Video monitoring system suitable for expressway and capable of clearly distinguishing expressway running objects

Abstract

The invention relates to the technical field of video monitoring, in particular to a video monitoring system applicable to highways and capable of clearly distinguishing high-speed running objects, and the system comprises a key frame selection module, a form distortion analysis module, an image synchronous fusion module and a dynamic illumination adjustment module. According to the method, key frames in the expressway monitoring video are selected through difference measurement, content with significant change is accurately extracted, it is ensured that data processing is efficient and key points are highlighted, the form distortion problem caused by the speed of a high-speed moving object is effectively solved through form recognition and distortion correction, and the accuracy of form distortion is improved by adjusting the form data of the object. The consistency of the image and the actual state is ensured, the overall coordination of the multi-view image is improved through the synchronous fusion of time and space, the dynamic continuous monitoring of traffic flow is realized, the illumination difference in the video is balanced through real-time illumination adjustment, the image definition under the complex illumination condition is enhanced, and the image quality is improved. And the effectiveness of road safety and traffic management is obviously enhanced.

Description

Technical Field

[0001] The present invention relates to the field of video monitoring technology, and in particular to a video monitoring system suitable for expressways that can clearly identify high-speed running objects. Background Art

[0002] Video surveillance technology refers to the technology of capturing images with video cameras and processing, analyzing and storing them through software and hardware systems. It includes camera design and deployment, image transmission technology, video data compression and transmission, real-time or stored video analysis and other aspects. It is commonly used in security monitoring, traffic monitoring, industrial automation and other application scenarios. Its technical development involves advanced technologies such as high-definition image processing, machine vision, artificial intelligence and machine learning to improve the accuracy of image recognition and the practicality of the system. With the advancement of technology, video surveillance systems have been able to achieve efficient image capture, automatic event detection and multi-dimensional data analysis.

[0003] Among them, the video surveillance system suitable for highways that can distinguish high-speed running objects is mainly used in highway environments. It uses high-definition video surveillance to identify and track high-speed vehicles or other objects. The core purpose is to improve road safety, reduce traffic accidents through real-time monitoring, and optimize traffic flow management. In addition, it can also support accident analysis and emergency response, and help relevant departments respond in time through accurate object recognition and behavior analysis to ensure smooth and safe highway traffic.

[0004] The existing technology has the problem of insufficient extraction of key content in highway monitoring. Conventional video monitoring systems fail to effectively identify and distinguish important content in the video when processing data, resulting in low data processing efficiency. The capture of high-speed moving objects is easily affected by speed, and the image distortion problem is difficult to solve. Especially in the case of morphological distortion, the recognition and tracking accuracy of objects is significantly reduced. In multi-view monitoring technology, due to the lack of effective synchronization and fusion mechanisms, the monitoring screen is prone to time dislocation and spatial discontinuity, making it difficult to accurately present the overall state of traffic flow. In addition, in scenes with complex lighting conditions, traditional monitoring technology cannot dynamically balance lighting changes, resulting in unstable image quality, and shadow and highlight problems further reduce the reliability of monitoring. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a video monitoring system suitable for highways that can clearly identify high-speed running objects.

[0006] In order to achieve the above-mentioned object, the present invention adopts the following technical solution: A video monitoring system applicable to highways and capable of distinguishing high-speed running objects comprises:

[0007] The key frame selection module analyzes the content difference between each frame and the previous frame in the highway monitoring video stream, measures the difference, selects the frame with the most significant content change, determines that each selected frame represents a key action event in the target video, and generates a key frame set;

[0008] The morphological distortion analysis module performs morphological recognition and analysis on high-speed moving objects in the image based on the key frame set, compares the standard morphology of each object with the current image representation, analyzes the consistency between the visual representation and the actual state of each object according to the comparison result, corrects the image distortion caused by the speed, corrects the morphological error, and obtains the corrected morphological data;

[0009] Based on the corrected morphological data, the image synchronization fusion module hierarchically sorts the surveillance images captured by multiple cameras according to timestamps, performs spatial and temporal alignment, and seamlessly stitches multi-view videos according to the alignment results to reveal the continuous highway traffic flow, perform all-round viewing angle monitoring, and obtain a seamless panoramic monitoring view;

[0010] Based on the seamless panoramic monitoring view, the dynamic lighting adjustment module monitors and analyzes the lighting changes in different time periods in the current video, adjusts the image contrast in real time, balances the visual effects of shadow and highlight areas according to the adjustment results, maintains the visual consistency of the video image under changing lighting, and obtains the optimized monitoring video output result.

