Monitoring image processing method and system for high-precision target tracking

Abstract

The invention discloses a monitoring image processing method and system for high-precision target tracking, and belongs to the technical field of monitoring processing, and the method specifically comprises the steps: collecting polarization data of incident light in a monitoring scene, and generating a physical attribute signature of a light source; carrying out space-time clustering analysis on pixel points of which the variable quantity exceeds an environment illumination baseline in the monitoring scene to form a physical boundary of an illumination mutation region; obtaining motion trail data of the target vehicle, and generating a motion trail space area of the target vehicle; marking an overlapping region of the physical boundary of the illumination abrupt change region and the motion trail space region as a dynamic interference source influence region; if the predicted position of the target vehicle is located in the dynamic interference source influence area, calling an exposure control parameter and a gain control parameter corresponding to the physical attribute signature, calculating a focusing control parameter according to the scene depth of the predicted position, and loading the focusing control parameter to a sensor control unit; according to the method, the problem of target key frame loss caused by AE / AF delay under dynamic illumination is effectively solved.

Description

Technical Field

[0001] The present invention relates to the field of monitoring processing technology, and in particular to a monitoring image processing method and system for high-precision target tracking. Background Art

[0002] In today's society, with the increasing demand for public safety, the need for high-precision target tracking is becoming increasingly urgent in the field of public security surveillance. Whether it's vehicle monitoring at traffic intersections or monitoring people in large public places, accurate and real-time target tracking is required to achieve effective security prevention and management. Surveillance image processing technology, as the core support for target tracking, has a direct impact on the effectiveness of public security surveillance systems.

[0003] Currently, several technologies have been developed in the field of surveillance image processing. For example, these technologies can distinguish between natural and artificial light sources by collecting data on the polarization angle and degree of polarization distribution of light, perform spatiotemporal pixel clustering analysis using changes in light intensity to identify areas of sudden illumination changes, and obtain target trajectory data for tracking. These technologies provide a foundation for surveillance image processing and, to a certain extent, enable target tracking and surveillance scene processing.

[0004] However, existing technologies still have some shortcomings in practical applications. When the camera faces dynamic lighting changes, such as flashing car lights, there is a 100-200ms delay in auto-exposure and auto-focus, which can cause keyframe loss and seriously affect the accuracy of target tracking and object recognition. Summary of the Invention

[0005] The purpose of the present invention is to provide a monitoring image processing method and system for high-precision target tracking to solve the following technical problems:

[0006] When the camera faces dynamic lighting changes, there is a certain delay in automatic exposure and autofocus, which will cause key frames to be lost, seriously affecting the accuracy of target tracking and object recognition.

[0007] The purpose of the present invention can be achieved through the following technical solutions:

[0008] A monitoring image processing method for high-precision target tracking includes the following steps:

[0009] S1. Real-time collection of polarization angle distribution data and polarization degree distribution data of incident light in the monitoring scene, distinguishing between constant natural light sources and dynamic artificial light sources, and generating physical property signatures of dynamic artificial light sources;

[0010] S2. Detect the change in light intensity at each pixel in the monitored scene, perform spatiotemporal clustering analysis on pixels whose changes exceed the ambient light baseline, and form the physical boundaries of the light mutation area.

[0011] S3, obtaining the motion trajectory data of the target vehicle in continuous image frames, extracting the position coordinate sequence and the motion direction sequence, and generating the motion trajectory space area of ​​the target vehicle;

[0012] S4. When the physical boundary of the illumination mutation area and the motion trajectory space area have geometric overlap, the overlapping area is marked as the dynamic interference source influence area;

[0013] S5. Predict the target vehicle's predicted position at the next image capture moment. If the predicted position is within the dynamic interference source's influence area, call the exposure control parameters and gain control parameters corresponding to the physical attribute signature, and calculate the focus control parameters based on the scene depth at the predicted position.

[0014] S6. During the vertical blanking period of the image sensor, the exposure control parameter, the gain control parameter, and the focus control parameter are loaded into the sensor control unit.

[0015] As a further solution of the present invention: in S1, the process of generating the physical property signature of the dynamic artificial light source is:

[0016] A database of polarization characteristic feature templates containing metal reflective light sources and transmissive material light sources is pre-built. The polarization angle distribution data and polarization degree distribution data collected in real time are input into the feature matching calculation unit for processing. The similarity matrix value between the current polarization data and the feature templates in the database is calculated. When the similarity value exceeds the preset natural light discrimination threshold, the binary code of the corresponding feature template is extracted.

[0017] The binary code includes combined information of a light source material type identifier and a spectral characteristic identifier. The complete binary code is output as a physical property signature and stored in a cache area.

