Automobile industry live broadcast real-time image extraction system and method based on AI technology

Abstract

The invention relates to the technical field of real-time image processing, in particular to an automobile industry live broadcast real-time image extraction system and method based on the AI technology, and the system comprises a contour recognition module, a dynamic energy adjustment module, a background updating module, an information extraction module, a behavior pattern recognition module and a real-time data output module. According to the invention, by analyzing the live video stream in the automobile industry in real time, the position coordinate and contour of the automobile can be calculated and adjusted more accurately, the rapid change of the environment, such as the influence of light, weather or emergencies, can be dealt with more flexibly, the image quality can be improved, and the dynamic foreground of the moving object can be tracked and analyzed more meticulously. The system is allowed to capture smaller motion differences, more accurate data support and deep analysis of dynamic prospects are provided in applications such as automatic driving assistance and remote diagnosis, richer and more useful information is provided for users, the application value of the system is enhanced, and the operation efficiency and safety are greatly improved.

Description

Technical Field

[0001] The present invention relates to the field of real-time image processing technology, and in particular to a live broadcast real-time image extraction system and method for the automotive industry based on AI technology. Background Art

[0002] Real-time image processing is a dynamic field of technology that focuses on the instant analysis and modification of video or image data, using efficient algorithms and hardware acceleration technology to process and parse image data with minimal latency, with a focus on image enhancement, feature extraction, object recognition, scene understanding, and real-time feedback. It has a wide range of applications, from monitoring systems, autonomous driving, medical diagnosis to media entertainment and other industries. Real-time image processing technology enables the system to complete the necessary analysis and response at the moment of receiving image data.

[0003] Among them, the live broadcast real-time image extraction system for the automotive industry based on AI technology is an advanced application that quickly and accurately extracts key image information by analyzing live broadcast videos of the automotive industry in real time, and uses deep learning, computer vision and pattern recognition technology to identify and process key images such as vehicles, driving behaviors, and road conditions. Uses include real-time monitoring, autonomous driving assistance, remote diagnosis and traffic management. Enterprises can improve operational efficiency and enhance safety while providing users with high-quality visual data analysis services.

[0004] Existing real-time image processing technology has been widely used in multiple industries, but it still faces some challenges when processing high-speed moving objects. For example, in autonomous driving and traffic management applications, existing technologies have difficulty processing complex environmental changes at extremely low latency, such as sudden weather changes or rapid changes in lighting conditions, which will affect the clarity and accuracy of the image, lead to errors in data analysis, and increase the risk of accidents. Existing technologies are not flexible enough in energy management, and it is difficult to adjust the energy distribution of the image in real time to adapt to environmental changes, which limits the application effect in dynamic environments. For example, when a vehicle enters a bright light area from a shadow, the outline of the image will become unclear due to overexposure, affecting the judgment and response of the entire system. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and propose a live broadcast real-time image extraction system and method for the automotive industry based on AI technology.

[0006] In order to achieve the above purpose, the present invention adopts the following technical solutions. The live broadcast real-time image extraction system for the automotive industry based on AI technology includes:

[0007] The contour recognition module analyzes the real-time video stream of the automotive industry live broadcast, uses AI technology to identify and calculate the position coordinates of the car, and adjusts the car contour according to the pixel changes of the real-time image to generate a contour data set;

[0008] The dynamic energy adjustment module adjusts the internal energy to maintain the smoothness of the contour line based on the contour data set, and generates adjusted energy parameters by adjusting the external energy to match the dynamic changes of the contour of the moving car;

[0009] The background update module iteratively optimizes the adjusted energy parameters, identifies and tracks moving objects in the live real-time image of the automotive industry, including vehicles and pedestrians, and generates dynamic foreground analysis results;

[0010] The information extraction module uses AI technology to extract moving images from the live broadcast of the automotive industry based on the dynamic prospect analysis results, calculate the direction change and speed fluctuation of the vehicle movement, and generate motion feature information;

[0011] The behavior pattern recognition module uses AI technology to analyze the real-time behavior patterns of cars in the live broadcast of the automotive industry based on the motion feature information and generates predictive behavior analysis results;

[0012] The real-time data output module updates the dynamic behavior of the automobile in real time and generates behavior trend update data based on the predicted behavior analysis results by monitoring and synchronizing the behavior trends in the live broadcast of the automobile industry.

[0013] As a further solution of the present invention, the contour data set includes the initial contour coordinates of the vehicle, the change records in the contour adjustment and the vehicle contour shape, the adjusted energy parameters include internal energy parameters, external energy parameters and the number of iterations of energy adjustment, the dynamic foreground analysis results include the position, movement trajectory and behavior pattern of the moving object in the live image, the motion feature information includes the speed vector, direction change rate and motion trajectory image of the car, the predicted behavior analysis results include the prediction of behavior patterns in future time periods, analysis of real-time behavior trends and identification of potential risk behaviors, and the behavior trend update data includes time series updates of behavior patterns and synchronous updates of dynamic behavior changes.

[0014] As a further solution of the present invention, the contour recognition module includes:

[0015] The video stream analysis submodule uses the real-time video stream of the automotive industry live broadcast to analyze the content of the live real-time image, track and locate the contour position of the car, and identify the contour features of the car to generate the car position information;

[0016] The coordinate calculation submodule records and calculates the vehicle position coordinates through the vehicle position information, analyzes the change trend of the vehicle coordinates, converts them into a data format, and obtains a coordinate data set;

[0017] The contour adjustment submodule compares the pixel point changes of the live real-time image of the automobile industry based on the coordinate data set, optimizes the matching degree of the automobile contour, dynamically adjusts the image to align with the real-time automobile contour, and generates a contour data set.

