Intelligent video monitoring method and system

Abstract

The present invention relates to the field of security monitoring technology, specifically to an intelligent video monitoring method and system, comprising the following steps: extracting pixel intensity changes and color histograms of each frame from a real-time video stream, calculating the information entropy value of each frame, analyzing pixel changes and color distribution, and generating information entropy data. In the present invention, by analyzing the information entropy data generated by the pixel intensity changes and color histograms of video frames in real time, frames with high information entropy values ​​are automatically screened out, so that abnormal behaviors or unconventional activities can be located more quickly. This processing logic reduces the dependence on continuous monitoring, reduces the consumption of human resources, and significantly improves the response speed and accuracy. By extracting the specific motion trajectory of dynamic objects and the relationship between multiple objects from the key frame set, a dynamic space-time graph is further constructed, which not only increases the adaptability of the monitoring system to complex scenes, but also provides data support for early warning of events.

Description

Technical Field

[0001] The present invention relates to the field of security monitoring technology, and in particular to an intelligent video monitoring method and system. Background Art

[0002] The field of security monitoring technology covers the use of various sensors and monitoring equipment to achieve real-time monitoring and management of public safety, private property and sensitive areas. This technology usually includes cameras, sensors, alarm systems and networked monitoring centers. These systems can capture video images and analyze image content through advanced algorithms to detect and respond to abnormal behavior or security threats. With the development of artificial intelligence and machine learning, security monitoring systems have been able to achieve more intelligent functions, such as automatic facial recognition, abnormal behavior detection and automatic alarm. In addition, these systems can also integrate data analysis functions to help security personnel conduct risk assessment and security management more effectively.

[0003] Among them, intelligent video surveillance methods refer to the technology of using video cameras to capture images and using software algorithms to analyze the images in real time. The main uses of this method include crime prevention, traffic monitoring, crowd management and accident prevention. Through intelligent video surveillance, the system can automatically identify suspicious behavior, track targets or detect specific events, thereby providing real-time response without the need for human intervention. This greatly improves the efficiency and security level of monitoring, which is particularly important in large-scale or high-risk environments.

[0004] Although existing technologies are widely used in many occasions, when dealing with large-scale or complex scenarios, such as large-scale events or busy traffic areas, they are often unable to accurately identify and track abnormal behaviors in a small area due to the limitations of basic content analysis functions. This limitation leads to insufficient behavior analysis in dense crowds, which may lead to untimely identification and response to key behaviors, increasing the difficulty and risk of safety management. In addition, existing systems mostly rely on post-manual intervention, and lack the ability to respond in real time and predict future scene changes, limiting their potential in preventing and controlling potential safety issues. If potential dangerous behaviors are not identified in a timely manner, it may lead to the inability to effectively prevent accidents, thereby affecting the overall safety prevention and control effect. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose an intelligent video monitoring method and system.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: an intelligent video monitoring method, comprising the following steps:

[0007] S1: Extract the pixel intensity change and color histogram of each frame from the real-time video stream, calculate the information entropy value of each frame, analyze the pixel change and color distribution, and generate information entropy data;

[0008] S2: Based on the information entropy data, identify frames whose information entropy values ​​exceed a set threshold, calculate the number of dynamic objects and motion intensity in each frame, screen out frames with abnormal behavior patterns and unconventional activities, and integrate them into a key frame set;

[0009] S3: extracting the position coordinates and motion trajectories of people and vehicles from the key frame set, calculating the moving speed and direction of multiple objects, comparing the position changes in consecutive frames, connecting the moving speed, direction and position change information, forming a spatial and temporal association, and constructing a dynamic space-time graph;

[0010] S4: performing multi-scale analysis on the dynamic spatiotemporal graph, calculating the short-term human behavior patterns and vehicle flow density, evaluating the occurrence frequency and location changes of abnormal behaviors, predicting scene changes, and generating behavior and flow density analysis results;

[0011] S5: Based on the behavior and flow density analysis results, analyze the correlation between the behavior pattern and the flow density data, identify abnormal behavior data points, integrate the behavior and flow data to build a simulation scenario, and generate scenario prediction results.

