Automobile tunnel automatic thermal imaging visual auxiliary method and related equipment

Abstract

The application is suitable for the field of image processing technology, and provides a kind of automobile tunnel automatic thermal imaging visual auxiliary method and related equipment.The method fuses and judges position information and ambient light intensity information to obtain tunnel trigger signal;According to the tunnel trigger signal, the non-uniformity correction parameter estimation of infrared focal plane array original frame sequence is carried out, and the parameter estimation value is used as the initial value of interframe iterative correction, the non-uniformity of focal plane array response is corrected, and the non-uniformity correction frame is obtained;According to the pixel neighborhood content statistics of each local area in the non-uniformity correction frame, the local and global gray scale mapping relationship is determined respectively, and the content characteristics of each local area are compared according to the weighted contrast adjustment, to obtain contrast enhancement frame and display output.This method cooperates non-uniformity correction parameter estimation and interframe iterative correction, so that effective feature points can be extracted under low contrast thermal field conditions in tunnel, and correction parameter converges normally.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to an automatic thermal imaging visual assistance method and related equipment for automobile tunnels. Background Technology

[0002] Modern cars increasingly rely on visible light cameras for environmental perception and driving assistance. However, when a vehicle enters or exits a tunnel, the intensity of ambient light changes drastically in a very short time. Due to the lag in exposure adaptation, visible light cameras suffer from large areas of overexposure or underexposure, resulting in the inability to provide effective visual information during critical periods. Thermal imaging technology, which does not rely on ambient visible light, has potential applications.

[0003] However, when the vehicle-mounted uncooled infrared focal plane array sensor enters or exits a tunnel, the thermal field distribution inside the tunnel is highly uniform, resulting in extremely low overall image contrast. Existing scene adaptive non-uniformity correction methods rely on extractable effective feature points in the image to drive the convergence of correction parameters. Under the low-contrast thermal field conditions in tunnels, effective feature points are scarce, and the correction parameters cannot converge normally, resulting in the inability to effectively eliminate fixed-mode noise and a serious deterioration in the quality of thermal imaging images. Summary of the Invention

[0004] This invention provides an automatic thermal imaging visual assistance method and related equipment for automobile tunnels, in order to solve the technical problems of feature extraction failure and ineffective convergence of correction parameters in existing scene adaptive non-uniformity correction methods under low-contrast thermal field conditions in tunnels.

[0005] In a first aspect, embodiments of this application provide an automatic thermal imaging visual assistance method for vehicle tunnels, including: The location information and ambient light intensity information collected during vehicle movement are fused and judged to obtain the tunnel trigger signal; Based on the tunnel trigger signal, the non-uniformity correction parameters of the original frame sequence of the infrared focal plane array output by the thermal imaging sensor are estimated, and the estimated parameter values ​​are used as the initial values ​​for inter-frame iterative correction to correct the non-uniformity of the focal plane array response of the original frame sequence, thereby obtaining the non-uniformity corrected frame. Based on the pixel neighborhood content statistics of each local region in the non-uniformity correction frame, the local grayscale mapping relationship and the global grayscale mapping relationship are determined for the non-uniformity correction frame respectively. Based on the content characteristics of each local area, the non-uniformity correction frame is weighted and its contrast is adjusted according to the local grayscale mapping relationship and the global grayscale mapping relationship to obtain a contrast-enhanced frame. The contrast-enhanced frame is then displayed and output to achieve visual assistance for automatic thermal imaging of automobile tunnels.

[0006] Secondly, embodiments of this application provide an automatic thermal imaging visual assistance device for vehicle tunnels, comprising: The trigger judgment module is used to fuse and judge the location information and ambient light intensity information collected during vehicle movement to obtain the tunnel trigger signal; The uniformity correction module is used to estimate the non-uniformity correction parameters of the original frame sequence of the infrared focal plane array output by the thermal imaging sensor according to the tunnel trigger signal, and use the parameter estimate as the initial value for inter-frame iterative correction to correct the non-uniformity of the focal plane array response of the original frame sequence to obtain the non-uniformity correction frame. The mapping determination module is used to determine the local grayscale mapping relationship and the global grayscale mapping relationship for the non-uniformity correction frame based on the pixel neighborhood content statistics of each local region in the non-uniformity correction frame. The contrast enhancement module is used to perform weighted contrast adjustment on the non-uniformity correction frame according to the content characteristics of each local area using the local grayscale mapping relationship and the global grayscale mapping relationship, to obtain a contrast enhancement frame, and to display the contrast enhancement frame to achieve visual assistance for automatic thermal imaging of automobile tunnels.