[0018] As a further solution of the present invention, the dynamic energy adjustment module includes:

[0019] The energy analysis submodule uses the contour data set to extract the characteristics of the automobile contour line of the live real-time image of the automobile industry, analyze the energy demand changes of contour recognition, compare the internal energy of the contour model with the demand of contour stability, and calculate the energy deviation to generate energy demand analysis results;

[0020] The energy output submodule adjusts the internal energy parameters of the contour model to maintain the smoothness of the contour line according to the energy demand analysis result, monitors the influence of the external energy of the contour model, and obtains the matching energy data set by adjusting the energy output in real time to match the dynamic contour changes of the moving car;

[0021] The parameter adjustment submodule optimizes the setting and response speed of the energy parameters of the contour model based on the matching energy data set, and generates adjusted energy parameters by analyzing the real-time changing moving image of the car.

[0022] As a further solution of the present invention, the background update module includes:

[0023] The image comparison submodule iteratively analyzes the adjusted energy parameters, matches the real-time changes of the live broadcast environment of the automotive industry by comparing the real-time image with the historical live broadcast image data, and generates an energy distribution map;

[0024] The object analysis submodule analyzes the energy characteristics of the moving objects in the real-time image through the energy distribution diagram, identifies the vehicles and pedestrians in the image by combining the shape and motion information in the image, and generates a moving object recognition result;

[0025] The dynamic tracking submodule utilizes the mobile object recognition result and adopts the Kalman filter algorithm to perform state estimation and position prediction of the mobile object to generate a dynamic foreground analysis result.

[0026] As a further solution of the present invention, the formula of the Kalman filter algorithm is as follows:

[0027]

[0028] in, is the state estimate at time k, is the predicted state before time k, K k is the Kalman gain, z k is the observed value at time k, H kis the observation model, α and β are weight coefficients, and Δz are the differentials of the predicted state and the observed value, respectively.

[0029] As a further solution of the present invention, the information extraction module includes:

[0030] The feature recognition submodule adopts the dynamic prospect analysis results and uses AI technology to extract dynamic images in the live broadcast of the automotive industry, identify the motion trajectory and key frames of the car, and analyze the behavior patterns under differentiated motion states to generate a motion path map;

[0031] The motion analysis submodule analyzes the motion parameters of the vehicle through the motion path diagram, calculates the direction change and speed fluctuation of the vehicle, evaluates the change trend of the vehicle, and obtains speed fluctuation data;

[0032] The information aggregation submodule combines the speed fluctuation data and the direction change information of the car, analyzes the motion characteristics of the movement in the real-time image of the automobile industry live broadcast, arranges and optimizes the characteristics according to the automobile industry live broadcast time series, and generates motion characteristic information.

[0033] As a further solution of the present invention, the behavior pattern recognition module includes:

[0034] The behavior recording submodule uses the motion feature information and AI technology to analyze the real-time video stream of the automotive industry live broadcast, identify the behavior sequence of the car, and generate behavior classification data;

[0035] The pattern analysis submodule analyzes the frequency and pattern of the vehicle behavior through the behavior classification data, checks the regularity of the vehicle behavior, adjusts the recognition parameters of the image, matches the differentiated behavior pattern, and obtains the periodic analysis results;

[0036] The data comparison submodule analyzes the behavior patterns of automobile movement in the future time period based on the periodic analysis results and combines the data of historical live broadcast and real-time live broadcast images to generate a predicted behavior analysis result.

[0037] As a further solution of the present invention, the real-time data output module includes:

[0038] The dynamic change analysis submodule uses the predicted behavior analysis results to monitor the real-time behavior in the live broadcast of the automotive industry, capture key dynamic changes, track and mark the behavior patterns of the car, and generate real-time monitoring data;

[0039] The behavior monitoring submodule uses the real-time monitoring data to analyze the time series data of the vehicle behavior, and generates behavior synchronization data by calculating and comparing the behavior changes at different time points, marking the direction and speed fluctuations of the behavior;

[0040] The data updating submodule analyzes the trend of the automobile behavior based on the behavior synchronization data, checks the real-time nature of the data according to the real-time automobile behavior dynamics, and generates behavior trend update data.

[0041] The method for extracting live and real-time images of the automobile industry based on AI technology is performed based on the above-mentioned system for extracting live and real-time images of the automobile industry based on AI technology, and includes the following steps:

[0042] S1: Based on the real-time video stream of the automotive industry live broadcast, the video content is analyzed, the position data and pixel change information of the car are analyzed, the dynamic changes of the image are matched, and the contour data set is generated;

[0043] S2: According to the contour data set, adjusting the internal and external energy parameters of the contour model to maintain the smoothness of the contour line, matching the dynamic changes of the moving car contour, and adjusting the energy distribution to generate adjusted energy parameters;

[0044] S3: using the adjusted energy parameters, matching the real-time changes in the live broadcast image of the automotive industry, identifying and tracking the moving vehicles and pedestrians in the image, updating the dynamic prospect of the live broadcast scene of the automotive industry, and generating dynamic prospect analysis results;

[0045] S4: Based on the dynamic foreground analysis result, analyzing and calculating the motion characteristics of the moving object, including direction changes and speed fluctuations, optimizing the motion tracking of the live broadcast image of the automotive industry by comparing the motion parameters, and generating motion characteristic information;

[0046] S5: Based on the motion feature information, analyze the behavior pattern of the car, analyze the time series and behavior trend of the car's behavior, mark and classify the dynamic behavior of the car, and generate behavior trend update data.