[0012] As a further solution of the present invention, the step of obtaining the information entropy data is specifically as follows:

[0013] S111: extracting images frame by frame from the real-time video stream, extracting color values ​​and intensity values ​​of pixels in each frame, generating a pixel intensity change matrix and a color histogram, and obtaining initial frame data;

[0014] S112: Based on the initial frame data, the color difference between each pixel and the previous frame is calculated, and the occurrence frequency of multiple colors in the current frame is counted to generate color difference data and color frequency data;

[0015] S113: Based on the color difference data and the color frequency data, the formula is used:

[0016]

[0017] Calculate the information entropy of each frame and generate information entropy data;

[0018] in, Represents the information entropy value of each frame, Represents the probability of a certain color in each frame, Represents the probability of the color difference from the previous frame, and are the weights of color frequency and difference.

[0019] As a further solution of the present invention, the step of acquiring the key frame set is specifically:

[0020] S211: Based on the information entropy data, by comparing the information entropy value of each frame with a set threshold, identifying frames whose information entropy values ​​exceed the threshold, and generating an over-threshold frame list;

[0021] S212: For each frame in the above-threshold frame list, the formula is used:

[0022]

[0023] Calculate the number and motion intensity of dynamic objects in the frame and generate dynamic data analysis results for each frame;

[0024] in, represents the dynamic data intensity of the i-th frame, is the information entropy value of the frame, is the set threshold, , , is the weight parameter that adjusts the calculation sensitivity;

[0025] S213: Using the dynamic data analysis results of each frame, screen frames including abnormal behavior patterns and unconventional activities, and integrate them into a key frame set.

[0026] As a further solution of the present invention, the steps of obtaining the dynamic spatiotemporal graph are specifically as follows:

[0027] S311: extracting the position coordinates of the person and the vehicle in each frame from the key frame set, recording the coordinates as initial position data, and obtaining an initial coordinate list;

[0028] S312: Based on the initial coordinate list, use the formula:

[0029]

[0030] Calculate the straight-line distance from the i-th frame to the i+1-th frame, and get the distance between each frame;

[0031] in, represents the straight-line moving distance from the i-th frame to the i+1-th frame, obtained by calculating the Euclidean distance of the object position between the two frames. is the coordinate position of the object in the i-th frame, representing the horizontal and vertical coordinates of the object in the frame in two-dimensional space, is the coordinate position of the object in the i+1th frame, representing the horizontal and vertical coordinates of the object in the two-dimensional space in the next frame;

[0032] S313: Combining the result of the distance between each frame with the timestamp data, analyzing the moving speed and direction of the object, connecting the speed and direction information, forming a temporal and spatial association, and constructing a dynamic spatiotemporal graph.

[0033] As a further solution of the present invention, the step of obtaining the scene prediction result is specifically:

[0034] S411: extracting multi-scale data from the dynamic spatiotemporal graph, preliminarily screening the behavior of people and the flow of vehicles based on the time window, extracting short-term behavior patterns and flow density data using a time series analysis method, and obtaining a preliminary behavior and flow pattern data set;

[0035] S412: Analyze the preliminary behavior and flow pattern data set using the formula:

[0036]

[0037] Calculate the frequency of occurrence of multiple patterns and generate frequency analysis results;

[0038] in, Indicates the number of occurrences of the target behavior or flow pattern in the sample. is the total number of samples, Representative The frequency of occurrence of the pattern;

[0039] S413: Integrate the frequency analysis results and dynamic spatiotemporal data to evaluate the occurrence frequency of abnormal behaviors and changes in associated locations, identify abnormal patterns through cluster analysis, connect data points to form a prediction model, and generate scenario prediction results.