[0007] Thirdly, embodiments of this application provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the above-described automatic thermal imaging visual assistance method for vehicle tunnels.

[0008] Fourthly, embodiments of this application provide a readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described automatic thermal imaging visual assistance method for vehicle tunnels.

[0009] Fifthly, embodiments of this application provide a computer program product, which includes a computer program that, when executed by a processor, enables the implementation of the steps of the above-described automatic thermal imaging visual assistance method for vehicle tunnels.

[0010] In one solution provided by the aforementioned automatic thermal imaging visual assistance method, device, equipment, medium, and program products for automobile tunnels, a tunnel trigger signal is obtained by fusing location information and ambient light intensity information. This precisely binds the start time of the thermal imaging processing flow to the tunnel scene, avoiding invalid processing in non-tunnel scenes. In the non-uniformity correction stage, this invention uses the estimated correction parameters as the initial values ​​for inter-frame iterative correction, ensuring an effective convergence starting point for the iterative process under low-contrast thermal field conditions. This overcomes the deficiency of traditional scene adaptive correction methods where parameters fail to converge due to scarce feature points. In the image enhancement stage, this invention simultaneously determines the local and global grayscale mapping relationship based on pixel neighborhood content statistics, and weights and adjusts both according to the content characteristics of each local region. This allows for differentiated contrast processing between the uniform thermal field background region and the target region, suppressing background noise amplification while preserving target details. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a schematic diagram of a vehicle tunnel automatic thermal imaging visual assistance system according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating an automatic thermal imaging visual assistance method for automobile tunnels according to an embodiment of the present invention; Figure 3 yes Figure 2 A schematic diagram of the implementation process of step S20; Figure 4 yes Figure 2 A schematic diagram of the implementation process of step S40; Figure 5 This is a schematic diagram of the structure of an automatic thermal imaging visual aid device for automobile tunnels according to an embodiment of the present invention; Figure 6 This is a schematic diagram of the structure of a computer device according to an embodiment of the present invention. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] It should be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or collections thereof. It should also be understood that, as used in this specification and the appended claims, the term "and / or" refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0015] Furthermore, in the description of this invention and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0016] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of the invention include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0017] It should be understood that the sequence number of each step in the following embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0018] To illustrate the technical solution of the present invention, specific embodiments are described below.

[0019] To address the problems mentioned above in the background art, this application provides an automatic thermal imaging visual assistance method, device, equipment, medium, and program product for vehicle tunnels. The automatic thermal imaging visual assistance method for vehicle tunnels provided by this invention can be applied to applications such as... Figure 1 The automatic thermal imaging visual assistance system for automobile tunnels shown includes a client and a server.

[0020] In one embodiment, such as Figure 2 As shown, an automatic thermal imaging visual assistance method for automobile tunnels is provided, which is applied to... Figure 1 Taking the automatic thermal imaging visual assistance system for automobile tunnels as an example, the following steps are included: S10: The location information and ambient light intensity information collected during vehicle movement are fused and judged to obtain the tunnel trigger signal; S20: Based on the tunnel trigger signal, perform non-uniformity correction parameter estimation on the original frame sequence of the infrared focal plane array output by the thermal imaging sensor, and use the parameter estimation value as the initial value for inter-frame iterative correction to correct the non-uniformity of the focal plane array response of the original frame sequence, thereby obtaining a non-uniformity corrected frame. S30: Based on the pixel neighborhood content statistics of each local region in the non-uniformity correction frame, determine the local gray-level mapping relationship and the global gray-level mapping relationship for the non-uniformity correction frame respectively. S40: Based on the content characteristics of each local area, the non-uniformity correction frame is weighted and its contrast is adjusted according to the local grayscale mapping relationship and the global grayscale mapping relationship to obtain a contrast-enhanced frame, and the contrast-enhanced frame is displayed and output to realize automatic thermal imaging visual assistance for automobile tunnels.

