analysis.

[0011] As a preferred technical solution for the identification method of the behavior analysis intelligent monitoring system, the S2 comprises the following steps:

[0012] S21: Preprocessing of the image captured in S1 to reduce the influence of the camera system on the identification of the captured image under different lighting conditions, including:

[0013] Gaussian filtering: Gaussian filtering can smooth the image and reduce noise, thereby removing the influence of light;

[0014] S22: Feature extraction, using LBP (Local Binary Patterns) to extract the local features of the image processed in S21 as the basis for discrimination:

[0015] The steps of the LBP feature extraction algorithm are as follows:

[0016] S221: Convert the image to a grayscale image, divide the grayscale image into 3x3 pixel blocks, and define a 3x3 neighborhood: take the center pixel as the center and the 3x3 pixel as the radius to form a 9-pixel circular neighborhood;

[0017] S222: Calculate the gray value of the center pixel;

[0018] S223: Compare the gray values of the 8 pixels in the neighborhood with the gray value of the center pixel;

[0019] If the pixel value around the center pixel is less than the gray value of the center pixel, the pixel position is marked as 0, otherwise marked as 1;

[0020] S224: Combine the results of the comparison to form an 8-bit binary number (usually converted to a decimal number, i.e. LBP code, a total of 256 kinds), i.e. the LBP value of the center pixel in the neighborhood is obtained;

[0021] S225: Use the LBP value obtained in S224 to reflect the texture information of the region;

[0022] S23: Target detection, compare the texture information of the real-time captured image obtained in S22 with the texture information of the image captured in the system storage under the non-abnormal state, concatenate all the LBP feature points to form the LBP feature vector of the picture;

[0023] Then, compare the LBP feature vectors of the two images using the Euclidean distance as the distance measure to determine the similarity of the two images;

[0024] Specifically, if the LBP feature vectors of the two images are closer, it means that their textures are more similar, and vice versa. The determination value is YT;

[0025] S24: target recognition and tracking, if the similarity obtained in S23 is lower than YT, it can be understood that an object has entered the camera area. LBP values of real-time images obtained in S22 are used for face recognition, target tracking, and image classification. Target tracking methods based on filtering and deep learning and target tracking methods based on appearance modeling and deep learning are used to recognize and track targets in images.

[0026] As a preferred technical solution for the behavior analysis intelligent monitoring system recognition method, S3 comprises the following steps:

[0027] Determining the spatial position of the object: based on the Yolov5 algorithm, the spatial position of the object in the region determined in S2 to have an object is determined, and it is determined whether the object entering the monitoring system region invades the sensitive region;

[0028] The method for analyzing the spatial position of the object based on the Yolov5 algorithm is as follows:

[0029] Obtain image data, input the image into the Yolov5 algorithm, and obtain the left upper corner coordinates (x, y) and the width w and height h of the object detection frame;

[0030] According to the left upper corner coordinates and the width w and height h of the detection frame, the center coordinates (x1, y1) of the detection frame are calculated, and the calculation formula is: x1=(x+w / 2), y1=(y+h / 2);

[0031] Taking the center coordinates (x1, y1) as the center point of the spatial position of the object, according to the width w and height h of the detection frame, the width w1 and height h1 of the object in the image are calculated, and the calculation formula is: w1=w / 2, h1=h / 2;

[0032] According to the center point coordinates (x1, y1) and the width w1 and height h1, the spatial position of the object can be determined;

[0033] It should be noted that in actual application, the Yolov5 algorithm needs to be optimized and improved according to specific circumstances to improve the accuracy and precision of object detection;

[0034] Sensitive region boundary determination:

[0035] The method for analyzing the spatial region boundary coordinates based on the Yolov5 algorithm is as follows:

[0036] Obtain image data, input the image into the Yolov5 algorithm, and obtain the left upper corner coordinates (x1, y1) and the width w and height h of the target detection frame;

[0037] According to the left upper corner coordinate (x1, y1) and the width w and height h of the detection frame, the left upper corner coordinate (x1, y1) and the right lower corner coordinate (x2, y2) of the target in the image can be determined, and the calculation formula is: x2=x1+w, y2=y1+h;

[0038] The center coordinate (x1, y1) of the target detection frame can be used as the center point coordinate of the target in the image;

[0039] Intrusion determination: when the coordinates of the object fall within the boundary of the sensitive area, an alarm is immediately issued, and the LBP feature extraction algorithm is used to record the physical features, and the camera is called to continuously track and record the object;

[0040] If the object does not invade the sensitive area, subsequent behavior determination is performed.