[0040] An intelligent video monitoring system, the intelligent video monitoring system is used to execute the above intelligent video monitoring method, the system comprises:

[0041] The information entropy analysis module extracts pixel intensity changes and color histograms from real-time video streams, calculates the information entropy value of the frame, analyzes pixel changes and color distribution, and obtains an information entropy data set;

[0042] The dynamic object screening module screens frames whose information entropy values ​​exceed a set threshold based on the information entropy data set, calculates the number of dynamic objects and the intensity of motion, screens frames with abnormal behavior patterns and unconventional activities, and generates a key frame set;

[0043] The trajectory analysis module extracts the position coordinates and motion trajectories of people and vehicles from the key frame set, calculates the moving speed and direction of multiple objects, compares the position changes in consecutive frames, connects the moving speed, direction and position change information, and constructs a dynamic spatiotemporal association graph;

[0044] The behavior pattern analysis module performs multi-scale analysis on the dynamic spatiotemporal correlation graph, calculates the short-term human behavior pattern and vehicle flow density, evaluates the frequency and location changes of abnormal behaviors, predicts scene changes, and generates behavior and flow density analysis results;

[0045] The prediction result output module analyzes the correlation between the behavior pattern and the flow density data based on the behavior and flow density analysis results, identifies abnormal behavior data points, integrates the behavior and flow data to construct a simulation scenario, and generates a scenario prediction result.

[0046] Compared with the prior art, the advantages and positive effects of the present invention are:

[0047] In the present invention, by analyzing the information entropy data generated by the pixel intensity changes and color histograms of the video frames in real time, frames with high information entropy values ​​are automatically screened out, which usually represent significant activities in the scene, so that abnormal behaviors or unconventional activities can be located more quickly. This processing logic reduces the reliance on continuous monitoring, reduces the consumption of human resources, and significantly improves the response speed and accuracy. By extracting the specific motion trajectories of dynamic objects and the relationships between multiple objects from the key frame set, a dynamic space-time graph is further constructed. This not only increases the adaptability of the monitoring system to complex scenes, but also provides data support for early warning of events, allowing security managers to take measures before problems occur, thereby improving the overall safety management efficiency and preventiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic diagram of the workflow of the present invention;

[0049] Figure 2 It is a flow chart of the steps of obtaining information entropy data of the present invention;

[0050] Figure 3 is a flow chart of the steps for obtaining a key frame set of the present invention;

[0051] Figure 4 A flowchart of the steps for obtaining the dynamic space-time graph of the present invention;

[0052] Figure 5 The present invention is a flowchart of the steps for obtaining the scenario prediction results. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0054] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships 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, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0055] Embodiment 1

[0056] See also Figure 1 The present invention provides a technical solution: an intelligent video monitoring method, comprising the following steps:

[0057] S1: Extract the pixel intensity change and color histogram of each frame from the real-time video stream, calculate the information entropy value of each frame, analyze the pixel change and color distribution, and generate information entropy data;

[0058] S2: Based on the information entropy data, identify frames whose information entropy values ​​exceed the set threshold, calculate the number of dynamic objects and motion intensity in each frame, filter out frames with abnormal behavior patterns and unconventional activities, and integrate them into a key frame set;

[0059] S3: Extract the position coordinates and motion trajectories of people and vehicles from the key frame set, calculate the moving speed and direction of multiple objects, compare the position changes in consecutive frames, connect the moving speed, direction and position change information, form spatial and temporal associations, and construct a dynamic space-time graph;

[0060] S4: Perform multi-scale analysis on dynamic spatiotemporal graphs, calculate short-term human behavior patterns and vehicle flow density, evaluate the frequency and location changes of abnormal behaviors, predict scene changes, and generate behavior and flow density analysis results;

[0061] S5: Based on the behavior and flow density analysis results, analyze the correlation between the behavior pattern and the flow density data, identify abnormal behavior data points, integrate the behavior and flow data to build a simulation scenario, and generate scenario prediction results.