[0047] Compared with the prior art, the advantages and positive effects of the present invention are:

[0048] In the present invention, by analyzing the video stream in real time, the position coordinates and outline of the car can be calculated and adjusted more accurately, and the rapid changes in the environment, such as the influence of light, weather or emergencies, can be more flexibly responded to. Dynamic energy adjustment further ensures the stability of the image outline, which is more important for objects moving at high speed. Not only does it improve the image quality, but it also reduces the probability of misidentification. By more carefully tracking and analyzing the dynamic prospects of moving objects, the system is allowed to capture more subtle motion differences and provide more accurate data support in applications such as automatic driving assistance and remote diagnosis. In-depth analysis of dynamic prospects, especially the detailed calculation of motion features and the identification of behavioral patterns, provides users with richer and more useful information, enhances the application value of the system, and by synchronously updating data, the system can adjust feedback in real time, greatly improving operational efficiency and safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 is a system flow chart of the present invention;

[0050] Figure 2 It is a schematic diagram of the system framework of the present invention;

[0051] Figure 3 It is a flow chart of the contour recognition module of the present invention;

[0052] Figure 4 It is a flow chart of the dynamic energy adjustment module of the present invention;

[0053] Figure 5 A flow chart of the background update module of the present invention;

[0054] Figure 6 It is a flow chart of the information extraction module of the present invention;

[0055] Figure 7 It is a flow chart of the behavior pattern recognition module of the present invention;

[0056] Figure 8 It is a flow chart of the real-time data output module of the present invention;

[0057] Fig. 9 It is a schematic diagram of the method steps of the present invention. DETAILED DESCRIPTION

[0058] 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.

[0059] 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.

[0060] Embodiment 1

[0061] See also Figure 1 to Figure 2 The present invention provides a technical solution, and the automotive industry live broadcast real-time image extraction system based on AI technology includes:

[0062] The contour recognition module analyzes the real-time video stream of the automotive industry live broadcast, uses AI technology to identify and calculate the position coordinates of the car, sets the initial contour line of the car, and adjusts the car contour according to the pixel changes of the real-time image to generate a contour data set;

[0063] The dynamic energy adjustment module calculates and adjusts the internal energy parameters of the contour model based on the contour data set to maintain the smoothness of the contour line. By adjusting the external energy of the contour model, it matches the dynamic changes of the contour of the moving car in the live broadcast real-time image of the automotive industry and generates the adjusted energy parameters.

[0064] The background update module iteratively optimizes the adjusted energy parameters, identifies and tracks moving objects in the live images of the automotive industry, including vehicles and pedestrians, and generates dynamic foreground analysis results;

[0065] The information extraction module uses AI technology to extract moving images from live broadcasts of the automotive industry based on the results of dynamic prospect analysis, calculate the direction changes and speed fluctuations of the vehicle's movement, and generate motion feature information;

[0066] The behavior pattern recognition module uses AI technology based on motion feature information to analyze the real-time behavior patterns of cars in live broadcasts of the automotive industry, predict the behavior trends of cars, identify potential changes in behavior patterns, and generate predictive behavior analysis results;

[0067] The real-time data output module updates the dynamic behavior of the car in real time, analyzes the periodicity of the car's behavior, and generates behavior trend update data based on the results of the predicted behavior analysis by monitoring and synchronizing the behavior trends in the live broadcast of the automotive industry.

[0068] The contour data set includes the initial contour coordinates of the vehicle, the change records in the contour adjustment and the vehicle contour shape. The adjusted energy parameters include internal energy parameters, external energy parameters and the number of iterations of energy adjustment. The dynamic prospect analysis results include the position, movement trajectory and behavior pattern of the moving object in the live image. The motion feature information includes the speed vector, direction change rate and motion trajectory image of the car. The predictive behavior analysis results include the prediction of behavior patterns in future time periods, analysis of real-time behavior trends and identification of potential risk behaviors. The behavior trend update data includes the time series update of behavior patterns and the synchronous update of dynamic behavior changes.

[0069] See also Figure 2 , 3 , the contour recognition module includes:

[0070] The video stream analysis submodule uses the real-time video stream of the automotive industry live broadcast to analyze the content of the live real-time image, track and locate the contour position of the car, and identify the contour features of the car. The execution process of generating the car position information is as follows;

[0071] The video stream analysis submodule uses the real-time video stream of the automotive industry live broadcast, initializes the video capture device, decodes the live stream, obtains frame-by-frame data, and uses image processing technology to process each frame of the image, including grayscale and noise filtering. Edge detection technology is applied to determine the rough outline of the car, and morphological operations such as corrosion and expansion are further used to refine the outline. Through template matching technology, the detected outline is compared with the pre-stored car model outline to identify and locate the specific position of the car. Repeat the execution until the video stream ends, generate position information and contour features for each identified car, analyze the content of the live real-time image, track and locate the contour position of the car, and identify the contour features of the car to generate the car position information. The formula is: Among them, S(i,j) is the local sensitivity score of pixel (i,j), I(k,l) is the grayscale value of the image at (k,l), μ and σ are the average grayscale value and standard deviation of the local window respectively, and δ is the parameter of the window size.