[0041] As a preferred technical solution for the behavior analysis and judgment of the intelligent monitoring system, the behavior analysis and judgment method is as follows: based on the method of steps S2 and S3, it is determined whether the object is a human body or a non-human body;

[0042] When the object is a non-human body, it is continuously monitored whether the coordinates invade the sensitive area;

[0043] If the object is a human body, it is continuously monitored whether the coordinates invade the sensitive area while the behavior of the object is analyzed and judged, and the AI model trained is used to identify the behavior of the human body and the LBP feature extraction algorithm is used to limit the analysis of the personnel outside the sensitive area: according to different behavior characteristics, the behavior of the human body is divided into normal behavior and abnormal behavior;

[0044] Among them, feature extraction is the most critical step in the whole process, and its effect directly affects the accuracy and reliability of behavior recognition;

[0045] Generally speaking, the extracted features should contain as much human motion information, posture information, and behavior speed as possible; in practical applications, the model needs to be customized and optimized according to specific scenes and application requirements;

[0046] Based on the above limit recognition to determine the behavior of the human body, when the behavior of the human body is abnormal, such as implementing a destructive behavior, the LBP feature extraction algorithm is used to record the physical features, and the camera is called to continuously track and issue an alarm.

[0047] As a preferred technical solution for the behavior analysis and judgment of the intelligent monitoring system, the alarm includes:

[0048] Screen alarm: when the detector senses a hidden danger, the alarm system will automatically pop up an alarm screen and display it in a conspicuous position, and notify the online operator in the form of flashing screen and text prompt;

[0049] Acousto-optic alarm: when an alarm event occurs, the on-site personnel are notified and reminded by the combination of the buzzer and LED acousto-optic flashing;

[0050] SMS alarm: set the mobile phone number for SMS alarm in the background, send it to the alarm host after collecting the alarm information, and send it to the mobile phone number through the host;

[0051] Email alarm: its alarm mode is the same as SMS alarm, fill in the email address in the contact person column in the management background, and automatically send the alarm information to the set email when the alarm condition occurs;

[0052] Mobile phone APP alarm: open the mobile phone APP function in the management background to provide instant information for users, and the system will immediately push a notification prompt once there is an alarm information on site.

[0053] The beneficial effects of the method for identifying the behavior analysis intelligent monitoring system of the application are as follows: LBP feature extraction algorithm and Yolov5 algorithm are used in sequence to comprehensively determine the condition of the monitoring area of the monitoring system, then AI power is invested in the human body of the personnel to determine whether the behavior is harmful, so as to reduce the power investment, so that lower-cost hardware can be invested and power consumption can be reduced. BRIEF DESCRIPTION OF DRAWINGS

[0054] In order to more clearly illustrate the technical solutions of the embodiments of the application, the drawings needed in the embodiment description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor. Among them:

[0055] Figure 1 The main flowchart for the method of the application is used;

[0056] Figure 2 The detailed flowchart of the data processing part of the application is used;

[0057] Figure 3 The detailed flowchart of the behavior recognition part of the application is used. DETAILED DESCRIPTION

[0058] In order to make the above-mentioned purposes, features and advantages of the application more obvious and easy to understand, the specific embodiments of the application will be described in detail below with reference to the drawings of the specification.

[0059] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without the specific details set forth in this description. In other instances, well-known methods, procedures, components, and circuits have not been described in detail so as not to unnecessarily obscure aspects of the present application.