[0062] Information entropy data includes pixel intensity changes, color histograms, and information entropy values. The key frame sets specifically include the number of dynamic objects, motion intensity, abnormal behavior patterns, and unconventional activities. The dynamic space-time diagram includes position coordinates, motion trajectory, moving speed, direction information, and space-time correlation. The scene prediction results specifically refer to character behavior patterns, vehicle flow density, abnormal behavior frequency, and position change predictions.

[0063] See also Figure 2 , the specific steps for obtaining information entropy data are:

[0064] S111: extracting images frame by frame from the real-time video stream, extracting color values ​​and intensity values ​​of pixels in each frame, generating a pixel intensity change matrix and a color histogram, and obtaining initial frame data;

[0065] Images are extracted frame by frame from the real-time video stream, and color and intensity data are extracted based on each frame. The color value and intensity value of each pixel are captured by the sensor to ensure the accuracy of data collection. The pixel data of each frame is analyzed by an automated script to generate two matrices: one records the pixel intensity change and the other records the color histogram. These data provide the basis for subsequent analysis to obtain the initial frame data.

[0066] S112: Based on the initial frame data, the color difference between each pixel and the previous frame is calculated, and the occurrence frequency of multiple colors in the current frame is counted to generate color difference data and color frequency data;

[0067] Based on the initial frame data, the color difference and color distribution frequency of each frame of the image are analyzed. By comparing the pixel differences between consecutive frames, the differences are calculated through image processing, and color frequency statistics are performed. The statistical method includes the number of occurrences and frequency of each color to ensure the breadth and depth of the analysis. The processing process of color data is optimized through data processing algorithms to generate color difference data and color frequency data.

[0068] S113: Starting from the color difference data and the color frequency data, the formula is used:

[0069]

[0070] Calculate the information entropy of each frame and generate information entropy data;

[0071] in, Represents the information entropy value of each frame, Represents the probability of a certain color in each frame, Represents the probability of the color difference from the previous frame, and are the weights of color frequency and difference.

[0072] formula:

[0073]

[0074] The usefulness of the formula is that by introducing the color frequency and color difference frequency The weighted sum of , evaluates the information entropy, where and The adjustment optimizes the entropy calculation according to the dynamic changes of the actual video content, thereby reflecting the richness of the information content;

[0075] Detailed explanation of the formula and the process of formula calculation and derivation:

[0076] In a target frame, it is assumed that red pixels account for 40% and blue pixels account for 60%. The red pixels increase by 5% from the previous frame, and the blue pixels decrease by 10%. , , , , weight , .

[0077] Calculate the weight of each color and the corresponding information entropy:

[0078]

[0079]

[0080] The total information entropy is

[0081] The calculation results show:

[0082]

[0083]

[0084] Total information entropy

[0085] The results show that the calculated information entropy value of 0.866 can quantify the complexity and information content of this frame of video, indicating the richness of the video content.

[0086] See also Figure 3 , the steps for obtaining the key frame set are as follows:

[0087] S211: Based on the information entropy data, by comparing the information entropy value of each frame with a set threshold, identifying frames whose information entropy values ​​exceed the threshold, and generating an over-threshold frame list;

[0088] In the analysis based on information entropy data, it is first necessary to extract the amount of information of each frame from the video stream data and express the amount of information in the form of entropy value. This process requires the calculation of the pixel distribution of each frame. The color change of each pixel will be converted into numerical information. By accumulating information, an information entropy value is generated for each frame. This value reflects the complexity of the frame image or the amount of information. This process involves pixel analysis and information conversion algorithms. The changes of each pixel will be calculated to ensure that the entropy value obtained reflects the actual situation as much as possible. By comparing according to the set threshold, frames with more information than the normal level can be effectively screened. This frame often includes more dynamic changes or key information.