[0018] As a further solution of the present invention: in S2, the process of forming the physical boundary of the illumination mutation area is:

[0019] Continuously record and output the valid pixel event coordinates where the change in light intensity exceeds the ambient light baseline, perform density clustering operations on valid event points that are continuous in time and adjacent in space, and calculate the convex hull boundary vertex coordinate set data for each event cluster group;

[0020] When the convex hull area change rate exceeds the upper threshold of natural light fluctuation, the effective light mutation area is confirmed. The physical boundary of the light mutation area is the closed polygon formed by connecting the coordinate sequence of the convex hull boundary vertices, and the coordinate value of the minimum circumscribed rectangle of the closed polygon is recorded.

[0021] As a further solution of the present invention: in S3, the process of generating the target vehicle's motion trajectory space area is:

[0022] The historical position coordinate sequence of the key points of the target vehicle's outline is obtained through a feature point tracking algorithm. The key point position coordinates of adjacent image frames are connected to generate a continuous motion trajectory polyline. The width value of the target vehicle's physical size is expanded to both sides with the motion trajectory polyline as the center line.

[0023] A strip buffer space area is generated to cover the uncertainty of the motion direction. The width parameter value of the strip buffer changes linearly with the target vehicle's motion speed. The final boundary of the motion trajectory space area is composed of an equidistant curve that is completely parallel to the trajectory line.

[0024] As a further solution of the present invention: in S5, the process of predicting the predicted position of the target vehicle at the next frame image acquisition moment is:

[0025] Determine the physical constraint parameter value of the minimum turning radius based on the target vehicle type, fit the motion direction angle change curve using the position point data of the last three consecutive frames, correct the extreme point position of the direction change curve based on the minimum turning radius physical constraint, and extrapolate the position point of the next frame acquisition time along the corrected curve direction as the prediction origin;

[0026] The range coordinate values ​​of the predicted position are generated with the prediction origin as the center point, and the long axis direction of the predicted position range is spatially aligned with the current movement direction of the target vehicle.

[0027] As a further solution of the present invention: in S5, the process of calling the exposure control parameters and gain control parameters corresponding to the physical attribute signature is:

[0028] A database table pre-stores the mapping relationship between physical property signatures and exposure parameters. The physical property signature of a metal reflective light source is mapped to a short exposure time parameter combination with an exposure time less than a first set threshold; the physical property signature of a transmissive material light source is mapped to a medium exposure parameter combination with an exposure time greater than the first set threshold and less than a second set threshold.

[0029] When the predicted position is covered by the influence areas of multiple dynamic interference sources, the parameters corresponding to the interference source with the largest peak light intensity are selected; the focus control parameters are calculated through the binocular stereo vision system to calculate the average depth value of the predicted position; and the depth value is converted into the curvature control voltage output instruction parameter required by the liquid lens device.

[0030] As a further solution of the present invention: in S6, the process of loading the parameters into the image sensor control unit is:

[0031] The sensor control unit receives all control parameter instructions at the start of the vertical blanking period. The exposure control parameters are converted into clock configuration instructions for the charge accumulation period of the image sensor. The gain control parameters are converted into impedance adjustment configuration instructions for the programmable amplifier circuit. The focus control parameters are directly input into the dedicated drive circuit interface of the liquid lens device. The drive circuit compensates the voltage output value according to the real-time temperature sensor data.

[0032] After completing the above conversion, a parameter loading completion ready signal is output, and the curvature stability state of the liquid lens device is verified in real time through a laser interferometer measurement device. When the exposure timing control signal is synchronized with the optical stability signal, the image acquisition action is triggered.

[0033] As a further solution of the present invention, the process of verifying the curvature stability of the liquid lens device is as follows:

[0034] A monochromatic coherent laser beam with a fixed wavelength is projected toward the predicted position. The high-speed image sensor unit captures the interference fringe pattern on the surface of the target object and calculates the correlation coefficient of the spatial phase distribution of the interference fringe pattern at adjacent acquisition moments. When the correlation coefficient value is continuously higher than the preset stability threshold, an optical stability signal is output. If the correlation coefficient value drops below the preset threshold during the exposure acquisition process, the acquisition process of the current image frame is immediately terminated.

[0035] As a further solution of the present invention: after outputting a parameter loading completion ready signal, dynamic light field coherence verification is performed, a theoretical polarization distribution matrix corresponding to the light source physical property signature is called, and a theoretical light field gradient is calculated based on a direction vector of the motion trajectory spatial region;

[0036] Acquire real-time phase distribution data from the laser interferometer. Perform a coherent matching operation on the theoretical light field gradient and the real-time phase data. When the matching similarity exceeds the dynamic environment threshold, the current parameters are maintained. If the matching similarity does not reach the threshold, a parameter reset instruction is sent to the sensor control unit.