[0060] Secondly, the present application is described in detail in combination with the schematic diagram. In the detailed description of the embodiments of the present application, the sectional view of the device structure is partially enlarged without the general proportion for the convenience of illustration, and the schematic diagram is only an example which should not limit the scope of protection of the present application. In addition, the three-dimensional spatial dimensions including length, width and depth should be included in the actual manufacture.

[0061] Thirdly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. The "in one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an independent or alternative embodiment which is mutually exclusive with other embodiments.

[0062] Embodiment one: comprising the following steps:

[0063] S1: data acquisition, image data is collected by the camera of the intelligent monitoring system;

[0064] S2: data processing, the image data collected by the camera of the intelligent monitoring system is processed and interpreted by the algorithm of the intelligent monitoring system, so as to identify and analyze specific behaviors subsequently;

[0065] S3: behavior recognition, the data processed by S2 is used to detect, track and recognize the behavior of personnel and objects in the image by the intelligent monitoring system;

[0066] S4: behavior analysis and judgment, according to the recognition result, specific behaviors are analyzed, such as abnormal behaviors or behaviors threatening public safety, and the intelligent monitoring system will issue a warning;

[0067] S5: data storage, important data in the process of S1-S4 is backed up and stored for subsequent viewing and analysis.

[0068] The S2 comprises the following steps:

[0069] S21: the image taken by S1 is preprocessed to reduce the influence of the camera system on the recognition of the taken image under different lighting conditions, including:

[0070] Histogram equalization: histogram equalization of the image can increase the contrast of the image, make the dark area brighter and the bright area darker, so as to remove the influence of light;

[0071] Gaussian filter: Gaussian filter can smooth the image, reduce noise, and thus remove the influence of light;

[0072] Median filter: Median filter can remove noise and spots in the image, making the image smoother and reducing the influence of light;

[0073] Color space conversion: Color space conversion can convert the image from RGB color space to other color spaces such as HSV, YUV, which are usually independent of light, thus reducing the influence of light;

[0074] S22: Feature extraction, using LBP (Local Binary Patterns) to extract the local features of the image processed in S21 as the basis for discrimination:

[0075] The steps of LBP feature extraction algorithm are as follows:

[0076] S221: Convert the image to a grayscale image, divide the grayscale image into 3x3 pixel blocks, and define a 3x3 neighborhood: take the center pixel as the center and the 3x3 pixel as the radius to form a 9-pixel circular neighborhood;

[0077] S222: Calculate the gray value of the center pixel;

[0078] S223: Compare the gray values of the 8 pixels in the neighborhood with the gray value of the center pixel;

[0079] If the pixel value around the center pixel is less than the gray value of the center pixel, the pixel position is marked as 0, otherwise marked as 1;

[0080] S224: Combine the results of the comparison to form an 8-bit binary number (usually converted to a decimal number, i.e. LBP code, a total of 256 kinds), i.e. the LBP value of the center pixel in the neighborhood is obtained;

[0081] S225: Use the LBP value obtained in S224 to reflect the texture information of the region;

[0082] S23: Target detection, compare the texture information of the real-time captured image obtained in S22 with the texture information of the image captured in the system under normal conditions, concatenate all the LBP features of the pixels to form the LBP feature vector of the image;

[0083] Then, compare the LBP feature vectors of the two images using Euclidean distance as the distance measure to determine the similarity of the two images;

[0084] Specifically, if the LBP feature vectors of the two images are closer, it means that their textures are more similar, and vice versa. The determination value is YT;

[0085] S24: target recognition and tracking, if the similarity obtained in S23 is lower than YT, it can be understood that an object has entered the camera area. LBP values of real-time images obtained in S22 are used for face recognition, target tracking, and image classification. Target tracking methods based on filtering and deep learning and target tracking methods based on appearance modeling and deep learning are used to recognize and track targets in images.