[0089] S212: For each frame in the super-threshold frame list, use the formula:

[0090]

[0091] Calculate the number and motion intensity of dynamic objects in the frame and generate dynamic data analysis results for each frame;

[0092] in, represents the dynamic data intensity of the i-th frame, is the information entropy value of the frame, is the set threshold, , , is the weight parameter that adjusts the calculation sensitivity;

[0093] formula:

[0094]

[0095] The benefit of the formula is that it not only refers to the information entropy value itself, but also by introducing a threshold and weight parameters To adjust the sensitivity to information entropy anomalies, this allows the formula to be flexibly adjusted in differentiated situations to match the needs of various dynamic scenarios.

[0096] Detailed explanation of the formula and the process of formula calculation and derivation:

[0097] Set the information entropy value measured for a video frame The threshold is set to 0.8. is 0.5. Select the weight parameter , and calculate:

[0098]

[0099]

[0100]

[0101]

[0102] The results show that the entropy value after weight adjustment exceeds the set threshold, which indicates that the frame includes potential critical information or abnormal activities and should be marked and analyzed iteratively.

[0103] S213: Using the dynamic data analysis results of each frame, screen frames including abnormal behavior patterns and unconventional activities and integrate them into a key frame set.

[0104] The marked and iteratively analyzed frames include the calculation of motion intensity and the identification of dynamic objects. This process includes not only conventional frame comparison, but also edge detection and object tracking. By counting the number of dynamic objects in each frame and analyzing their motion intensity, it is possible to identify patterns of unconventional behavior or abnormal activities, sudden movements of an object, or batch movements at abnormal times. These data will be integrated into a key frame set for subsequent security monitoring or analysis processing to ensure the response speed and accuracy of the video surveillance system.

[0105] See also Figure 4 , the specific steps for obtaining the dynamic space-time graph are:

[0106] S311: extracting the position coordinates of people and vehicles in each frame from the key frame set, recording the coordinates as initial position data, and obtaining an initial coordinate list;

[0107] The process of extracting the position coordinates of people and vehicles in each frame from the key frame set requires first setting up an automatic recognition mechanism. The mechanism analyzes video frames through computer vision algorithms to detect and identify people and vehicles in the images. During this process, the algorithm will apply edge detection and shape recognition techniques to separate the images of people and vehicles and extract their position coordinates in each frame. The coordinate data is recorded and marked, and then used to iteratively analyze the dynamic trajectories of people and vehicles. The extraction of position coordinates is key because it directly affects the accuracy of motion trajectory analysis. In order to ensure the accuracy of the data, the video quality needs to be preprocessed, including denoising and image enhancement, so that people and vehicles can be identified more clearly. This series of operations ensures the accurate collection and effective use of position data, which is crucial for subsequent motion analysis and behavior recognition.

[0108] S312: Based on the initial coordinate list, use the formula:

[0109]

[0110] Calculate the straight-line distance from the i-th frame to the i+1-th frame, and get the distance between each frame;

[0111] in, represents the straight-line moving distance from the i-th frame to the i+1-th frame, obtained by calculating the Euclidean distance of the object position between the two frames. is the coordinate position of the object in the i-th frame, representing the horizontal and vertical coordinates of the object in the frame in two-dimensional space, is the coordinate position of the object in the i+1th frame, representing the horizontal and vertical coordinates of the object in the two-dimensional space in the next frame;

[0112] formula:

[0113]

[0114] The benefit of the formula is that by quantifying the position change between two frames, the moving distance of the object can be calculated, providing basic data for speed and direction analysis. The formula uses the Euclidean distance calculation method to simply and effectively convert image coordinates into actual moving distance.

[0115] Detailed explanation of the formula and the process of formula calculation and derivation:

[0116] Set the position coordinates of the object in two consecutive frames to be and , then the straight-line moving distance from the first frame to the second frame can be calculated by the Euclidean distance:

[0117] Calculate the x-axis distance difference:

[0118] Calculate the y-axis distance difference:

[0119] Apply the Pythagorean theorem to calculate straight-line distance:

[0120] In the monitoring data, the position of the first frame is (100,200) and the position of the second frame is (300,600), then:

[0121]

[0122]

[0123] unit

[0124] The results show that from the first frame to the second frame, the object moved a straight-line distance of approximately 447.21 units. This information is crucial for analyzing the object's movement speed and trajectory.