[0021] In this embodiment, the vehicle continuously collects location information and ambient light intensity information while in motion. The location information is output in real time by the onboard GPS module, and the ambient light intensity information is extracted from the brightness statistics of the current frame by the onboard visible light image sensor. The system continuously monitors the above two sources of information in the background and only initiates the subsequent thermal imaging processing flow when the trigger conditions are met, in order to avoid unnecessary computational overhead in non-tunnel scenarios.

[0022] Based on the above explanation, the triggering judgment does not rely solely on one piece of information, but rather fuses location information and light intensity information to output a tunnel trigger signal. This is because relying solely on GPS location judgment suffers from map data update lag, while relying solely on light intensity abrupt changes may misidentify scenes such as passing through tree shade or entering an underground parking lot as tunnel entrances. The fusion of the two pieces of information effectively reduces the probability of false triggers. For example, when GPS positioning shows that the vehicle is less than 50 meters from a known tunnel entrance, and the visible light frame brightness statistics show a monotonically decreasing trend over several consecutive frames, the system outputs a valid tunnel trigger signal and initiates the thermal imaging processing; if only one piece of information meets the conditions, the system continues to wait without triggering.

[0023] In this embodiment, upon receiving the tunnel trigger signal, the system immediately initiates non-uniformity correction processing on the original frame sequence of the infrared focal plane array output by the thermal imaging sensor. It should be noted that the vehicle-mounted uncooled infrared focal plane array faces a step change in ambient temperature when entering and exiting a tunnel, causing a shift in the response characteristics of each pixel, which manifests as fixed-pattern stripe noise in the image, either horizontally or vertically. Traditional correction methods based on shutter occlusion require interrupting image acquisition, making them unsuitable for real-time processing during driving. Therefore, this embodiment employs a scene-adaptive non-uniformity correction method, utilizing scene information between consecutive frames to drive the estimation and updating of correction parameters.

[0024] Specifically, non-uniformity correction is divided into two stages: parameter estimation and iterative correction. In the parameter estimation stage, the system preprocesses the original frame sequence to suppress noise interference, extracts feature matching relationships between adjacent frames to obtain inter-frame displacement, and then estimates the gain correction coefficient and bias correction coefficient of each pixel based on the inter-frame displacement, obtaining estimated non-uniformity correction parameters. In the iterative correction stage, the estimated parameters are used as the initial values ​​for the iterative process, linearly correcting the response values ​​of each pixel in the original frame. Through multiple frame iterations, residual fixed-pattern noise is gradually eliminated, resulting in a non-uniformity corrected frame.

[0025] In this embodiment, the non-uniformity correction frame has eliminated fixed-pattern stripe noise. However, due to the highly uniform thermal field distribution within the tunnel, the overall grayscale dynamic range of the correction frame is still compressed into a narrow range, resulting in insufficient grayscale difference between the target and the background. When displayed directly, it is difficult for the driver to distinguish targets such as pedestrians and vehicles. Therefore, subsequent processing further enhances the contrast of the correction frame.

[0026] Contrast enhancement processing targets the pixel distribution characteristics of various local regions in non-uniformity correction frames, calculating both local and global gray-level mapping relationships. Local gray-level mapping relationships are calculated based on the statistical content of pixel neighborhoods within each sub-block, capturing the local gray-level distribution features within that sub-block. Global gray-level mapping relationships, on the other hand, are calculated based on statistics across the entire image, reflecting the overall gray-level distribution of the entire frame. Each has its emphasis: local mapping is beneficial for highlighting details, while global mapping is beneficial for suppressing noise amplification in uniform background areas.

[0027] The system weights and fuses the local and global grayscale mapping relationships based on the content characteristics of each local region, and then applies this weighted fusion to the non-uniformity correction frame to obtain the contrast-enhanced frame. For areas rich in detail, the local grayscale mapping relationship has a higher weight, ensuring that the target outline is fully stretched; for uniformly heated background areas, the global grayscale mapping relationship has a higher weight, preventing background noise from being excessively amplified. For example, in the area where a pedestrian is located in a tunnel, the pedestrian's body temperature is higher than the background wall temperature, resulting in a larger difference in pixel grayscale within the corresponding sub-block. The local mapping weight is higher, effectively enhancing the pedestrian's outline; while in the sub-block corresponding to a large area of ​​uniformly heated tunnel wall, the pixel grayscale is highly concentrated, resulting in a higher global mapping weight, and no noise amplification occurs in the wall area. Finally, the contrast-enhanced frame is output to the vehicle's central control screen for display.