[0011] As a further solution of the present invention, the step of acquiring the key frame set is:

[0012] Analyze the highway monitoring video stream, extract each frame in the video and its previous frame, perform continuous frame correspondence, and obtain the original frame sequence set;

[0013] The pixel difference analysis and color histogram comparison are performed on the original frame sequence set, using the formula:

[0014]

[0015] Calculate the content difference value Δ between each pair of consecutive frames i , and obtain the difference measurement result, where n represents the total number of pixels in the frame, and f i,x and f i-1,x Represent the color values ​​of the corresponding pixels in the i-th frame and the i-1-th frame respectively;

[0016] According to the difference measurement result, a difference threshold is set, the frames with the most significant content changes are screened, and the representative key action events are determined to generate a key frame set.

[0017] As a further solution of the present invention, the comparison step between the standard form and the current image representation is:

[0018] Based on the key frame set, edge detection is performed, each frame is processed and the contour of the high-speed moving object in the image is identified to generate an object morphology contour data set;

[0019] The object shape contour data set is subjected to shape matching analysis and compared with the standard shape in the database using the formula:

[0020]

[0021] Calculate the morphological matching score S of each object and get the morphological matching result, where R i Represents the shape contour data points of the object, O i represents the data point corresponding to the standard form, and N represents the number of points involved in the comparison;

[0022] Using the morphological matching results, the deviation between the current image representation of each object and its standard morphology is evaluated using the formula:

[0023] D = max(1-S, 0);

[0024] Calculate the morphological deviation value D and obtain the morphological data comparison and analysis results.

[0025] As a further solution of the present invention, the step of acquiring the corrected morphological data is:

[0026] Based on the comparative analysis results of the morphological data, by comparing the image data with the standard model data, the formula is used:

[0027]

[0028] Calculate the visual consistency score V to obtain the visual consistency evaluation data, where S is the morphological matching score and T is the consistency threshold;

[0029] The image distortion metric is calculated using the visual consistency evaluation data using the formula:

[0030]

[0031] Calculate and output the morphology correction coefficient C d ;

[0032] Apply the morphology correction factor C d , correct the morphological error of each object, and generate corrected morphological data by adjusting the image data to match the actual state.

[0033] As a further solution of the present invention, the space and time alignment steps are:

[0034] Using the corrected morphological data, the surveillance images captured by multiple cameras are preliminarily sorted according to timestamps to confirm that the time sequence of each image is correct, and obtain timestamp sorting data;

[0035] Based on the timestamp sorting data, spatial alignment is performed, spatial overlapping areas and key alignment points between images are analyzed, continuity and accuracy of the images in space are checked, and spatial alignment data is generated;

[0036] The spatial alignment data is integrated with the timestamp sorting data, time alignment is performed, the consistency of the image sequence on the time axis is confirmed, and the spatial and temporal alignment results are obtained.

[0037] As a further solution of the present invention, the steps of acquiring the seamless panoramic monitoring view are:

[0038] Based on the spatial and temporal alignment results, multi-view video stitching is started, the image boundary output by the camera is processed, image edge blending is performed to smoothly transition the boundary area, and preliminary seamless video stitching data is generated;

[0039] Based on the preliminary seamless video stitching data, image color adjustment is performed to match the hue deviation in the camera's viewing angle, and exposure compensation is applied to balance the overall brightness, using the formula:

[0040]

[0041] Calculate the multiple C of the new color value of each pixel to the original color value i , and get the optimized multi-view spliced ​​video, where R i Represents the red component of the i-th pixel, G i Represents the green component of the i-th pixel, B i Represents the blue component of the i-th pixel;

[0042] The optimized multi-view stitched video is used for all-round monitoring, real-time analysis of video data, identification of traffic flow, vehicle speed and vehicle density information, automatic detection and annotation of key parameters of traffic flow, and acquisition of a seamless panoramic monitoring view.

[0043] As a further solution of the present invention, the steps of adjusting the image brightness and contrast are:

[0044] Based on the seamless panoramic monitoring view, a time series frame of the video is collected, and the light intensity of each frame is sampled to obtain light sampling data;

[0045] Apply time series analysis to the illumination sampling data to determine the difference in illumination intensity between consecutive frames using the formula:

[0046] ΔI t,t-1=|I t -I t-1 |;

[0047] Calculate the light intensity difference ΔI t,t-1 , generate the illumination change index, where I t and I t-1 Respectively represent the light intensity of two consecutive frames;

[0048] According to the illumination change index, the image contrast is readjusted using the formula:

[0049]

[0050] Calculate and output the adjusted new contrast C new , get the real-time image contrast adjustment result, where C old Indicates the previous image contrast.

[0051] As a further solution of the present invention, the steps for obtaining the optimized monitoring video output result are:

[0052] Based on the real-time image contrast adjustment result, identifying highlight and shadow areas in the video through image analysis, analyzing area boundaries and intensity changes, and generating highlight and shadow area analysis data;

[0053] Based on the highlight and shadow area analysis data, high dynamic range processing is applied to match visual requirements under different lighting conditions, balance light and dark details in the image, and generate HDR processed image data;

[0054] Based on the image data processed by HDR, local image sharpening and color correction are performed to ensure image consistency and natural transition of visual effects in all lighting environments, thereby generating optimized monitoring video output results.