[0125] S313: Combine the result of the distance between each frame with the timestamp data, analyze the moving speed and direction of the object, connect the speed and direction information, form a temporal and spatial association, and construct a dynamic spatiotemporal graph.

[0126] The distance between each frame is used to combine with the timestamp to quantitatively analyze the speed of the object. The calculation of speed not only refers to the displacement in space, but also combines the time factor, that is, the displacement per unit time. The calculation formula of speed here involves , where S is the spatial displacement and t is the time interval. This formula is simple but very effective, and can directly reflect the speed of an object's movement. In order to improve the calculation accuracy, it is also necessary to verify and synchronize the accuracy of the timestamp to ensure that the time stamp of each frame is accurate. Through these processes, the dynamic motion state of each object can be analyzed, including its speed change and moving direction. These data will be directly used to generate a dynamic space-time diagram to provide support for behavior analysis and decision-making.

[0127] See also Figure 5 , the specific steps for obtaining the scene prediction results are:

[0128] S411: Extract multi-scale data from the dynamic spatiotemporal graph, perform preliminary screening of human behavior and vehicle flow based on the time window, and use time series analysis methods to extract short-term behavior patterns and flow density data to obtain preliminary behavior and flow pattern data sets;

[0129] The process of extracting multi-scale data from dynamic spatiotemporal graphs involves the control of time windows and the application of data screening algorithms. First, it is necessary to identify the behavior patterns of people and the flow density of vehicles through time series analysis methods. The execution of this step relies on the support of big data analysis technology and programming languages. Python or R can be used to process batches of frame data and identify the frequency and pattern of behavior. Then, statistical methods are used to preliminarily classify the extracted data. According to the differentiated characteristics of behavior and flow, the data is divided into several categories. Each category corresponds to a behavior pattern or flow density of a target. Through the classification of categories, it is easier to monitor and analyze pattern changes in the short term. This analysis not only helps to analyze the short-term trend of behavior patterns, but also predicts changes in behavior patterns in future time periods, thereby providing data support for subsequent urban planning and traffic management.

[0130] S412: Analyze the preliminary behavior and flow pattern data set using the formula:

[0131]

[0132] Calculate the frequency of occurrence of multiple patterns and generate frequency analysis results;

[0133] in, Indicates the number of occurrences of the target behavior or flow pattern in the sample. is the total number of samples, Representative The frequency of occurrence of the pattern;

[0134] formula:

[0135]

[0136] The benefit of the formula is that by calculating the frequency of occurrence of target behaviors or flow patterns, the prevalence and abnormality of the patterns can be quantitatively analyzed, thereby identifying abnormal behaviors or congestion points that require attention, which is particularly critical for security monitoring and urban management.

[0137] Detailed explanation of the formula and the process of formula calculation and derivation:

[0138] In the monitoring area of ​​a target, there are the following behavior records: walking (n1=150 times), running (n2=30 times), car movement (n3=300 times), and the total number of records N=480 times.

[0139] The walking frequency is calculated as

[0140] The frequency of running is

[0141] The frequency of car movement is .

[0142] The results show that most of the monitored behaviors are car movements, while running behaviors are less, indicating that the area has busy traffic but is not suitable for running. Such data analysis provides a scientific basis for urban planners to make decisions on facility layout and traffic control.

[0143] S413: Integrate frequency analysis results and dynamic spatiotemporal data to evaluate the frequency of abnormal behaviors and associated location changes, identify abnormal patterns through cluster analysis, connect data points to form a prediction model, and generate scenario prediction results.