[0072] The coordinate calculation submodule records and calculates the vehicle position coordinates through the vehicle position information, analyzes the change trend of the vehicle coordinates, and converts them into data format. The execution process of obtaining the coordinate data set is as follows;

[0073] The coordinate calculation submodule uses the car position information to establish a coordinate system, and converts the position of the car in the video into coordinate points. It performs time series analysis on the coordinate points in consecutive frames, calculates the moving speed and direction of the car position, and uses dynamic window technology to update the real-time position of the car to cope with the fast-moving scenes in the live video. Each coordinate point is stored in the database and updated in real time. Through the data conversion program, the coordinate data is standardized and converted into a data format that can be used for further analysis. The car position coordinates are recorded and calculated, the change trend of the car coordinates is analyzed, and converted into a data format to obtain a coordinate data set. The formula is: C t =αC t-1 +(1-α)(x t ,y t ), where C T is the vehicle coordinate at time t, x t and t is the coordinate value of the current frame, and α is the smoothing coefficient.

[0074] The contour adjustment submodule compares the pixel changes of the live real-time image of the automotive industry based on the coordinate data set, optimizes the matching degree of the car contour, dynamically adjusts the image to align with the real-time car contour, and generates the contour data set. The execution process is as follows;

[0075] The contour adjustment submodule initializes the contour matching algorithm based on the coordinate data set, compares the car contour in the real-time video stream with the pre-stored standard car model, and adjusts the threshold and matching parameters according to the pixel changes to optimize the matching process. By analyzing each frame in the live video in real time, the parameters of the contour matching algorithm are dynamically adjusted to ensure the accuracy and stability of contour matching. The model is trained using machine learning methods to automatically adjust the deviation in the contour alignment process and continuously improve the matching degree. After each successful match, the contour data of the video frame is generated and stored in the contour data set to provide support for subsequent advanced applications such as autonomous driving, compare the pixel changes of the real-time images of the automotive industry, optimize the matching degree of the car contour, dynamically adjust the image and the real-time car contour alignment, and generate a contour data set. The formula is: Among them, M(d) is the adjusted matching degree, P(d) is the original matching degree, Q(d) is the weight factor, and w d is the weight of the contour difference, D is the maximum contour difference distance, and β is the weight of the adjustment influencing factor.

[0076] See also Figure 2 , 4 , the dynamic energy adjustment module includes:

[0077] The energy analysis submodule uses the contour data set to extract the characteristics of the car contour line of the live real-time image of the automobile industry, analyzes the energy demand changes of contour recognition, compares the internal energy of the contour model with the demand for contour stability, and calculates the energy deviation. The execution process of generating the energy demand analysis results is as follows;

[0078] The energy analysis submodule uses the contour data set to evaluate the energy demand of each contour point in the data set. It estimates the dynamic energy demand by analyzing the curvature and length of the car contour line in the live image. It uses the energy evaluation model to calculate the deviation between the theoretical energy and the actual detection energy of each contour point. It uses statistical methods such as standard deviation to analyze the consistency of energy deviation. It also considers the impact of external factors such as illumination and viewing angle changes on energy demand, and dynamically updates the energy demand model to more accurately reflect the energy required for the stability of the car contour and generate energy demand analysis results. The formula is: E(c) = α·A(c) + β·K(c), where E(c) is the energy demand of contour point c, A(c) is the area contribution of contour point c, K(c) is the curvature contribution of contour point c, and α and β are weight coefficients.

[0079] The energy output submodule adjusts the internal energy parameters of the contour model to maintain the smoothness of the contour line through the energy demand analysis results, monitors the influence of the external energy of the contour model, and adjusts the energy output in real time to match the dynamic contour changes of the moving car. The execution process of obtaining the matching energy data set is as follows;

[0080] The energy output submodule initializes the energy regulation device based on the energy demand analysis results, and adjusts the internal energy parameters of the contour model, such as current and voltage, according to the energy demand analysis results to maintain the smoothness of the contour line. A closed-loop control strategy is adopted to adjust the energy output of the contour model according to real-time feedback to ensure that the energy output matches the contour stability requirements. The external environmental changes, such as temperature and humidity, are monitored by sensors, and the energy output of the contour model is dynamically adjusted to resist the influence of external changes. Through a series of steps, the energy output process is continuously optimized to obtain a matching energy data set. The formula is: P(t) = γ·E(t) + δ·ΔE(t), where P(t) is the energy output at time t, E(t) is the energy demand calculated at time t, ΔE(t) is the rate of change of energy demand, and γ and δ are weight coefficients.

[0081] The parameter adjustment submodule optimizes the setting and response speed of the energy parameters of the contour model based on the matching energy data set, and generates the adjusted energy parameters by analyzing the real-time changing moving images of the car. The execution process is as follows;

[0082] The parameter adjustment submodule analyzes the effect of the energy parameter settings of the current profile model based on the matching energy data set, and identifies the profile model energy parameters that need to be adjusted based on the real-time moving image of the car. It uses optimization algorithms such as gradient descent to adjust the key parameters of the energy output of the profile model, such as power coefficient and response time, to improve the accuracy and response speed of energy adjustment. Through simulation and experimental data analysis, the preset parameter settings are verified to ensure the effectiveness and stability in practical applications, and the adjusted energy parameters are generated. The formula is: Θ(t) = η·P(t) + (1-η)·Θ(t-1), where Θ(t) is the adjusted parameter at time t, P(t) is the energy output at time t, Θ(t-1) is the parameter value at the previous moment, and η is a coefficient between 0 and 1.