[0055] Compared with the prior art, the advantages and positive effects of the present invention are:

[0056] In the present invention, key frames in highway monitoring videos are selected through difference measurement, and content with significant changes is accurately extracted to ensure efficient and focused data processing. Morphological recognition and distortion correction effectively solve the morphological distortion problem of high-speed moving objects caused by speed. By adjusting the object morphological data, the consistency between the image and the actual state is ensured. The synchronous fusion of time and space improves the overall coordination of multi-view images and realizes dynamic and continuous monitoring of traffic flow. Real-time lighting adjustment balances the lighting differences in the video and enhances the image clarity under complex lighting conditions. The overall process optimization from content extraction, morphological analysis, perspective fusion to dynamic adjustment is realized, which improves the recognition accuracy, fluency and adaptability of video monitoring in highway environments and significantly enhances the effectiveness of road safety and traffic management. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 is a system flow chart of the present invention;

[0058] Figure 2 is a flowchart of obtaining a key frame set of the present invention;

[0059] Figure 3 is a flow chart for comparing the standard form of the present invention with the current image representation;

[0060] Figure 4 is a flow chart for obtaining the corrected morphological data of the present invention;

[0061] Figure 5 It is a flow chart of the space and time alignment of the present invention;

[0062] Figure 6 A flowchart of obtaining a seamless panoramic monitoring view of the present invention;

[0063] Figure 7 The following is a flow chart of adjusting the image brightness and contrast of the present invention;

[0064] Figure 8 This is a flow chart for obtaining the optimized monitoring video output results of the present invention. DETAILED DESCRIPTION

[0065] 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.

[0066] 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.

[0067] See also Figure 1 A video monitoring system applicable to highways and capable of distinguishing high-speed running objects comprises:

[0068] The key frame selection module analyzes the content difference between each frame and the previous frame in the highway monitoring video stream, measures the difference, selects the frame with the most significant content change, determines that each selected frame represents a key action event in the target video, and generates a key frame set;

[0069] The morphological distortion analysis module performs morphological recognition and analysis on high-speed moving objects in the image based on the key frame set, compares the standard morphology of each object with the current image representation, analyzes the consistency between the visual representation and the actual state of each object based on the comparison results, corrects the image distortion caused by speed, corrects the morphological error, and obtains the corrected morphological data;

[0070] Based on the corrected morphological data, the image synchronization fusion module hierarchically sorts the surveillance images captured by multiple cameras according to timestamps, performs spatial and temporal alignment, and seamlessly stitches multi-view videos based on the alignment results to reveal the continuous highway traffic flow, perform all-round perspective monitoring, and obtain a seamless panoramic monitoring view;

[0071] The dynamic lighting adjustment module is based on a seamless panoramic monitoring view. It monitors and analyzes the lighting changes in different time periods in the current video, adjusts the image contrast in real time, balances the visual effects of shadow and highlight areas according to the adjustment results, maintains the visual consistency of the video image under changing lighting, and obtains the optimized monitoring video output results.

[0072] The key frame sets include high motion density frames, intensive content change frames and visual impact frames; the corrected morphological data include geometric alignment indicators, morphological correction coefficients and high-speed movement correction data; the seamless panoramic monitoring view includes time synchronization layers, spatial fusion layers and perspective integrity layers; the optimized monitoring video output results include light intensity standards, color temperature balance setting records and brightness adjustment level records.

[0073] See also Figure 2, the steps to obtain the key frame set are:

[0074] Analyze the highway monitoring video stream, extract each frame in the video and its previous frame, perform continuous frame correspondence, and obtain the original frame sequence set;

[0075] First, the highway surveillance video stream is analyzed and the number of video frames per second is extracted, such as 30 frames per second. Then, each frame is digitized and stored together with the previous frame. The raw image data captured by the high-resolution camera is processed with preliminary color correction and noise filtering to ensure that the image quality meets the analysis standards. Next, each frame is converted into digital data, which involves encoding the RGB value of each pixel, for example, converting color information into a value between 0 and 255. The data is then transmitted to the main server via a data link, where the original frame sequence set is stored on a high-speed data storage device, and each frame is indexed by time series to facilitate subsequent rapid retrieval and comparative analysis.

[0076] Perform pixel difference analysis and color histogram comparison on the original frame sequence set using the formula:

[0077]

[0078] Calculate the content difference value Δ between each pair of consecutive frames i , and obtain the difference measurement result, where n represents the total number of pixels in the frame, and f i,x and f i-1,x Represent the color values ​​of the corresponding pixels in the i-th frame and the i-1-th frame respectively;

[0079] For a specific video frame pair, 10 specific pixel differences are extracted. The difference data directly reflects the changes from the previous frame to the current frame. The difference values ​​are: [5, 3, 10, 6, 8, 2, 7, 4, 9, 1] respectively.