[0037] The present invention also includes a high-precision target tracking monitoring image processing system for implementing the above-mentioned high-precision target tracking monitoring image processing method, comprising:

[0038] The light source acquisition module is used to collect the polarization angle distribution data and polarization degree distribution data of the incident light in the monitoring scene in real time, distinguish between constant natural light sources and dynamic artificial light sources, and generate the physical property signature of the dynamic artificial light source;

[0039] The mutation detection module is used to detect the change in light intensity at each pixel in the monitoring scene, and perform spatiotemporal clustering analysis on pixels whose changes exceed the ambient light baseline to form the physical boundaries of the light mutation area;

[0040] The trajectory analysis module is used to obtain the motion trajectory data of the target vehicle in continuous image frames, extract the position coordinate sequence and the motion direction sequence, and generate the motion trajectory space area of ​​the target vehicle;

[0041] The trajectory comparison module is used to mark the overlapping area as the dynamic interference source influence area when the physical boundary of the illumination mutation area overlaps with the motion trajectory space area;

[0042] The parameter calling module is used to predict the target vehicle's predicted position at the next frame image acquisition moment. If the predicted position is within the influence area of ​​the dynamic interference source, the exposure control parameters and gain control parameters corresponding to the physical property signature are called, and the focus control parameters are calculated based on the scene depth at the predicted position.

[0043] The parameter loading module is used to load the exposure control parameters, gain control parameters and focus control parameters into the sensor control unit during the vertical blanking period of the image sensor.

[0044] Beneficial effects of the present invention:

[0045] The present invention generates a dynamic artificial light source physical property signature by real-time collection of light polarization angle and polarization degree distribution data, forms the physical boundary of the illumination mutation area by combining spatiotemporal clustering analysis of the illumination intensity variation, and then obtains the target vehicle motion trajectory data to generate a trajectory space area. When the illumination mutation area and the trajectory space area geometrically overlap, the dynamic interference source influence area is marked, and the next frame position is predicted based on the target vehicle motion trajectory. If the predicted position is within the influence area, the exposure and gain control parameters corresponding to the physical property signature are called, and the focus control parameters are calculated according to the scene depth. Finally, the parameters are loaded during the vertical blanking period of the image sensor and the curvature stability of the liquid lens is verified. This effectively solves the problem of key frame loss caused by AE / AF delay under dynamic illumination of traditional cameras, realizes intelligent processing of the relationship between the illumination mutation area and the target motion trajectory, improves the accuracy and timeliness of parameter loading to the sensor control unit and equipment status verification, ensures the quality of image acquisition and tracking accuracy during target tracking, and avoids problems such as trajectory breakage and feature loss caused by dynamic illumination interference. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The present invention will be further described below with reference to the accompanying drawings.

[0047] Figure 1 It is a flow chart of a monitoring image processing method for high-precision target tracking of the present invention;

[0048] Figure 2 It is a module schematic diagram of a monitoring image processing system for high-precision target tracking of the present invention. DETAILED DESCRIPTION

[0049] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0050] See also Figure 1 As shown, the present invention is a monitoring image processing method for high-precision target tracking, comprising the following steps:

[0051] S1. Real-time acquisition of the polarization angle and polarization degree distribution data of the incident light in the monitoring scene. Feature matching is performed through a pre-built light source polarization characteristic template database. The binary code of the feature template that exceeds the natural light distinction threshold is extracted, and the physical property signature of the dynamic artificial light source is generated and stored to distinguish natural light sources from artificial light sources.

[0052] S2. Detect the change in illumination intensity of each pixel, perform spatiotemporal density clustering on the valid pixel event coordinates that exceed the environmental baseline, calculate the convex hull boundary of the cluster grouping, and confirm that the physical boundary of the illumination mutation area is a closed polygon when the convex hull area change rate exceeds the natural fluctuation threshold.

[0053] S3. Obtain the trajectory of key points of the target vehicle's contour by tracking feature points, generate a strip buffer with the trajectory polyline as the center and the width adjusted according to the vehicle speed, and form a motion trajectory space area.

[0054] S4. When the illumination mutation area geometrically overlaps with the motion trajectory space area, the overlapping area is marked as the dynamic interference source influence area.

[0055] S5. Determine the minimum turning radius constraint based on the target vehicle type. Fit the motion direction curve using the three frames of position data and correct for extreme points. Extrapolate the prediction origin and generate a predicted position range aligned with the motion direction. If the predicted position is within the impact zone, invoke the exposure and gain parameters corresponding to the physical property signature. Depth is calculated using binocular vision and converted into a curvature control voltage for the liquid lens.

[0056] S6. During the vertical blanking period of the image sensor, the exposure and gain parameters are converted into sensor clock and amplifier configuration instructions. The focus parameters are input into the liquid lens driver circuit and temperature compensation is performed. After verifying the stability of the lens curvature through laser interferometry, the exposure and stability signals are synchronized to trigger image acquisition.