[0083] See also Figure 2 , 5 , the background update module includes:

[0084] The image comparison submodule iteratively analyzes the adjusted energy parameters, and generates the energy distribution map by comparing the real-time image with the historical live image data to match the real-time changes in the live environment of the automotive industry. The execution process is as follows;

[0085] The image comparison submodule iteratively analyzes the adjusted energy parameters, sets the analysis framework, selects appropriate image processing algorithms to process the current real-time image, such as grayscale and edge enhancement, to improve the analysis accuracy, and uses image comparison algorithms, such as histogram comparison, to compare the current real-time image with the stored historical live images pixel by pixel, analyze and identify the difference in energy distribution between the real-time image and the historical image, which is caused by environmental changes such as lighting or weather conditions. According to the difference, the image processing parameters, such as contrast and brightness, are adjusted to adapt to the real-time environmental changes. Through the process of iteratively adjusting the image processing parameters, the real-time changes of the live environment are accurately matched to generate a detailed energy distribution map. The formula is: S(i,j) = ω·D(i,j)+(1-ω)·H(i,j), where S(i,j) is the similarity score between image frames i and j, D(i,j) is the direct pixel difference between image frames i and j, H(i,j) is the similarity between image frames i and j, and ω is the weight coefficient.

[0086] The object analysis submodule analyzes the energy characteristics of moving objects in real-time images through energy distribution diagrams, and identifies vehicles and pedestrians in images by combining shape and motion information in the images. The execution process of generating moving object recognition results is as follows;

[0087] The object analysis submodule uses the energy distribution map to initialize the object recognition algorithm and detect moving objects in the real-time image, including the location and analysis of high energy value areas. By combining the shape recognition technology in the image and the motion detection algorithm, such as optical flow technology, dynamic objects in the image are identified, mainly including vehicles and pedestrians. A detailed analysis of the energy characteristics of the object, such as the relationship between the energy peak area and the moving speed, and the relationship between the energy change and the shape of the object, helps to confirm the identity and status of each object in the image, and integrates all data to generate the mobile object recognition result. The formula is: O(x,y) = σ·E(x,y)+τ·M(x,y), where O(x,y) is the object recognition score at position (x,y), E(x,y) is the energy feature score, M(x,y) is the shape matching score, and σ and τ are weight coefficients.

[0088] The dynamic tracking submodule uses the mobile object recognition results and adopts the Kalman filter algorithm to perform state estimation and position prediction of the mobile object, and the execution process of generating dynamic prospect analysis results is as follows;

[0089] The dynamic tracking submodule uses the mobile object recognition results to set tracking parameters, such as the tracking window size and tracking speed. State estimation techniques, such as Kalman filtering, are applied to continuously estimate the position and speed of the identified moving objects. The state information at the previous moment and the current observation data are used to predict the state at the next moment, effectively dealing with measurement noise and temporary occlusion problems. Through the process of continuously iterating the prediction of the state at the next moment, the tracking parameters are dynamically adjusted to adapt to the motion changes of the object, ensuring the continuity and accuracy of the tracking, and generating detailed dynamic foreground analysis results.

[0090] The formula of the Kalman filter algorithm is as follows:

[0091]

[0092] in, is the state estimate at time k, is the predicted state before time k, K k is the Kalman gain, z k is the observed value at time k, H k is the observation model, α and β are weight coefficients, and Δz are the differentials of the predicted state and the observed value, respectively.

[0093] The execution flow is as follows:

[0094] State estimation based on the previous time point and the current observation z k Update the prediction and set the prediction error Correction, add differential adjustment of state and observation, state differential The observation differential Δz is obtained by calculating the change of two consecutive states or observations, multiplied by weight coefficients α and β respectively. The two coefficients are optimized by historical data analysis to minimize the estimation error and update the state estimate

[0095] See also Figure 2 , 6 , the information extraction module includes:

[0096] The feature recognition submodule uses the dynamic foreground analysis results and AI technology to extract dynamic images from live broadcasts of the automotive industry, identify the motion trajectory and key frames of the car, and analyze the behavior patterns under differentiated motion states. The execution process of generating a motion path diagram is as follows;

[0097] The feature recognition submodule uses the dynamic foreground analysis results, sets image analysis parameters, and introduces image processing algorithms to segment video frames and extract key frames. It uses edge detection and target tracking technology to identify the motion trajectory of the car from the dynamic foreground analysis results. By analyzing the changes in speed and acceleration in the car's motion trajectory, different behavior patterns such as acceleration, deceleration, and turning are identified. By comparing the differences between key frames, the dynamic changes of the car in a specific state are further analyzed to identify the transition point of the behavior pattern. Through comprehensive analysis, a motion path diagram describing the car's motion state and behavior pattern is generated. The formula is: R(n) = ξ·V(n) + ρ·C(n), where R(n) is the motion feature recognition score of the nth frame, V(n) is the motion speed analysis score of the vehicle in the nth frame, C(n) is the direction change analysis score of the vehicle in the nth frame, and ξ and ρ are weight coefficients.

[0098] The motion analysis submodule analyzes the motion parameters of the car through the motion path diagram, calculates the direction change and speed fluctuation of the car's motion, and evaluates the change trend of the car. The execution process of obtaining speed fluctuation data is as follows;

[0099] The motion analysis submodule initializes the motion parameter analysis model through the motion path diagram, and uses vector field analysis to quantitatively describe the direction change of the car's motion. By calculating the vector difference of the car's position between two consecutive key frames, the direction change information is obtained. At the same time, using time series analysis methods, such as sliding window technology, the speed standard deviation in each window is calculated to evaluate the volatility of the speed. It helps to understand the causes of the instability of the car's motion, such as driving behavior or road conditions, integrate the parameters, and generate detailed speed fluctuation data. The formula is: M(v,θ)=μ·Δv+ν·Δθ, where M(v,θ) is the motion analysis score based on speed v and direction θ, Δv is the change in speed, Δθ is the change in direction, and μ and ν are weight coefficients.