[0080] Compute the sum of the absolute values ​​of the differences:

[0081] 5+3+10+6+8+2+7+4+9+1=55

[0082] Average this sum over the 10 pixels:

[0083]

[0084] This value of 5.5 means that among these 10 sample pixels, the average color value of each pixel between frame i and frame i-1 changes by 5.5 units, which can help evaluate the degree of visual difference between the pair of frames.

[0085] According to the difference measurement results, the difference threshold is set to filter the frames with the most significant content changes, determine the key action events represented, and generate a key frame set;

[0086] In the process of determining the key frame set, the difference measurement results of each pair of consecutive frames obtained from the previous step are first analyzed. During this analysis, a difference threshold is set, for example, the threshold is set to a certain percentage of the highest difference value, such as 95%. This means that only those frames with difference values ​​higher than this threshold will be considered as frames with significant content changes. The specific operation is to sort all the calculated inter-frame difference measurement results and select the top 5% of the results as candidates. Then, these frames are marked as key frames and further content review is carried out. During the review process, the visual content of these frames is manually checked to confirm whether they truly reflect important actions or events. In addition, in order to improve the accuracy of key frame selection, object motion information and scene changes in the frame are also considered, such as using edge detection and object tracking technology to assist in judging the importance of the frame. Finally, the strictly screened key frames form a key frame set, which provides key data support for subsequent video content analysis and event reproduction.

[0087] See also Figure 3 , the comparison steps between the standard form and the current image representation are:

[0088] Based on the key frame set, edge detection is performed, each frame is processed and the contours of high-speed moving objects in the image are identified to generate an object morphology contour dataset;

[0089] After adopting the key frame set, the image preprocessing steps are first applied to enhance the recognizability of the object contours in the frame, including the use of local histogram equalization to increase the contrast of the image, making the edges of high-speed moving objects more clearly visible. Subsequently, edge detection processing is performed. This processing does not directly use pre-packaged library functions, but manually identifies the gradient change points in the image by constructing a gradient intensity matrix. The construction of the gradient matrix relies on the differential operation of the grayscale values ​​of the pixels around each pixel. The operation ensures that the precise contours of the object are effectively separated from various background noises. Finally, these contour data are integrated into a set. This dataset will be used for subsequent shape matching analysis. Every step of the image processing is precisely controlled throughout the process, from grayscale adjustment to edge recognition. Each link undergoes strict parameter tuning and effect verification to ensure the high quality and availability of the generated object morphology dataset.

[0090] Perform shape matching analysis on the object shape contour dataset and compare it with the standard shape in the database using the formula:

[0091]

[0092] Calculate the morphological matching score S of each object and get the morphological matching result, where R i Represents the shape contour data points of the object, O i represents the data point corresponding to the standard form, and N represents the number of points involved in the comparison;

[0093] There is a small sample data set, where N = 5, and the object's morphological contour data points R i and the standard morphological data point O in the database i As shown below:

[0094] R = [1, 3, 5, 7, 9]

[0095] O = [1, 2, 5, 6, 9]

[0096] Perform the calculation:

[0097]

[0098] The result S≈0.632 represents the morphology matching score, showing that there is a slight difference between the current image and the standard morphology. The closer the score is to 1, the more similar the morphology is. The score of 0.632 indicates that although most of the key points are similar in position, there are still some deviations.

[0099] Using the morphological matching results, the deviation between the current image representation of each object and its standard morphology is evaluated using the formula:

[0100] D = max(1-S, 0);

[0101] Calculate the morphological deviation value D and obtain the morphological data comparison analysis results;

[0102] The similarity score S=0.632 obtained by the above calculation is still used.

[0103] Insert the values ​​to calculate:

[0104] D=max(1-0.632,0)=max(0.368,0)=0.368

[0105] The obtained D=0.368 represents the morphological deviation value. This value indicates that it is necessary to adjust and optimize the morphology between the current image representation and the standard morphology to reduce this deviation, suggesting that there is still room for improvement.

[0106] See also Figure 4 , the steps to obtain the corrected morphological data are:

[0107] Based on the comparative analysis results of morphological data, by comparing image data with standard model data, the formula is used:

[0108]

[0109] Calculate the visual consistency score V to obtain the visual consistency evaluation data, where S is the morphological matching score and T is the consistency threshold;

[0110] S = 0.632 (morphology matching score)

[0111] T = 0.85 (consistency threshold)

[0112] Interpolation is used to calculate:

[0113]

[0114] The result V≈1.058 represents the visual consistency score. A value greater than 1 indicates that there is a significant difference between the current image and the standard form, and further correction or analysis is needed to improve the consistency between the image and the standard form.