[0057] In a preferred embodiment of the present invention, the process of generating the physical property signature of the dynamic artificial light source in S1 is:

[0058] First, a polarization characteristic feature template database is constructed, which pre-includes the polarization characteristic data of metal reflection light sources and transmissive material light sources under different environmental conditions. During the construction, technicians use professional optical measurement equipment to collect polarization angle distribution data and polarization degree distribution data of metal reflection light sources (such as vehicle headlights, billboard lights) and transmissive material light sources (such as glass curtain wall reflection light, plastic material transmission) in a variety of lighting scenarios. After preprocessing and feature extraction, a standard feature template is formed.

[0059] When the system is running, it collects real-time polarization angle and degree of polarization distribution data of the incident light within the monitoring scene and inputs this real-time data into the feature matching calculation unit. This calculation unit performs precise numerical calculations, comparing the current real-time polarization data with the feature templates stored in the database on a template-by-template basis. Using a specific similarity algorithm, it calculates the similarity matrix between the current polarization data and each feature template. This similarity algorithm is based on the principles of polarization optics and comprehensively considers the differences in polarization angle distribution and changes in polarization intensity to accurately measure the degree of match between the real-time data and the template.

[0060] When the calculated similarity value exceeds the pre-set natural light distinction threshold, the system will automatically extract the binary code of the corresponding feature template. This binary code uses specific encoding rules and contains a combination of information about the light source material type identifier and the spectral characteristic identifier. Among them, the light source material type identifier is used to distinguish whether the light source is a metal reflective material or a transmissive material, and the spectral characteristic identifier reflects the spectral distribution characteristics of the light source, such as the visible light band, infrared band, etc. Finally, the system will output the complete binary code as the physical property signature of the dynamic artificial light source and store it in the cache area for subsequent processing flow calls, thereby achieving accurate distinction between constant natural light sources and dynamic artificial light sources.

[0061] In another preferred embodiment of the present invention, in S2, the process of forming the physical boundary of the illumination mutation area is:

[0062] The system continuously monitors the change in light intensity at each pixel in the monitored scene in real time. When the change in light intensity at a pixel exceeds the ambient light baseline, the pixel is identified as a valid pixel event point, and the system records and outputs its coordinates. The ambient light baseline is obtained by statistically analyzing background light data over a period of time and reflects the range of normal ambient light variations.

[0063] For these valid event points that are continuous in time and adjacent in space, the system performs a density clustering operation. Specifically, a density clustering algorithm is used to divide eligible valid event points into different cluster groups based on the set spatial distance threshold and time interval threshold. Each cluster group corresponds to a possible illumination mutation area. Next, the system calculates the convex hull boundary vertex coordinate set data for each event cluster group. The convex hull calculation can determine the maximum spatial boundary range of the cluster group, thereby preliminarily outlining the shape of the illumination mutation area.

[0064] To mitigate the effects of natural light fluctuations, the system calculates the area change rate of each convex hull. Natural light fluctuations typically exhibit a certain regularity and a limited range of variation, so an upper threshold for natural light fluctuations is set accordingly. When the convex hull area change rate exceeds this threshold, it indicates that the light change in the area is not caused by natural factors but is a true sudden change in light. The system then identifies the area as a valid sudden change region. The physical boundary of this valid sudden change region is formed by connecting the coordinates of the convex hull's boundary vertices in sequence to form a closed polygon. The system also calculates and records the coordinates of the minimum bounding rectangle of this closed polygon. The minimum bounding rectangle provides a more concise geometric description for subsequent region location and analysis, facilitating the determination of spatial relationships with other regions, thereby accurately defining the physical boundary of the sudden change region.

[0065] In another preferred embodiment of the present invention, in S3, the process of generating the target vehicle's motion trajectory space area is:

[0066] First, a feature point tracking algorithm is used to obtain a historical coordinate sequence of key points on the target vehicle's outline. In practice, this algorithm identifies and tracks distinctive features on the target vehicle's outline, such as vehicle corners, headlight positions, and license plate edges. The system locates these features in each image frame and records their coordinates. Over time, this creates a historical coordinate sequence of key points on the target vehicle's outline.

[0067] Next, the coordinates of the keypoint positions in adjacent image frames are connected sequentially to generate a continuous motion trajectory polyline. This polyline intuitively reflects the target vehicle's motion path over a period of time. To more comprehensively describe the target vehicle's spatial range, the system uses this motion trajectory polyline as the center line and expands the width of the target vehicle's physical dimensions in both directions. The target vehicle's physical dimensions are pre-set based on standard dimensions for different vehicle types, such as the width of a small car or a large truck. This expansion creates a preliminary area encompassing the actual space occupied by the target vehicle.