[0100] The information aggregation submodule combines the speed fluctuation data and the direction change information of the car to analyze the motion characteristics of the mobile in the real-time image of the automotive industry live broadcast, and arranges and optimizes the characteristics according to the automotive industry live broadcast time series. The execution process of generating motion characteristic information is as follows;

[0101] The information aggregation submodule combines speed fluctuation data and the direction change information of the car, integrates multi-source information through data fusion technology to enhance the accuracy and reliability of the data, and uses pattern recognition algorithms, such as cluster analysis, to identify typical motion patterns and abnormal patterns from the fused data. Through pattern analysis, the motion characteristics of the car in different environments and different driving behaviors are extracted. According to the time series data of the live broadcast of the automotive industry, the characteristics are time-aligned and optimized to ensure the timeliness and relevance of the information and generate motion feature information. The formula is: I(t) = ψ·S(t) + φ·D(t), where I(t) is the comprehensive motion information score at time point t, S(t) is the speed fluctuation analysis result at time point t, D(t) is the direction change analysis result at time point t, and ψ and φ are weight coefficients.

[0102] See also Figure 2 , 7 , the behavior pattern recognition module includes:

[0103] The behavior recording submodule uses motion feature information and applies AI technology to analyze the real-time video stream of the automotive industry live broadcast, identify the behavior sequence of the car, and generate the execution process of behavior classification data as follows;

[0104] The behavior recording submodule uses motion feature information to initialize the video analysis engine and configure relevant image recognition algorithms, such as convolutional neural networks (CNNs), which are specifically used to identify and classify behavior patterns in dynamic images. By analyzing the real-time video stream frame by frame, the behavior characteristics of the car, such as acceleration, deceleration, parking, turning, etc., are identified. According to the identified behavior, each behavior is assigned to a predefined classification, and the frequency and duration of each behavior are determined through statistical analysis to generate behavior classification data. The formula is: B(f) = θ·L(f)+(1-θ)·A(f), where B(f) is the behavior score of the f-th frame, L(f) is the position information analysis score of the f-th frame, A(f) is the activity level analysis score of the f-th frame, and θ is the weight factor.

[0105] The pattern analysis submodule analyzes the frequency and pattern of vehicle behavior through behavior classification data, checks the regularity of vehicle behavior, adjusts the recognition parameters of the image, matches the differentiated behavior pattern, and obtains the execution process of periodic analysis results as follows;

[0106] The pattern analysis submodule processes the behavior classification data, including normalization and denoising, to improve the accuracy of the analysis. It applies frequency analysis methods, such as fast Fourier transform (FFT), to identify the main frequency components in the behavior data, which helps to reveal the periodicity and regularity of the behavior. Based on the analysis results, the image recognition parameters, such as threshold and filter settings, are further adjusted to better adapt to different behavior patterns and changing conditions. Through the adjustment, the periodic analysis results are generated. The formula is: P(x) = κ·F(x)+(1-κ)·G(x), where P(x) is the pattern score of the car behavior x, F(x) is the frequency analysis score of x, G(x) is the regularity analysis score of behavior x, and κ is the weight coefficient.

[0107] The data comparison submodule analyzes the behavior of vehicles in the future time period based on the periodic analysis results and combines the data of historical live broadcast and real-time live broadcast images to generate the predicted behavior analysis results. The execution process is as follows;

[0108] The data comparison submodule integrates data based on the periodic analysis results, processes and compares the real-time live data with the historical live data in parallel, evaluates the time dependence and trend of the car behavior through time series analysis methods such as moving average or exponential smoothing, and applies machine learning algorithms such as random forest or support vector machine (SVM) to model the prediction model of car behavior. The prediction model predicts future behavior patterns based on historical and current behavior data and generates prediction behavior analysis results. The formula is: Q(t) = α·H(t)+β·I(t), where Q(t) is the comprehensive behavior prediction analysis score at time point t, H(t) is the historical data analysis score, I(t) is the instant behavior analysis score, and α and β are weight coefficients.

[0109] See also Figure 2 , 8 , the real-time data output module includes:

[0110] The dynamic change analysis submodule uses the predictive behavior analysis results to monitor the real-time behavior in the live broadcast of the automotive industry, capture key dynamic changes, track and mark the behavior patterns of the car, and generate the execution process of real-time monitoring data as follows;

[0111] The dynamic change analysis submodule uses the results of predictive behavior analysis to set real-time data monitoring parameters and deploy dynamic change capture algorithms, such as differential image analysis, to detect behavioral changes in video streams. By comparing images between consecutive frames, it quickly identifies behavioral change points, such as sudden stops or rapid changes in direction of the car, and focuses on monitoring and predicting behavioral patterns marked as high-risk or critical in the model. It tracks every behavioral change of the car in real time, marks and records it, and generates real-time monitoring data. The formula is: V(t) = π·X(t) + (1-π)·Y(t), where V(t) is the dynamic change analysis score at time point t, X(t) is the real-time analysis score, Y(t) is the expected behavior score, and π is the weight factor.

[0112] The behavior monitoring submodule uses real-time monitoring data to analyze the time series data of vehicle behavior, calculates and compares the behavior changes at different time points, marks the direction and speed fluctuations of the behavior, and generates the execution flow of behavior synchronization data as follows;

[0113] The behavior monitoring submodule uses real-time monitoring data to perform a more detailed time analysis of the car behavior data through advanced time series analysis techniques such as sliding window analysis, including calculating the time interval, duration and frequency between each behavior event, and comparing the data with historical behavior data. In particular, by comparing the behavior data at different time points, identifying the changing trend of the behavior pattern, such as speed increase or decrease, sharp change of direction, etc., it is used to mark the key turning points of the behavior, including significant fluctuations in direction and speed, and generate behavior synchronization data. The formula is: W(z) = ξ·U(z) + ζ·V(z), where W(z) is the monitoring score of z behavior, U(z) is the variability score, V(z) is the direction and speed fluctuation analysis score of the behavior, and ξ and ζ are weight coefficients.