[0144] In the process of identifying abnormal behaviors by clustering analysis methods by combining the frequency analysis results with spatiotemporal data, it is first necessary to standardize the behavioral data to eliminate the dimensional influence between differentiated behaviors so that the data can be compared under the same standard. Then the K-means clustering algorithm is applied to classify the behaviors. The selection of the K value in the algorithm is based on the intrinsic distribution characteristics of the data and can be determined by the elbow rule. Once the classification is completed, each category will correspond to a behavior pattern. By analyzing the density and distribution position of the data points in each category, abnormal behaviors or unconventional activities in the cluster can be effectively identified, including a sudden increase in behaviors or frequent changes in positions in a certain category. These are all abnormal signals. Through this analysis, a prediction map reflecting the dynamic changes of the current monitoring scene can be constructed to provide early warnings for events that are expected to occur in the future, thereby playing a key role in security monitoring and emergency response.

[0145] An intelligent video monitoring system, which is used to execute the above intelligent video monitoring method, comprises:

[0146] The information entropy analysis module extracts pixel intensity changes and color histograms from real-time video streams, calculates the information entropy value of the frame, analyzes pixel changes and color distribution, and obtains an information entropy data set;

[0147] The dynamic object screening module screens frames whose information entropy values ​​exceed the set threshold based on the information entropy data set, calculates the number of dynamic objects and the intensity of motion, screens frames with abnormal behavior patterns and unconventional activities, and generates a key frame set;

[0148] The trajectory analysis module extracts the position coordinates and motion trajectories of people and vehicles from the key frame set, calculates the moving speed and direction of multiple objects, compares the position changes in consecutive frames, connects the moving speed, direction and position change information, and constructs a dynamic spatiotemporal correlation graph;

[0149] The behavior pattern analysis module performs multi-scale analysis on dynamic spatiotemporal correlation graphs, calculates short-term human behavior patterns and vehicle flow density, evaluates the frequency and location changes of abnormal behaviors, predicts scene changes, and generates behavior and flow density analysis results;

[0150] The prediction result output module is based on the behavior and flow density analysis results, analyzes the correlation between behavior patterns and flow density data, identifies abnormal behavior data points, integrates behavior and flow data to build simulation scenarios, and generates scenario prediction results.

[0151] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. An intelligent video monitoring method, characterized in that: The following steps are involved: Extract the pixel intensity change and color histogram of each frame from the real-time video stream, calculate the information entropy value of each frame, analyze the pixel change and color distribution, and generate information entropy data; Based on the information entropy data, frames whose information entropy values ​​exceed a set threshold are identified, the number of dynamic objects and the intensity of motion in each frame are calculated, and frames with abnormal behavior patterns and unconventional activities are screened out to be integrated into a key frame set; Extracting the position coordinates and motion trajectories of people and vehicles from the key frame set, calculating the moving speed and direction of multiple objects, comparing the position changes in consecutive frames, connecting the moving speed, direction and position change information, forming a spatial and temporal association, and constructing a dynamic space-time graph; Perform multi-scale analysis on the dynamic spatiotemporal graph, calculate the short-term human behavior patterns and vehicle flow density, evaluate the frequency and location changes of abnormal behaviors, predict scene changes, and generate behavior and flow density analysis results; Based on the behavior and flow density analysis results, analyze the correlation between the behavior pattern and the flow density data, identify abnormal behavior data points, integrate the behavior and flow data to build a simulation scenario, and generate scenario prediction results; The steps of obtaining the key frame set are specifically as follows: Based on the information entropy data, by comparing the information entropy value of each frame with a set threshold, identifying frames whose information entropy values ​​exceed the threshold, and generating a list of super-threshold frames; For each frame in the above-threshold frame list, the formula is used: Calculate the number and motion intensity of dynamic objects in the frame and generate dynamic data analysis results for each frame; in, represents the dynamic data intensity of the i-th frame, is the information entropy value of the frame, is the set threshold, , , is the weight parameter that adjusts the calculation sensitivity; Frames including abnormal behavior patterns and unconventional activities are screened using the dynamic data analysis results of each frame and integrated into a key frame set.