[0068] Taking into account the uncertainty of the target vehicle's direction of motion during its movement, a strip-shaped buffer zone is generated to cover this uncertainty. Specifically, the width parameter of the strip-shaped buffer zone adjusts linearly with the target vehicle's speed. When the target vehicle is moving faster, the uncertainty of its direction of motion is relatively large, and the width of the strip-shaped buffer zone increases accordingly. Conversely, when the target vehicle is moving slower, its direction of motion is more stable, and the width of the strip-shaped buffer zone decreases accordingly. Ultimately, the boundary of the trajectory spatial region is formed by equidistant curves that are completely parallel to the trajectory polyline. These equidistant curves maintain a fixed distance from the trajectory polyline. This distance is the expansion width determined by the target vehicle's physical size and speed, thus forming a complete region that accurately describes the target vehicle's spatial range.

[0069] In another preferred embodiment of the present invention, in S5, the process of predicting the predicted position of the target vehicle at the next frame image acquisition moment is:

[0070] The physical constraint parameter value for the minimum turning radius is determined based on the target vehicle type. Different vehicle types have different minimum turning radii due to factors such as body structure and wheelbase. For example, a small car has a relatively small minimum turning radius, while a large truck has a larger minimum turning radius. The system pre-stores the minimum turning radius parameters for various vehicle types. Once the target vehicle type is identified, the corresponding minimum turning radius value is used as the physical constraint parameter.

[0071] The system then fits a curve of angular changes in the direction of motion using the position data from the three most recent consecutive frames. By analyzing the target vehicle's position changes in these three frames, the angular changes in the vehicle's direction of motion can be calculated and a curve can be fitted to reflect this angular change trend. To ensure that the predictions are more consistent with actual physical laws, the system adjusts the positions of the extreme points of the angular change curve based on the physical constraints of the minimum turning radius. If the fitted curve contains extreme angular changes that exceed the vehicle's minimum turning radius, the system adjusts these extreme points to conform to the vehicle's actual motion limitations.

[0072] After completing the curve correction, the position point at the next frame acquisition moment is extrapolated along the direction of the corrected curve and used as the prediction origin. This prediction origin is calculated based on the historical motion trajectory and physical constraints of the target vehicle, and can more accurately predict the position of the vehicle in the next frame. Finally, the system generates the range coordinate values ​​of the predicted position with the prediction origin as the center point. Taking into account the possible errors in the prediction process and the uncertainty of the target vehicle's movement, the predicted position range is not a precise point, but an area. The long axis direction of this area is spatially aligned with the current direction of movement of the target vehicle, which can better reflect the uncertainty of the target vehicle in the direction of movement, thereby providing a more reliable basis for subsequent image processing and parameter adjustment.

[0073] In another preferred embodiment of the present invention, in S5, the process of calling the exposure control parameters and gain control parameters corresponding to the physical attribute signature is:

[0074] A database table of mapping relationships between physical property signatures and exposure parameters is pre-built and stored. This table is established based on a large amount of experimental data on light source characteristics, which clearly defines the correspondence between the physical property signatures of different types of light sources and the optimal exposure parameters. Specifically, for metal reflective light sources, such as vehicle headlights and metal surface reflections, the physical property signatures are mapped in the table to a short exposure time parameter combination with an exposure time less than a first set threshold, so as to address the problem that the strong light of metal reflective light sources easily leads to overexposure of the image; and the physical property signatures of transmissive material light sources, such as glass curtain wall transmission light and plastic material transmission light, are mapped to a medium exposure parameter combination with an exposure time greater than the first set threshold and less than the second set threshold, to ensure that the image details in the transmitted light scene are clearly presented. The first set threshold and the second set threshold here are determined through statistical analysis of the lighting characteristics of various natural light sources and artificial light sources, which can effectively distinguish different light source types and match the corresponding exposure strategies.

[0075] When the predicted location is covered by the influence areas of multiple dynamic interference sources, the present invention prioritizes the light interference that has the most significant impact on image quality by comparing the peak light intensity of each interference source and selecting the exposure parameters corresponding to the interference source with the largest light intensity peak. Regarding the acquisition of focus control parameters, the present invention utilizes a binocular stereo vision system to measure the depth of the scene at the predicted location and calculates the parallax information of the left and right camera images to obtain the average depth value of the predicted location. This depth value is then converted into the required curvature control voltage output instruction parameters according to the optical characteristics of the liquid lens device, thereby achieving precise adjustment of the liquid lens curvature and ensuring clear imaging of the target at the predicted location.

[0076] In another preferred embodiment of the present invention, the process of loading the parameters in S6 into the image sensor control unit is as follows:

[0077] When entering the vertical blanking period of the image sensor, the sensor control unit will receive all control parameter instructions from the front-end processing module at the beginning of this period. These instructions include exposure control parameters, gain control parameters, and focus control parameters. Among them, the exposure control parameters need to be converted into clock configuration instructions for the charge accumulation period of the image sensor. Specifically, the length of the exposure time corresponds to the clock frequency and pulse width of the charge accumulation period. By accurately configuring the clock signal, the exposure time of the image sensor can be controlled to adapt to the imaging requirements under different lighting conditions. The gain control parameters will be converted into impedance adjustment configuration instructions for the programmable amplifier circuit. The impedance of the programmable amplifier directly affects the amplification factor of the image signal. By adjusting the impedance parameters, precise control of the image gain can be achieved to ensure that the image maintains appropriate brightness under different light intensities.