[0086] The S3 comprises the following steps:

[0087] Determining the spatial position of the object: based on the Yolov5 algorithm, the spatial position of the object in the region determined in S2 to have an object is analyzed to determine whether the object entering the monitoring system region has invaded the sensitive region;

[0088] The method for analyzing the spatial position of the object based on the Yolov5 algorithm is as follows:

[0089] Obtain image data, input the image into the Yolov5 algorithm, and obtain the left upper corner coordinates (x, y) and width w and height h of the object detection frame;

[0090] According to the left upper corner coordinates and width w and height h of the detection frame, the center coordinates (x1, y1) of the detection frame are calculated, and the calculation formula is: x1=(x+w / 2), y1=(y+h / 2);

[0091] Taking the center coordinates (x1, y1) as the center point of the spatial position of the object, according to the width w and height h of the detection frame, the width w1 and height h1 of the object in the image are calculated, and the calculation formula is: w1=w / 2, h1=h / 2;

[0092] According to the center point coordinates (x1, y1) and the width w1 and height h1, the spatial position of the object can be determined;

[0093] It should be noted that in actual application, the Yolov5 algorithm needs to be optimized and improved according to specific circumstances to improve the accuracy and precision of object detection;

[0094] Sensitive region boundary determination:

[0095] The method for analyzing the spatial region boundary coordinates based on the Yolov5 algorithm is as follows:

[0096] Obtain image data, input the image into the Yolov5 algorithm, and obtain the left upper corner coordinates (x1, y1) and width w and height h of the target detection frame;

[0097] According to the left upper corner coordinates (x1, y1) and the width w and height h of the detection frame, the left upper corner coordinates (x1, y1) and the right lower corner coordinates (x2, y2) of the target in the image can be determined, and the calculation formula is: x2 = x1 + w, y2 = y1 + h;

[0098] The center coordinates (x1, y1) of the target detection frame can be used as the center point coordinates of the target in the image;

[0099] Intrusion determination: when the coordinates of the object fall within the boundary of the sensitive area, an alarm is immediately issued, and the LBP feature extraction algorithm is used to record its physical features, and the camera is called to continuously track and record it;

[0100] If the object does not invade the sensitive area, subsequent behavior determination is performed.

[0101] The behavior analysis and determination method is: based on the methods of steps S2 and S3, it is determined whether the object is a human body or a non-human body;

[0102] When the object is a non-human body, it is continuously monitored whether its coordinates invade the sensitive area;

[0103] If the object is a human body, it is continuously monitored whether its coordinates invade the sensitive area while its behavior is analyzed and determined, and the AI model trained is used to identify the behavior of the human body and the LBP feature extraction algorithm is used to limit the analysis of personnel outside the sensitive area: according to different behavior characteristics, the behavior of the human body is divided into normal behavior and abnormal behavior;

[0104] Among them, feature extraction is the most critical step in the whole process, and its effect directly affects the accuracy and reliability of behavior recognition;

[0105] Generally speaking, the extracted features should contain as much human motion information, posture information, behavior speed, etc. as possible; in practical applications, the model needs to be customized and optimized according to specific scenes and application requirements;

[0106] Based on the above limit recognition, the behavior of the human body is determined, and when the behavior of the human body is abnormal, such as implementing a destructive behavior, the LBP feature extraction algorithm is used to record its physical features, and the camera is called to continuously track and issue an alarm;

[0107] The following methods are adopted to optimize the algorithm and reduce redundant calculations, so as to reduce the energy consumption of YOLOv5 algorithm:

[0108] Data enhancement: through pasting, cropping, mosaic, affine transformation, color space conversion and other ways to enhance the sample, increase the diversity of the target, and improve the detection and classification accuracy of the model;

[0109] Optimization strategy: use SAM optimizer, with inherent robustness;

[0110] Loss function: use Varifocal Loss loss function, mainly train dense target detector to make IOU-aware classification score (IASC) regression, to improve detection accuracy.