2. The intelligent video monitoring method according to claim 1, characterized in that: The steps for obtaining the information entropy data are specifically as follows: Extract images frame by frame from the real-time video stream, extract the color value and intensity value of the pixels of each frame, generate a pixel intensity change matrix and a color histogram, and obtain the initial frame data; Based on the initial frame data, the color difference between each pixel and the previous frame is calculated, and the occurrence frequency of multiple colors in the current frame is counted to generate color difference data and color frequency data; Starting from the color difference data and color frequency data, the formula is adopted: Calculate the information entropy of each frame and generate information entropy data; in, Represents the information entropy value of each frame, Represents the probability of a certain color in each frame, Represents the probability of the color difference from the previous frame, and are the weights of color frequency and difference.

3. The intelligent video monitoring method according to claim 2, characterized in that: The steps for obtaining the dynamic space-time graph are specifically as follows: Extracting the position coordinates of the person and the vehicle in each frame from the key frame set, recording the coordinates as initial position data, and obtaining an initial coordinate list; Based on the initial coordinate list, the formula is used: Calculate the straight-line distance from the i-th frame to the i+1-th frame, and get the distance between each frame; in, represents the straight-line moving distance from the i-th frame to the i+1-th frame, obtained by calculating the Euclidean distance of the object position between the two frames. is the coordinate position of the object in the i-th frame, representing the horizontal and vertical coordinates of the object in the frame in two-dimensional space, is the coordinate position of the object in the i+1th frame, representing the horizontal and vertical coordinates of the object in the two-dimensional space in the next frame; The result of the distance between each frame is combined with the timestamp data, the moving speed and direction of the object are analyzed, the speed and direction information are connected, the association in time and space is formed, and a dynamic space-time diagram is constructed.

4. The intelligent video monitoring method according to claim 3, characterized in that: The steps for obtaining the scene prediction results are specifically as follows: Extract multi-scale data from the dynamic spatiotemporal graph, perform preliminary screening of human behavior and vehicle flow based on the time window, extract short-term behavior patterns and flow density data using a time series analysis method, and obtain a preliminary behavior and flow pattern data set; The preliminary behavioral and flow pattern data set was analyzed using the formula: Calculate the frequency of occurrence of multiple patterns and generate frequency analysis results; in, Indicates the number of occurrences of the target behavior or flow pattern in the sample. is the total number of samples, Representative The frequency of occurrence of the pattern; The frequency analysis results and dynamic spatiotemporal data are combined to evaluate the frequency of abnormal behaviors and associated position changes, identify abnormal patterns through cluster analysis, connect data points to form a prediction model, and generate scenario prediction results.

5. An intelligent video surveillance system, characterized in that: For implementing the intelligent video surveillance method according to any one of claims 1 to 4, the system comprises: The information entropy analysis module extracts pixel intensity changes and color histograms from real-time video streams, calculates the information entropy value of the frame, analyzes pixel changes and color distribution, and obtains an information entropy data set; The dynamic object screening module screens frames whose information entropy values ​​exceed a set threshold based on the information entropy data set, calculates the number of dynamic objects and the intensity of motion, screens frames with abnormal behavior patterns and unconventional activities, and generates a key frame set; The trajectory analysis module extracts the position coordinates and motion trajectories of people and vehicles from the key frame set, calculates the moving speed and direction of multiple objects, compares the position changes in consecutive frames, connects the moving speed, direction and position change information, and constructs a dynamic spatiotemporal association graph; The behavior pattern analysis module performs multi-scale analysis on the dynamic spatiotemporal correlation graph, calculates the short-term human behavior pattern and vehicle flow density, evaluates the frequency and location changes of abnormal behaviors, predicts scene changes, and generates behavior and flow density analysis results; The prediction result output module analyzes the correlation between the behavior pattern and the flow density data based on the behavior and flow density analysis results, identifies abnormal behavior data points, integrates the behavior and flow data to construct a simulation scenario, and generates a scenario prediction result.

Images