[0078] Focus control parameters are directly input into the liquid lens device's dedicated drive circuit interface. The curvature of the liquid lens changes depending on the voltage signal output by the drive circuit. Because changes in ambient temperature affect the physical properties of the liquid lens, causing the relationship between the drive voltage and curvature to shift, the drive circuit needs to compensate for the voltage output value based on data collected by the real-time temperature sensor to ensure that the liquid lens can accurately adjust to the target curvature and achieve precise focus.

[0079] After completing the conversion and compensation operations for all the above parameters, the sensor control unit will output a parameter loading completion signal, indicating that the hardware system is ready for image acquisition. To ensure imaging quality, the system needs to verify the curvature stability of the liquid lens device in real time using a laser interferometer measurement device. Specifically, the system projects a monochromatic coherent laser beam of a fixed wavelength to the predicted position. When this laser beam hits the surface of the target object, it produces an interference fringe pattern, which is captured by the high-speed image sensor unit. By calculating the correlation coefficient of the spatial phase distribution of the interference fringe pattern at adjacent acquisition moments, it can be determined whether the curvature of the liquid lens is stable. When the correlation coefficient value is continuously higher than the system's preset stability threshold, it indicates that the curvature of the liquid lens has stabilized, and an optical stability signal is output. Only when the exposure timing control signal and the optical stability signal are synchronized and valid will the system trigger the image acquisition action to ensure that the captured image is clear and accurate.

[0080] In a preferred embodiment of this embodiment, the process of verifying the curvature stability of the liquid lens device has strict technical logic. The wavelength of the monochromatic coherent laser is carefully selected to ensure that its interference effect in the target scene is obvious and stable. The sampling frequency of the high-speed image sensor unit needs to match the laser projection frequency to ensure that the dynamic changes of the interference fringes can be accurately captured. The calculation method of the spatial phase distribution correlation coefficient is based on the principle of optical interference and can sensitively reflect the slight changes in the curvature of the liquid lens, thereby providing a reliable basis for judging the stable state.

[0081] In another preferred embodiment of this embodiment, after outputting a parameter loading completion signal, the system performs dynamic light field coherence verification. First, the theoretical polarization distribution matrix corresponding to the light source's physical property signature is called. This matrix is ​​pre-established based on the physical properties of the dynamic artificial light source and contains the theoretical polarization distribution values ​​of the light source at different spatial locations. Then, the theoretical light field gradient is calculated based on the directional vectors in the motion trajectory spatial region. The directional vectors in the motion trajectory spatial region reflect the motion trend of the target vehicle. By calculating the theoretical light field gradient, the changes in the light field within the target area can be predicted.

[0082] Next, the real-time phase distribution data output by the laser interferometer measurement device is obtained. This data reflects the actual state of the current light field. The theoretical light field gradient is coherently matched with the real-time phase data. The coherent matching operation comprehensively considers the differences between the theoretical model and the actual measurement data and calculates the matching similarity between the two. When the matching similarity exceeds the dynamic environment threshold, it means that the current parameter settings can adapt well to the actual light field environment, so the current parameters are maintained unchanged. If the matching similarity does not reach the threshold, it indicates that there is a large deviation between the actual light field and the theoretical prediction, and the current parameter settings may not achieve the ideal imaging effect. At this time, the system will send a parameter reset instruction to the sensor control unit, triggering the parameter reloading and adjustment process to ensure that the system can adapt to the dynamically changing light field environment and guarantee the image acquisition quality.

[0083] The dynamic light field coherence verification process fully considers the complexity and dynamics of the light field in actual monitoring scenarios. The theoretical polarization distribution matrix is ​​established based on a large amount of experimental data and theoretical analysis, and can accurately reflect the polarization characteristics of different types of light sources. The direction vector calculation of the motion trajectory space area combines the motion trajectory and speed information of the target vehicle, making the prediction of the theoretical light field gradient more in line with the actual situation. The laser interferometer measurement device can accurately obtain the phase distribution data of the light field in real time, providing a reliable measurement basis for coherent matching operations. The setting of the dynamic environment threshold takes into account the range of light field changes in different monitoring scenarios, enabling the system to maintain good adaptability and stability in various complex environments.