[0115] Using visual consistency evaluation data, image distortion metric calculation is performed using the formula:

[0116]

[0117] Calculate and output the morphology correction coefficient C d ;

[0118] The given parameters are:

[0119] V=1.058

[0120] Interpolation is used to calculate:

[0121]

[0122] Get C d ≈0.742 represents the morphological correction factor, which is used to adjust the distortion in the image due to speed. Less than 1 indicates that the applied correction does not completely compensate for the morphological distortion caused by the rapid movement of the object, but still effectively reduces most of the distortion. The correction factor of 0.742 indicates that the correction operation eliminates about 74.2% of the image distortion, and some distortions still need further correction or optimization.

[0123] Apply the morphology correction factor C d , correct the morphological error of each object, and generate corrected morphological data by adjusting the image data to match the actual state;

[0124] The parts of the dynamic object are identified by analyzing each pixel of the image, and then the position and shape of these pixels are adjusted according to the correction factor through a proportional adjustment. In specific operation, if a pixel deviates from its proper position due to the rapid movement of the object, the adjustment will calculate its new position according to the moving direction and velocity vector diagram of the object. The new position is obtained by multiplying the original position vector with the morphological correction coefficient to compensate for the morphological distortion caused by speed. Finally, all the adjusted pixel data are recombined into an image, making the final image visually closer to the actual appearance of the object, significantly improving the visual error caused by rapid movement, and providing a clearer and more accurate visual representation, thereby ensuring the accuracy and consistency of the data and ensuring that the image truly reflects the actual state of the object.

[0125] See also Figure 5 , the spatial and temporal alignment steps are:

[0126] Using the corrected morphological data, the surveillance images captured by multiple cameras are preliminarily sorted according to timestamps to confirm that the time sequence of each image is correct and obtain timestamp sorted data;

[0127] Based on the corrected morphological data obtained, the surveillance images captured by multiple cameras are processed. The corrected morphological data mainly covers the optimization of image resolution, the adjustment of contrast, and the standardization of brightness. The data is obtained through a series of image processing techniques, including edge enhancement processing, which improves the overall clarity of the image by identifying and highlighting the edges in the image; at the same time, color correction processing is also applied to adjust the image color deviation to ensure the color consistency of images captured by different cameras. This stage of processing ensures that each image meets certain standards in visual effects. After that, the surveillance images are sorted according to timestamps. The sorting operation includes extracting the time tags in the metadata of each image and rearranging the images in the order of these time tags to ensure the logic and accuracy of the image sequence on the time axis, laying the foundation for the next step of spatial alignment.

[0128] Sort the data based on timestamps, perform spatial alignment, analyze the spatial overlap areas and key alignment points between images, verify the spatial continuity and accuracy of the images, and generate spatial alignment data;

[0129] The images sorted based on timestamps are spatially aligned. The process requires a detailed analysis of the spatial overlapping areas between images, especially the specific feature points in the images, such as traffic signs and corners of roadside buildings, which are key references for spatial alignment. By extracting these feature points, geometric transformation techniques are applied, including image rotation and scaling, to ensure that the feature points of each image are aligned in space. The steps are manually set, relying on image processing experts to adjust the misalignment between images one by one, ensuring the perfect docking of images before stitching. Accurate spatial alignment ensures the visual continuity of the image sequence, providing high-quality basic data for subsequent video stitching.

[0130] Integrate the spatial alignment data with the time stamp sorting data, perform time alignment, confirm the consistency of the image sequence on the time axis, and obtain the spatial and temporal alignment results;

[0131] By integrating the spatially aligned data with the timestamp sorted data, temporal alignment is performed, which involves adjusting the time each image is displayed to eliminate the time differences caused by processing delays. Temporal alignment is accomplished by analyzing the processing and transmission time of each image in detail, ensuring that the image display time is completely synchronized with the actual time. The steps pay special attention to the time delays in data transmission and processing, and adjust the time parameters to ensure continuity by monitoring the timestamp information of the data stream in real time. The work at this stage provides an accurate timeline for the integration of video data, ensuring that subsequent analysis and use can rely on this series of strictly aligned image data to obtain spatial and temporal alignment results, ensuring the real-time and accuracy of the monitoring system.

[0132] See also Figure 6 ,The steps to obtain seamless panoramic monitoring view are:

[0133] Based on the spatial and temporal alignment results, multi-view video stitching is started, the image boundaries output by the camera are processed, image edge blending is performed to smoothly transition the boundary area, and preliminary seamless video stitching data is generated;

[0134] The multi-view video stitching process is started based on the spatial and temporal alignment results. Specifically, edge detection technology is used to identify the boundary lines of each image, and then pixel-level weight adjustment is used to achieve seamless transition between images. This involves dynamically adjusting the brightness and color of edge pixels of adjacent images so that the image from one camera perspective to another can transition visually smoothly. In addition, the image fusion technology used also includes multi-scale fusion, which analyzes the details and structural information of the image at different scales to ensure the consistency of details and overall harmony during the stitching process. The implementation of technical processing not only depends on the precise design of the algorithm, but also needs to adjust parameters according to the actual image content to optimize the stitching effect, thereby generating preliminary seamless video stitching data.