[0111] The alarm includes:

[0112] Screen alarm: when the detector senses a hidden danger, the alarm system will automatically pop up an alarm screen and display it in a conspicuous position, and notify the online operator in the form of flashing screen and text prompt;

[0113] Acoustic and light alarm: when the alarm event occurs, the on-site personnel are warned through the combination of buzzer and LED sound and light flickering to remind the on-site duty personnel;

[0114] SMS alarm: set the mobile phone number for SMS alarm in the background, and send the alarm information to the alarm host after collecting the alarm information, and send the mobile phone number through the host;

[0115] Email alarm: its alarm mode is the same as SMS alarm, fill in the email address in the contact person column in the management background, and automatically send the alarm information to the set email when the alarm condition occurs;

[0116] Mobile phone APP alarm: open the mobile phone APP function in the management background, which can provide instant information for users, and the system will immediately push the notification prompt once there is alarm information on the scene.

[0117] Example two:

[0118] The second embodiment of the present application provides a behavior recognition algorithm code based on the first embodiment, a camera human body posture estimation code based on Python, using OpenCV and OpenPose library:

[0119] pythonimport cv2

[0120] import numpy as np

[0121] import openpose as op

[0122] # Load OpenPose model

[0123] params=dict()

[0124] params['model_folder']='path / to / openpose / models'

[0125] opWrapper = op.WrapperPython()

[0126] opWrapper.loadModelFromFolder(params['model_folder'])

[0127] # Open camera

[0128] cap = cv2.VideoCapture(0)

[0129] # Loop to detect human pose

[0130] while True:

[0131] # Read camera image

[0132] ret, frame = cap.read()

[0133] if not ret:

[0134] break

[0135] # Image preprocessing

[0136] frame = cv2.resize(frame, (432, 368))

[0137] rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)

[0138] rgb = np.transpose(rgb, (2, 0, 1))

[0139] rgb = np.expand_dims(rgb, axis=0)

[0140] # Calculate human pose keypoint coordinates

[0141] datum = op.Datum()

[0142] datum.cvInputData = rgb

[0143] opWrapper.emplaceAndPop([datum])

[0144] # Draw human pose keypoint

[0145] for i in range(datum.poseKeypoints.shape[0]):

[0146] if i == 0: # head keypoint

[0147] cv2.circle(frame, (int(datum.poseKeypoints[i, 0]), int(datum.poseKeypoints[i, 1])), 3, (0, 0, 255), -1)

[0148] elif i == 17: # hip keypoint

[0149] cv2.circle(frame, (int(datum.poseKeypoints[i, 0]), int(datum.poseKeypoints[i, 1])), 3, (0, 255, 0), -1)

[0150] else: # other keypoint

[0151] cv2.circle(frame, (int(datum.poseKeypoints[i, 0]), int(datum.poseKeypoints[i, 1])), 3, (255, 0, 0), -1)

[0152] cv2.line(frame, (int(datum.poseKeypoints[i, 0]), int(datum.poseKeypoints[i, 1])), (int(datum.poseKeypoints[(i - 1) % 18, 0]), int(datum.poseKeypoints[(i - 1) % 18, 1])), (0, 0, 255), 2)

[0153] cv2.line(frame, (int(datum.poseKeypoints[i, 0]), int(datum.poseKeypoints[i, 1])), (int(datum.poseKeypoints[(i + 1) % 18, 0]), int(datum.poseKeypoints[(i + 1) % 18, 1])), (0, 255, 0), 2)

[0154] cv2.line(frame, (int(datum.poseKeypoints[i, 0]), int(datum.poseKeypoints[i, 1])), (int(datum.poseKeypoints[(i + 9) % 18, 0]), int(datum.poseKeypoints[(i + 9) % 18, 1])), (255, 0, 0), 2)

[0155] cv2.line(frame, (int(datum.poseKeypoints[i, 0]), int(datum.poseKeypoints[i, 1])), (int(datum.poseKeypoints[(i - 9) % 18, 0]), int(datum.poseKeypoints[(i - 9) % 18,

[0156] In the above code, we first load a pre-trained TensorFlow model that has been trained to recognize human behaviors; then we use the OpenCV library to open the camera and read the camera image in an infinite loop; we preprocess each image and use the model to predict the human behavior category in the image; if the prediction result is abnormal behavior, draw a red rectangle on the image to mark the abnormal behavior area; finally, we display the result image and wait for user input, if the user presses the "q" key, the program will exit; finally, we release the camera resource and close the window.