[0084] Please refer to Figure 2As shown, the present invention also includes a high-precision target tracking monitoring image processing system, which is used to implement the above-mentioned high-precision target tracking monitoring image processing method, including:

[0085] The light source acquisition module is used to collect the polarization angle distribution data and polarization degree distribution data of the incident light in the monitoring scene in real time, distinguish between constant natural light sources and dynamic artificial light sources, and generate the physical property signature of the dynamic artificial light source;

[0086] The mutation detection module is used to detect the change in light intensity at each pixel in the monitoring scene, and perform spatiotemporal clustering analysis on pixels whose changes exceed the ambient light baseline to form the physical boundaries of the light mutation area;

[0087] The trajectory analysis module is used to obtain the motion trajectory data of the target vehicle in continuous image frames, extract the position coordinate sequence and the motion direction sequence, and generate the motion trajectory space area of ​​the target vehicle;

[0088] The trajectory comparison module is used to mark the overlapping area as the dynamic interference source influence area when the physical boundary of the illumination mutation area overlaps with the motion trajectory space area;

[0089] The parameter calling module is used to predict the target vehicle's predicted position at the next frame image acquisition moment. If the predicted position is within the influence area of ​​the dynamic interference source, the exposure control parameters and gain control parameters corresponding to the physical property signature are called, and the focus control parameters are calculated based on the scene depth at the predicted position.

[0090] The parameter loading module is used to load the exposure control parameters, gain control parameters and focus control parameters into the sensor control unit during the vertical blanking period of the image sensor.

[0091] The above is a detailed description of an embodiment of the present invention. However, the content described is only a preferred embodiment of the present invention and should not be considered to limit the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A monitoring image processing method for high-precision target tracking, characterized in that: The following steps are involved: S1. Real-time collection of polarization angle distribution data and polarization degree distribution data of incident light in the monitoring scene, distinguishing between constant natural light sources and dynamic artificial light sources, and generating physical property signatures of dynamic artificial light sources; S2. Detect the change in light intensity at each pixel in the monitored scene, perform spatiotemporal clustering analysis on pixels whose changes exceed the ambient light baseline, and form the physical boundaries of the light mutation area. S3, obtaining the motion trajectory data of the target vehicle in continuous image frames, extracting the position coordinate sequence and the motion direction sequence, and generating the motion trajectory space area of ​​the target vehicle; S4. When the physical boundary of the illumination mutation area and the motion trajectory space area have geometric overlap, the overlapping area is marked as the dynamic interference source influence area; S5. Predict the target vehicle's predicted position at the next image capture moment. If the predicted position is within the dynamic interference source's influence area, call the exposure control parameters and gain control parameters corresponding to the physical attribute signature, and calculate the focus control parameters based on the scene depth at the predicted position. S6. During the vertical blanking period of the image sensor, the exposure control parameter, the gain control parameter, and the focus control parameter are loaded into the sensor control unit.

2. The monitoring image processing method for high-precision target tracking according to claim 1 is characterized in that: In S1, the process of generating the physical property signature of the dynamic artificial light source is as follows: A database of polarization characteristic feature templates containing metal reflective light sources and transmissive material light sources is pre-built. The polarization angle distribution data and polarization degree distribution data collected in real time are input into the feature matching calculation unit for processing. The similarity matrix value between the current polarization data and the feature templates in the database is calculated. When the similarity value exceeds the preset natural light discrimination threshold, the binary code of the corresponding feature template is extracted. The binary code includes combined information of a light source material type identifier and a spectral characteristic identifier. The complete binary code is output as a physical property signature and stored in a cache area.

3. The monitoring image processing method for high-precision target tracking according to claim 1 is characterized in that: In S2, the process of forming the physical boundary of the illumination mutation area is as follows: Continuously record and output the valid pixel event coordinates where the change in light intensity exceeds the ambient light baseline, perform density clustering operations on valid event points that are continuous in time and adjacent in space, and calculate the convex hull boundary vertex coordinate set data for each event cluster group; When the convex hull area change rate exceeds the upper threshold of natural light fluctuation, the effective light mutation area is confirmed. The physical boundary of the light mutation area is the closed polygon formed by connecting the coordinate sequence of the convex hull boundary vertices, and the coordinate value of the minimum circumscribed rectangle of the closed polygon is recorded.

4. The monitoring image processing method for high-precision target tracking according to claim 1 is characterized in that: In S3, the process of generating the target vehicle's motion trajectory space area is as follows: The historical position coordinate sequence of the key points of the target vehicle's outline is obtained through a feature point tracking algorithm. The key point position coordinates of adjacent image frames are connected to generate a continuous motion trajectory polyline. The width value of the target vehicle's physical size is expanded to both sides with the motion trajectory polyline as the center line. A strip buffer space area is generated to cover the uncertainty of the motion direction. The width parameter value of the strip buffer changes linearly with the target vehicle's motion speed. The final boundary of the motion trajectory space area is composed of an equidistant curve that is completely parallel to the trajectory line.