[0114] The data update submodule analyzes the trend of vehicle behavior based on the behavior synchronization data, checks the real-time nature of the data according to the real-time vehicle behavior dynamics, and generates the execution flow of behavior trend update data as follows;

[0115] The data update submodule deploys trend analysis algorithms, such as time series prediction or regression analysis, based on behavior synchronization data to determine the future trend of vehicle behavior. Using the analysis results, it adjusts the real-time monitoring strategy to better respond to changes in vehicle behavior and ensure the real-time update and accuracy of data. By comparing the synchronization data with the real-time behavior data, anomalies or deviations in the data can be discovered in a timely manner, and the data processing process can be optimized to generate behavior trend update data. The formula is: Z(t) = λ·Q(t)+μ·R(t), where Z(t) is the data update score at time point t, Q(t) is the behavior trend score, R(t) is the data accuracy score, and λ and μ are weight coefficients.

[0116] See also Fig. 9 , the method for extracting live real-time images of the automobile industry based on AI technology, the method for extracting live real-time images of the automobile industry based on AI technology is executed based on the above-mentioned system for extracting live real-time images of the automobile industry based on AI technology, and includes the following steps:

[0117] S1: Based on the real-time video stream of the automotive industry live broadcast, the video content is analyzed, the position data and pixel change information of the car are analyzed, the dynamic changes of the image are matched, and the contour data set is generated;

[0118] S2: According to the contour data set, adjust the internal and external energy parameters of the contour model to maintain the smoothness of the contour line, match the dynamic changes of the moving car contour, and adjust the energy distribution to generate adjusted energy parameters;

[0119] S3: Use the adjusted energy parameters to match the real-time changes in the live broadcast images of the automotive industry, identify and track the moving vehicles and pedestrians in the images, update the dynamic foreground of the live broadcast scenes of the automotive industry, and generate dynamic foreground analysis results;

[0120] S4: Based on the dynamic foreground analysis results, the motion characteristics of the moving object are analyzed and calculated, including direction changes and speed fluctuations. By comparing the motion parameters, the motion tracking of the live broadcast images of the automotive industry is optimized and the motion characteristic information is generated;

[0121] S5: Based on the motion feature information, analyze the behavior pattern of the car, analyze the time series and behavior trend of the car's behavior, mark and classify the dynamic behavior of the car, and generate behavior trend update data.

[0122] 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. The real-time image extraction system for the automotive industry based on AI technology is characterized by: The system comprises: The contour recognition module analyzes the real-time video stream of the automotive industry live broadcast, uses AI technology to identify and calculate the position coordinates of the car, and adjusts the car contour according to the pixel changes of the real-time image to generate a contour data set; The dynamic energy adjustment module adjusts the internal energy of the contour model based on the contour data set to maintain the smoothness of the contour line, and generates adjusted energy parameters by adjusting the external energy of the contour model to match the dynamic contour changes of the moving car; The background update module iteratively optimizes the adjusted energy parameters, identifies and tracks moving objects in the live real-time image of the automotive industry, including vehicles and pedestrians, and generates dynamic foreground analysis results; The information extraction module uses AI technology to extract moving images from the live broadcast of the automotive industry based on the dynamic prospect analysis results, calculate the direction change and speed fluctuation of the vehicle movement, and generate motion feature information; The behavior pattern recognition module uses AI technology to analyze the real-time behavior patterns of cars in the live broadcast of the automotive industry based on the motion feature information and generates predictive behavior analysis results; The real-time data output module updates the dynamic behavior of the automobile in real time and generates behavior trend update data based on the predicted behavior analysis results by monitoring and synchronizing the behavior trends in the live broadcast of the automobile industry.

2. The AI-based real-time image extraction system for the automotive industry according to claim 1 is characterized in that: The contour data set includes the initial contour coordinates of the vehicle, the change records in the contour adjustment and the vehicle contour shape, the adjusted energy parameters include internal energy parameters, external energy parameters and the number of iterations of energy adjustment, the dynamic prospect analysis results include the position, movement trajectory and behavior pattern of the moving object in the live image, the motion feature information includes the speed vector, direction change rate and motion trajectory image of the car, the predicted behavior analysis results include the prediction of behavior patterns in future time periods, analysis of real-time behavior trends and identification of potential risk behaviors, and the behavior trend update data includes time series updates of behavior patterns and synchronous updates of dynamic behavior changes.

3. The AI-based automotive industry live broadcast real-time image extraction system according to claim 1 is characterized in that: The contour recognition module comprises: The video stream analysis submodule uses the real-time video stream of the automotive industry live broadcast to analyze the content of the live real-time image, track and locate the contour position of the car, and identify the contour features of the car to generate the car position information; The coordinate calculation submodule records and calculates the vehicle position coordinates through the vehicle position information, analyzes the change trend of the vehicle coordinates, converts them into a data format, and obtains a coordinate data set; The contour adjustment submodule compares the pixel point changes of the live real-time image of the automobile industry based on the coordinate data set, optimizes the matching degree of the automobile contour, dynamically adjusts the image to align with the real-time automobile contour, and generates a contour data set.