[0135] Based on the preliminary seamless video stitching data, image color adjustment is performed to match the hue deviation in the camera's field of view, and exposure compensation is applied to balance the overall brightness, using the formula:

[0136]

[0137] Calculate the multiple C of the new color value of each pixel to the original color value i , and get the optimized multi-view spliced ​​video, where R i Represents the red component of the i-th pixel, G i Represents the green component of the i-th pixel, B i Represents the blue component of the i-th pixel;

[0138] Assume that the RGB value of a pixel is R i =100, G i =150, B i = 200. According to the formula, find the maximum and minimum of the three values:

[0139] max(R i ,G i ,B i )=200

[0140] min(R i ,G i ,B i )=100

[0141] Compute the ratio of the maximum to the minimum value:

[0142]

[0143] Compute the sum of all color channels:

[0144] R i +G i +B i =100+150+200=450

[0145] Substitute the values ​​into the formula to calculate C i :

[0146]

[0147] The results show that after color adjustment, the new color value of the pixel is 0.0032 times the original color value. By emphasizing color contrast and keeping the total color intensity unchanged, the overall color distribution of the image is effectively balanced, which can significantly improve the color continuity at the image joints and reduce the hard edge effect of color transition, thereby achieving a smoother and more natural visual effect.

[0148] Use optimized multi-view stitched video for all-round monitoring, analyze video data in real time, identify traffic flow, speed and vehicle density information, automatically detect and annotate key parameters of traffic flow, and obtain seamless panoramic monitoring views;

[0149] The optimized multi-view stitched video is used for all-round monitoring. The analysis process at this stage involves complex image recognition technology, mainly including functions such as vehicle detection, vehicle classification and vehicle speed estimation. The specific operation is to first separate the vehicle image in the video frame from the background through image segmentation technology, and then use machine learning models, such as convolutional neural networks, to identify different types of vehicles. In this process, each vehicle model has a pre-trained model to identify specific shapes and features. Then, the vehicle's movement trajectory is determined by tracking technology to calculate the vehicle speed. This series of processes requires not only detailed analysis of each frame of the image, but also continuous tracking of vehicle dynamics to ensure the real-time and accuracy of the data. The final seamless panoramic monitoring view provides comprehensive traffic flow information and supports traffic management departments to make real-time decisions and responses, thereby effectively improving the overall performance and response speed of the monitoring system.

[0150] See also Figure 7 , the steps to adjust the image contrast are:

[0151] Based on the seamless panoramic monitoring view, the time series frames of the video are collected, and the light intensity of each frame is sampled to obtain the light sampling data;

[0152] From the seamless panoramic monitoring view, the time series frames of the video are collected. The process is completed by a high-precision image sampling program. The program first decodes the video frames and then extracts the brightness value in each frame. Sampling the light intensity of each frame involves analyzing the brightness of each pixel and calculating the total light intensity of the frame by averaging the brightness of the pixels in the entire frame. The value reflects the lighting conditions of the video at a specific point in time for subsequent analysis of lighting changes. The data is then encoded and temporarily stored in a specially designed data cache area to ensure data integrity and traceability. The cached data is regularly updated and cleaned to ensure the efficient use of system resources and reduce data processing delays. The detailed data collection and processing process ensures the accuracy and real-time nature of the lighting data, providing reliable input data for subsequent lighting change analysis to obtain lighting sampling data.

[0153] Apply time series analysis to the illumination sampling data to determine the difference in illumination intensity between consecutive frames using the formula:

[0154] ΔI t,t-1 =|I t -I t-1 |;

[0155] Calculate the light intensity difference ΔI t,t-1 , generate the illumination change index, where I t and I t-1 Respectively represent the light intensity of two consecutive frames;

[0156] Through time series analysis technology, using the light sampling data obtained in the previous step, we perform light change analysis. This analysis is based on the comparison of light intensity between each frame, using the calculation formula: ΔI t,t-1 =|I t -I t-1 | For example, if the light intensity is 300 units at time t and 250 units at time t-1, then ΔI t,t-1 =|300-250|=50. This difference value indicates the absolute change in light intensity between two consecutive frames, which helps to identify rapid or gradual changes in light in video surveillance. The generated light change index provides a quantitative view for evaluating light conditions at different time periods, providing a basis for subsequent image processing.