[0157] It should be appreciated that in the development of any actual implementation, as in any engineering or design project, numerous implementation-specific decisions can be made. Such development efforts might be complex and time-consuming, but would nevertheless be a routine undertaking for those of ordinary skill in the art having the benefit of this disclosure.

[0158] It should be noted that the above examples are only used to illustrate the technical solutions of the present application rather than limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, and all such modifications or replacements should be covered in the scope of the claims of the present application.

Claims

1. A method for identifying behavior analysis in an intelligent monitoring system, characterized by: The following steps are involved: S1: Data acquisition, collecting image data through the camera of the smart monitoring system; S2: Data processing, which uses the smart monitoring system's algorithms to process and interpret the image data collected by the smart monitoring system's cameras in order to subsequently identify and analyze specific behaviors; S3: Behavior recognition, using the data processed by S2 to detect, track, and identify the behavior of people and objects in the image through the intelligent monitoring system; S4: Behavior analysis and judgment: Based on the recognition results, specific behaviors are analyzed. If abnormal behaviors or behaviors that threaten public safety are found, the intelligent monitoring system will issue an early warning; S5: Data storage, backing up and storing important data from S1-S4 for subsequent review and analysis.

2. The method for identifying a behavior analysis intelligent monitoring system according to claim 1, characterized in that: The S2 comprises the following steps: S21: Preprocess the image captured by S1 to reduce the impact of the camera system on the recognition of the captured image under different lighting conditions, including: Gaussian filtering: Gaussian filtering can smooth the image, reduce noise, and thus remove the influence of lighting; S22: Feature extraction, using LBP (Local Binary Patterns) to extract local features of the image processed in S21 as a basis for discrimination: The steps of the LBP feature extraction algorithm are as follows: S221: Convert the image into a grayscale image, divide the grayscale image into 3x3 pixel blocks, and define a 3x3 neighborhood: a 9-pixel circular neighborhood is formed with the central pixel as the center and the 3x3 pixel points as the radius; S222: Calculate the grayscale value of the center pixel; S223: Compare the grayscale values ​​of the eight pixels in the neighborhood with the grayscale value of the central pixel; If the surrounding pixel values ​​are less than the grayscale value of the central pixel, the pixel position is marked as 0, otherwise it is marked as 1; S224: Combining the comparison results to form an 8-bit binary number (usually converted to a decimal number, i.e., an LBP code, of which there are 256 types), thus obtaining the LBP value of the central pixel of the neighborhood; S225: Use the LBP value obtained in S224 to reflect the texture information of the area; S23: Target detection, using S22 to obtain the real-time image and the texture information of the image taken when there is no abnormality stored by the system, and then compare them, and concatenate the LBP features of all pixels to form the LBP feature vector of the image; Then, the LBP feature vectors of the two images are compared, and the Euclidean distance is used as the distance metric to determine the similarity between the two images; Specifically, if the LBP feature vectors of two images are closer, it means that their textures are more similar, otherwise it means that their textures are more different, and the judgment value is YT; S24: Target recognition and tracking. The similarity obtained in S23 is lower than YT, which can be understood as an object entering the camera area. The LBP value of the real-time captured image obtained by S22 is used for face recognition, target tracking, and image classification. The target tracking method based on filtering and deep learning and the target tracking method based on appearance modeling and deep learning are used to identify and track the target in the image.