4. The AI-based real-time image extraction system for the automotive industry according to claim 1 is characterized in that: The dynamic energy adjustment module comprises: The energy analysis submodule uses the contour data set to extract the characteristics of the automobile contour line of the live real-time image of the automobile industry, analyze the energy demand changes of contour recognition, compare the internal energy of the contour model with the demand of contour stability, and calculate the energy deviation to generate energy demand analysis results; The energy output submodule adjusts the internal energy parameters of the contour model to maintain the smoothness of the contour line according to the energy demand analysis result, monitors the influence of the external energy of the contour model, and obtains the matching energy data set by adjusting the energy output in real time to match the dynamic contour changes of the moving car; The parameter adjustment submodule optimizes the setting and response speed of the energy parameters of the contour model based on the matching energy data set, and generates adjusted energy parameters by analyzing the real-time changing moving image of the car.

5. The AI-based automotive industry live broadcast real-time image extraction system according to claim 1 is characterized in that: The background update module comprises: The image comparison submodule iteratively analyzes the adjusted energy parameters, matches the real-time changes of the live broadcast environment of the automotive industry by comparing the real-time image with the historical live broadcast image data, and generates an energy distribution map; The object analysis submodule analyzes the energy characteristics of the moving objects in the real-time image through the energy distribution diagram, identifies the vehicles and pedestrians in the image by combining the shape and motion information in the image, and generates a moving object recognition result; The dynamic tracking submodule utilizes the mobile object recognition result and adopts the Kalman filter algorithm to perform state estimation and position prediction of the mobile object to generate a dynamic foreground analysis result.

6. The AI-based automotive industry live broadcast real-time image extraction system according to claim 5 is characterized in that: The formula of the Kalman filter algorithm is as follows: in, is the state estimate at time k, is the predicted state before time k, K k is the Kalman gain, z k is the observed value at time k, H k is the observation model, α and β are weight coefficients, and Δz are the differentials of the predicted state and the observed value, respectively.

7. The AI-based automotive industry live broadcast real-time image extraction system according to claim 1 is characterized in that: The information extraction module comprises: The feature recognition submodule adopts the dynamic prospect analysis results and uses AI technology to extract dynamic images in the live broadcast of the automotive industry, identify the motion trajectory and key frames of the car, and analyze the behavior patterns under differentiated motion states to generate a motion path map; The motion analysis submodule analyzes the motion parameters of the vehicle through the motion path diagram, calculates the direction change and speed fluctuation of the vehicle, evaluates the change trend of the vehicle, and obtains speed fluctuation data; The information aggregation submodule combines the speed fluctuation data and the direction change information of the car, analyzes the motion characteristics of the movement in the real-time image of the automobile industry live broadcast, arranges and optimizes the characteristics according to the automobile industry live broadcast time series, and generates motion characteristic information.

8. The AI-based automotive industry live broadcast real-time image extraction system according to claim 1 is characterized in that: The behavior pattern recognition module includes: The behavior recording submodule uses the motion feature information and AI technology to analyze the real-time video stream of the automotive industry live broadcast, identify the behavior sequence of the car, and generate behavior classification data; The pattern analysis submodule analyzes the frequency and pattern of the vehicle behavior through the behavior classification data, checks the regularity of the vehicle behavior, adjusts the recognition parameters of the image, matches the differentiated behavior pattern, and obtains the periodic analysis results; The data comparison submodule analyzes the behavior patterns of automobile movement in the future time period based on the periodic analysis results and combines the data of historical live broadcast and real-time live broadcast images to generate a predicted behavior analysis result.

9. The AI-based automotive industry live broadcast real-time image extraction system according to claim 1 is characterized in that: The real-time data output module comprises: The dynamic change analysis submodule uses the predicted behavior analysis results to monitor the real-time behavior in the live broadcast of the automotive industry, capture key dynamic changes, track and mark the behavior patterns of the car, and generate real-time monitoring data; The behavior monitoring submodule uses the real-time monitoring data to analyze the time series data of the vehicle behavior, and generates behavior synchronization data by calculating and comparing the behavior changes at different time points, marking the direction and speed fluctuations of the behavior; The data updating submodule analyzes the trend of the automobile behavior based on the behavior synchronization data, checks the real-time nature of the data according to the real-time automobile behavior dynamics, and generates behavior trend update data.

10. A method for extracting live real-time images in the automotive industry based on AI technology, characterized in that: According to any one of claims 1 to 9, the live broadcast real-time image extraction system for the automotive industry based on AI technology is implemented, comprising the following steps: Based on the real-time video stream of the automotive industry live broadcast, the video content is analyzed, the position data and pixel change information of the car are analyzed, the dynamic changes of the image are matched, and the contour data set is generated; According to the contour data set, adjusting internal and external energy parameters of the contour model to maintain the smoothness of the contour line, matching the dynamic changes of the contour of the moving car, and adjusting the energy distribution to generate adjusted energy parameters; Using the adjusted energy parameters, matching the real-time changes in the live broadcast image of the automotive industry, identifying and tracking the moving vehicles and pedestrians in the image, updating the dynamic prospect of the live broadcast scene of the automotive industry, and generating dynamic prospect analysis results; Based on the dynamic foreground analysis results, the motion characteristics of the moving object are analyzed and calculated, including direction changes and speed fluctuations, and the motion tracking of the live broadcast image of the automotive industry is optimized by comparing the motion parameters to generate motion characteristic information; Based on the motion feature information, the behavior pattern of the car is analyzed, the time series and behavior trend of the car's behavior are analyzed, the dynamic behavior of the car is marked and classified, and behavior trend update data is generated.

Images