[0157] According to the illumination change index, the image contrast is readjusted using the formula:

[0158]

[0159] Calculate and output the adjusted new contrast C new , get the real-time image contrast adjustment result, where C old represents the previous image contrast;

[0160] C old Represents the old image contrast, usually the initial or standard contrast value set for the video surveillance system; ΔI t,t-1 is the difference in illumination intensity between the two frames calculated from the previous step, which is 50. old =1.0, then:

[0161]

[0162] The new contrast value is 0.0199, which is significantly lower than the original contrast, reflecting that the contrast is automatically reduced due to the increased illumination to maintain the stability of image quality. The results show that adjusting the image contrast through actual illumination change data can effectively adapt to environmental changes and ensure that the quality of the surveillance video matches the illumination conditions.

[0163] See also Figure 8 , the steps to obtain the optimized surveillance video output results are:

[0164] Based on the real-time image contrast adjustment results, the highlight and shadow areas in the video are identified through image analysis, the area boundaries and intensity changes are analyzed, and the highlight and shadow area analysis data are generated;

[0165] Video obtained from the seamless panoramic surveillance view is processed using an image analysis program based on edge detection technology. By analyzing the lighting distribution in each frame, highlight and shadow areas are automatically identified. The process involves calculating the brightness value of each pixel in the image and comparing these values ​​with preset brightness thresholds. Pixels with brightness values ​​above the upper threshold are marked as highlight areas, and those below the lower threshold are marked as shadow areas. This method generates an accurate map of all highlight and shadow areas, which records the location, size and shape of each area in detail. The data is then used to guide the application of high dynamic range processing algorithms, ensuring that they can be optimized for specific areas in the image.

[0166] Based on the highlight and shadow area analysis data, high dynamic range processing is applied to match the visual needs under different lighting conditions, balance the light and dark details in the image, and generate HDR processed image data;

[0167] Apply high dynamic range (HDR) processing to previously identified highlight and shadow areas to balance the light contrast in the image. This algorithm is specifically designed to adjust the extreme brightness values ​​in the image to make the brightness of the entire image more even. This operation involves lowering the brightness of the highlight areas and raising the brightness of the shadow areas, thereby reducing the image contrast and enhancing the visibility of details. This process utilizes local contrast adjustment technology to automatically analyze the brightness of each area and calculate the necessary adjustment amount. By applying these adjustments frame by frame, the HDR algorithm ensures that the image maintains good visual effects in any lighting conditions, making details more clearly visible, thus providing a good foundation for the final image enhancement processing.

[0168] Perform local image sharpening and color correction based on the image data processed by HDR to ensure image consistency and natural transition of visual effects in all lighting environments, and generate optimized surveillance video output results;

[0169] Image enhancement technology is used to optimize the HDR processed images. The process includes local sharpening and color correction to ensure that the image is visually consistent and attractive. Local sharpening enhances the clarity of the image by emphasizing the detailed features of the image, such as edges and textures. Color correction adjusts the color temperature and saturation of the image to offset the color cast caused by different lighting conditions and ensure the true reproduction of color. The application of technology is achieved through pixel-by-pixel analysis to ensure that each part is properly processed. Through meticulous adjustments, the overall visual quality of the image is significantly improved, and the final output surveillance video can show the best visual effects under various ambient lighting conditions.

[0170] 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. A video monitoring system suitable for highways that can clearly identify high-speed moving objects, characterized in that: The system comprises: The key frame selection module analyzes the content difference between each frame and the previous frame in the highway monitoring video stream, measures the difference, selects the frame with the most significant content change, determines that each selected frame represents a key action event in the target video, and generates a key frame set; The morphological distortion analysis module performs morphological recognition and analysis on high-speed moving objects in the image based on the key frame set, compares the standard morphology of each object with the current image representation, analyzes the consistency between the visual representation and the actual state of each object according to the comparison result, corrects the image distortion caused by the speed, corrects the morphological error, and obtains the corrected morphological data; Based on the corrected morphological data, the image synchronization fusion module hierarchically sorts the surveillance images captured by multiple cameras according to timestamps, performs spatial and temporal alignment, and seamlessly stitches multi-view videos according to the alignment results to reveal the continuous highway traffic flow, perform all-round viewing angle monitoring, and obtain a seamless panoramic monitoring view; Based on the seamless panoramic monitoring view, the dynamic lighting adjustment module monitors and analyzes the lighting changes in different time periods in the current video, adjusts the image contrast in real time, balances the visual effects of shadow and highlight areas according to the adjustment results, maintains the visual consistency of the video image under changing lighting, and obtains the optimized monitoring video output result.

2. The video monitoring system for highways capable of distinguishing high-speed moving objects according to claim 1 is characterized in that: The steps of obtaining the key frame set are: Analyze the highway monitoring video stream, extract each frame in the video and its previous frame, perform continuous frame correspondence, and obtain the original frame sequence set; The pixel difference analysis and color histogram comparison are performed on the original frame sequence set, using the formula: Calculate the content difference value Δ between each pair of consecutive frames i , and obtain the difference measurement result, where n represents the total number of pixels in the frame, and f i,x and f i-1,x Represent the color values ​​of the corresponding pixels in the i-th frame and the i-1-th frame respectively; According to the difference measurement result, a difference threshold is set, the frames with the most significant content changes are screened, and the representative key action events are determined to generate a key frame set.