3. The method for identifying a behavior analysis intelligent monitoring system according to claim 1, characterized in that: The S3 includes the following steps: Determine the spatial position of an object: Based on the Yolov5 algorithm analysis S2, determine the spatial position of an object in an area where there is an object, and determine whether the object entering the monitoring system area has invaded a sensitive area; The method for analyzing the spatial position of an object based on the Yolov5 algorithm is as follows: Obtain image data and input the image into the Yolov5 algorithm to obtain the upper left corner coordinates (x, y) and width w and height h of the object detection box; Calculate the center coordinates (x1, y1) of the detection frame based on its upper left corner coordinates, width w, and height h. The calculation formula is: x1 = (x + w / 2), y1 = (y + h / 2). The center coordinates (x1, y1) are used as the center point of the object's spatial position. Based on the width w and height h of the detection box, the width w1 and height h1 of the object in the image are calculated using the following formulas: w1 = w / 2, h1 = h / 2. The spatial position of the object can be determined based on the center point coordinates (x1, y1) and the width w1 and height h1; It should be noted that in practical applications, the Yolov5 algorithm needs to be optimized and improved according to specific circumstances to improve the accuracy and precision of object detection; Determination of sensitive area boundaries: The method for analyzing the boundary coordinates of a spatial region based on the Yolov5 algorithm is as follows: Obtain image data, input the image into the Yolov5 algorithm, and obtain the upper left corner coordinates (x1, y1) and width w and height h of the target detection box; According to the upper left corner coordinates (x1, y1) and width w and height h of the detection box, the upper left corner boundary coordinates (x1, y1) and lower right corner boundary coordinates (x2, y2) of the target in the image can be determined. The calculation formula is: x2 = x1 + w, y2 = y1 + h; The center coordinates (x1, y1) of the target detection frame can be used as the center coordinates of the target in the image; Intrusion detection: When the coordinates of an object fall within the boundaries of a sensitive area, an alarm is immediately issued, and the LBP feature extraction algorithm is used to record its physical features. The camera is then called to continuously track and record it. If the object does not invade the sensitive area, subsequent behavior determination is performed.

4. The method for identifying a behavior analysis intelligent monitoring system according to claim 2, characterized in that: The specific method of behavior analysis and determination is as follows: based on the method of steps S2 and S3, comprehensively determine whether the object is a human body or a non-human body; When the object is not a human body, continue to monitor its coordinates to see if it intrudes into sensitive areas; If the object is a human body, the system will continue to monitor its coordinates to see if it intrudes into sensitive areas while analyzing and judging its behavior. The analysis and judgment specifically uses a trained AI model to identify human behavior and cooperates with the LBP feature extraction algorithm to perform position analysis on people outside the sensitive area: human behavior is divided into normal behavior and abnormal behavior based on different behavioral characteristics; Among them, feature extraction is the most critical step in the whole process, and its effect directly affects the accuracy and reliability of behavior recognition; Generally speaking, the extracted features should include as much information as possible about the body's motion, posture, and speed. In practical applications, the model needs to be customized and optimized based on specific scenarios and application requirements. Based on the above-mentioned limit recognition and judgment of human behavior, when human behavior is abnormal, such as committing destructive behavior, the LBP feature extraction algorithm is used to record its physical features, and the camera is called to continuously track it and issue an alarm.

5. A method for identifying behavior analysis in an intelligent monitoring system according to any one of claims 1 to 4, characterized in that: The alerts include: Screen alarm: When the detector senses a hidden danger, the alarm system will automatically pop up an alarm screen and display it in a conspicuous position, notifying the online operator by flashing the screen and text prompts; Sound and light alarm: When an alarm event occurs, the buzzer and LED sound and light flashing are used to alert the on-site personnel to remind the on-site staff on duty; SMS alarm: set the SMS alarm mobile phone number in the background, and send the collected alarm information to the alarm host, which will then send it to the mobile phone number; Email alarm: The alarm method is the same as SMS alarm. Fill in the email address in the notification contact column of the management background. When an alarm situation occurs, the alarm information will be automatically sent to the set email address; Mobile APP alarm: Only by activating the mobile APP function in the management background can real-time information be provided to users. Once there is an alarm information on site, the system will immediately push a notification prompt.

Images