5. The monitoring image processing method for high-precision target tracking according to claim 1 is characterized in that: In S5, the process of predicting the target vehicle's position at the next frame image acquisition moment is as follows: Determine the physical constraint parameter value of the minimum turning radius based on the target vehicle type, fit the motion direction angle change curve using the position point data of the last three consecutive frames, correct the extreme point position of the direction change curve based on the minimum turning radius physical constraint, and extrapolate the position point of the next frame acquisition time along the corrected curve direction as the prediction origin; The range coordinate values ​​of the predicted position are generated with the prediction origin as the center point, and the long axis direction of the predicted position range is spatially aligned with the current movement direction of the target vehicle.

6. The monitoring image processing method for high-precision target tracking according to claim 1, characterized in that: In S5, the process of calling the exposure control parameters and gain control parameters corresponding to the physical attribute signature is as follows: A database table pre-stores the mapping relationship between physical property signatures and exposure parameters. The physical property signature of a metal reflective light source is mapped to a short exposure time parameter combination with an exposure time less than a first set threshold; the physical property signature of a transmissive material light source is mapped to a medium exposure parameter combination with an exposure time greater than the first set threshold and less than a second set threshold. When the predicted position is covered by the influence areas of multiple dynamic interference sources, the parameters corresponding to the interference source with the largest peak light intensity are selected; the focus control parameters are calculated through the binocular stereo vision system to calculate the average depth value of the predicted position; and the depth value is converted into the curvature control voltage output instruction parameter required by the liquid lens device.

7. The monitoring image processing method for high-precision target tracking according to claim 1, characterized in that: In S6, the process of loading the parameters into the image sensor control unit is as follows: The sensor control unit receives all control parameter instructions at the start of the vertical blanking period. The exposure control parameters are converted into clock configuration instructions for the charge accumulation period of the image sensor. The gain control parameters are converted into impedance adjustment configuration instructions for the programmable amplifier circuit. The focus control parameters are directly input into the dedicated drive circuit interface of the liquid lens device. The drive circuit compensates the voltage output value according to the real-time temperature sensor data. After completing the above conversion, a parameter loading completion ready signal is output, and the curvature stability state of the liquid lens device is verified in real time through a laser interferometer measurement device. When the exposure timing control signal is synchronized with the optical stability signal, the image acquisition action is triggered.

8. The monitoring image processing method for high-precision target tracking according to claim 7 is characterized in that: The process of verifying the curvature stability of the liquid lens device is as follows: A monochromatic coherent laser beam with a fixed wavelength is projected toward the predicted position. The high-speed image sensor unit captures the interference fringe pattern on the surface of the target object and calculates the correlation coefficient of the spatial phase distribution of the interference fringe pattern at adjacent acquisition moments. When the correlation coefficient value is continuously higher than the preset stability threshold, an optical stability signal is output. If the correlation coefficient value drops below the preset threshold during the exposure acquisition process, the acquisition process of the current image frame is immediately terminated.

9. The monitoring image processing method for high-precision target tracking according to claim 8, characterized in that: After the parameter loading is completed and the ready signal is output, the dynamic light field coherence verification is performed, the theoretical polarization distribution matrix corresponding to the physical property signature of the light source is called, and the theoretical light field gradient is calculated based on the direction vector of the motion trajectory space area; Acquire real-time phase distribution data from the laser interferometer. Perform a coherent matching operation on the theoretical light field gradient and the real-time phase data. When the matching similarity exceeds the dynamic environment threshold, the current parameters are maintained. If the matching similarity does not reach the threshold, a parameter reset instruction is sent to the sensor control unit.

10. A high-precision target tracking monitoring image processing system, used to implement the high-precision target tracking monitoring image processing method according to any one of claims 1 to 9, characterized in that: include: The light source acquisition module is used to collect the polarization angle distribution data and polarization degree distribution data of the incident light in the monitoring scene in real time, distinguish between constant natural light sources and dynamic artificial light sources, and generate the physical property signature of the dynamic artificial light source; The mutation detection module is used to detect the change in light intensity at each pixel in the monitoring scene, and perform spatiotemporal clustering analysis on pixels whose changes exceed the ambient light baseline to form the physical boundaries of the light mutation area; The trajectory analysis module is used to obtain the motion trajectory data of the target vehicle in continuous image frames, extract the position coordinate sequence and the motion direction sequence, and generate the motion trajectory space area of ​​the target vehicle; The trajectory comparison module is used to mark the overlapping area as the dynamic interference source influence area when the physical boundary of the illumination mutation area overlaps with the motion trajectory space area; The parameter calling module is used to predict the target vehicle's predicted position at the next frame image acquisition moment. If the predicted position is within the influence area of ​​the dynamic interference source, the exposure control parameters and gain control parameters corresponding to the physical property signature are called, and the focus control parameters are calculated based on the scene depth at the predicted position. The parameter loading module is used to load the exposure control parameters, gain control parameters and focus control parameters into the sensor control unit during the vertical blanking period of the image sensor.

Images