3. The video monitoring system for highways capable of distinguishing high-speed moving objects according to claim 2 is characterized in that: The steps of comparing the standard form with the current image representation are: Based on the key frame set, edge detection is performed, each frame is processed and the contour of the high-speed moving object in the image is identified to generate an object morphology contour data set; The object shape contour data set is subjected to shape matching analysis and compared with the standard shape in the database using the formula: Calculate the morphological matching score S of each object and get the morphological matching result, where R i Represents the shape contour data points of the object, O i represents the data point corresponding to the standard form, and N represents the number of points involved in the comparison; Using the morphological matching results, the deviation between the current image representation of each object and its standard morphology is evaluated using the formula: D = max(1-S, 0); Calculate the morphological deviation value D and obtain the morphological data comparison and analysis results.

4. The video monitoring system for highways capable of distinguishing high-speed moving objects according to claim 3 is characterized in that: The steps for obtaining the corrected morphological data are as follows: Based on the comparative analysis results of the morphological data, by comparing the image data with the standard model data, the formula is used: Calculate the visual consistency score V to obtain the visual consistency evaluation data, where S is the morphological matching score and T is the consistency threshold; The image distortion metric is calculated using the visual consistency evaluation data using the formula: Calculate and output the morphology correction coefficient C d ; Apply the morphology correction factor C d , correct the morphological error of each object, and generate corrected morphological data by adjusting the image data to match the actual state.

5. The video monitoring system for highways capable of distinguishing high-speed moving objects according to claim 4 is characterized in that: The steps of space and time alignment are: Using the corrected morphological data, the surveillance images captured by multiple cameras are preliminarily sorted according to timestamps to confirm that the time sequence of each image is correct, and obtain timestamp sorting data; Based on the timestamp sorting data, spatial alignment is performed, spatial overlapping areas and key alignment points between images are analyzed, continuity and accuracy of the images in space are checked, and spatial alignment data is generated; The spatial alignment data is integrated with the timestamp sorting data, time alignment is performed, the consistency of the image sequence on the time axis is confirmed, and the spatial and temporal alignment results are obtained.

6. The video monitoring system for highways capable of distinguishing high-speed moving objects according to claim 5, characterized in that: The steps of obtaining the seamless panoramic monitoring view are as follows: Based on the spatial and temporal alignment results, multi-view video stitching is started, the image boundary output by the camera is processed, image edge blending is performed to smoothly transition the boundary area, and preliminary seamless video stitching data is generated; Based on the preliminary seamless video stitching data, image color adjustment is performed to match the hue deviation in the camera's viewing angle, and exposure compensation is applied to balance the overall brightness, using the formula: Calculate the multiple C of the new color value of each pixel to the original color value i , and get the optimized multi-view spliced ​​video, where R i Represents the red component of the i-th pixel, G i Represents the green component of the i-th pixel, B i Represents the blue component of the i-th pixel; The optimized multi-view stitched video is used for all-round monitoring, real-time analysis of video data, identification of traffic flow, vehicle speed and vehicle density information, automatic detection and annotation of key parameters of traffic flow, and acquisition of a seamless panoramic monitoring view.

7. The video monitoring system for highways capable of distinguishing high-speed moving objects according to claim 6, characterized in that: The steps for adjusting the image brightness and contrast are as follows: Based on the seamless panoramic monitoring view, a time series frame of the video is collected, and the light intensity of each frame is sampled to obtain light sampling data; Apply time series analysis to the illumination sampling data to determine the difference in illumination intensity between consecutive frames using the formula: ΔI t,t-1 =|I t -I t-1 |; Calculate the light intensity difference ΔI t,t-1 , generate the illumination change index, where I t and I t-1 Respectively represent the light intensity of two consecutive frames; According to the illumination change index, the image contrast is readjusted using the formula: Calculate and output the adjusted new contrast C new , get the real-time image contrast adjustment result, where C old Represents the previous image contrast.

8. The video monitoring system for highways capable of distinguishing high-speed moving objects according to claim 7, characterized in that: The steps for obtaining the optimized monitoring video output result are: Based on the real-time image contrast adjustment result, identifying highlight and shadow areas in the video through image analysis, analyzing area boundaries and intensity changes, and generating highlight and shadow area analysis data; Based on the highlight and shadow area analysis data, high dynamic range processing is applied to match visual requirements under different lighting conditions, balance light and dark details in the image, and generate HDR processed image data; Based on the image data processed by HDR, local image sharpening and color correction are performed to ensure image consistency and natural transition of visual effects in all lighting environments, thereby generating optimized monitoring video output results.

Images