[0028] In one embodiment, the step of fusing and judging the location information and ambient light intensity information collected during vehicle movement to obtain the tunnel trigger signal includes: The location information collected during vehicle movement is compared with pre-stored tunnel geographical location data to obtain the location determination result; Based on the position determination result, the brightness statistics of the visible light frames collected by the vehicle image sensor are compared with thresholds to obtain the illumination change judgment result. The tunnel trigger signal is obtained by logically fusing the location determination result with the illumination change determination result.

[0029] In this embodiment, the system continuously reads the real-time location coordinates output by the GPS module during vehicle operation and compares these coordinates with a pre-stored tunnel geographic location database. The tunnel geographic location database records the geographic coordinates of each tunnel entrance and exit, as well as the corresponding warning buffer distance. The system calculates the distance between the current location coordinates and the coordinates of each tunnel entrance in the database. If the distance is less than a preset location warning threshold, the location judgment result is set to valid; otherwise, it is set to invalid.

[0030] It should be noted that a valid location determination result only indicates that the vehicle is geographically close to the tunnel; it does not directly trigger the thermal imaging process. Instead, it serves as a condition for initiating subsequent illumination change detection. The reason for this design is that illumination change detection itself has a certain computational overhead. If it were to run continuously throughout the entire journey, it would consume unnecessary processing resources. Using the location determination result as a prerequisite limits the activation range of illumination change detection to the area near the tunnel entrance.

[0031] Provided the location determination result is valid, the system extracts brightness statistics from the visible light frames currently captured by the vehicle-mounted image sensor. Specifically, it calculates the average grayscale value of the pixels in the current frame to obtain the average brightness of the current frame, and calculates the difference between this average and the average brightness of the previous several frames to obtain the inter-frame brightness change. If the inter-frame brightness change exceeds a preset illumination change threshold, and multiple consecutive frames show the same brightness change trend, the illumination change determination result is set to valid; otherwise, it is set to invalid. For example, when a vehicle enters a tunnel from a bright outdoor environment, the average brightness of the visible light frames will continuously and rapidly decrease within several frames, and the inter-frame brightness change exceeds the threshold, triggering the illumination change determination result to be valid.

[0032] The system logically fuses the location determination result and the illumination change determination result, outputting a valid tunnel trigger signal only when both are valid. If the location determination result is invalid, the thermal imaging process will not be triggered even if an illumination change occurs, thus eliminating interference from illumination changes in non-tunnel scenes. If the location determination result is valid but the illumination change determination result is invalid, the system will also not trigger, avoiding false triggering due to map data deviations. When both determination results meet the conditions, the system outputs a tunnel trigger signal, and the subsequent non-uniformity correction and contrast enhancement processes are then initiated.

[0033] In one embodiment, such as Figure 3 As shown, step S20 specifically includes the following steps: S21: The gain correction coefficient is recursively calculated for the pixel response values ​​of each frame in the original frame sequence according to the median ratio constant constraint, and the calculation results are low-pass spatially filtered to obtain the pre-enhanced frame sequence. S22: Extract and match feature points in adjacent frames of the pre-enhanced frame sequence to obtain the inter-frame displacement. S23: Based on the inter-frame displacement, the offset correction coefficient of each pixel is recursively estimated using Kalman filtering. The estimated value is used as the initial offset correction coefficient. Based on the inter-frame displacement, the gain correction coefficient and offset correction coefficient of each pixel are iteratively updated using the least mean square criterion to obtain the estimated value of the non-uniformity correction parameter. S24: Based on the gain correction coefficient and bias correction coefficient in the estimated non-uniformity correction parameters, perform a linear affine transformation on the response values ​​of each pixel in the original frame sequence to obtain an initial corrected frame sequence; S25: Calculate the pixel residuals in the overlapping areas of adjacent frames in the initial correction frame sequence, and determine the learning rate gating coefficients for each pixel position based on the residual amplitude and the local residual standard deviation to obtain the gating mask; S26: Selectively iteratively update the gain correction coefficient and offset correction coefficient of the initial correction frame sequence according to the gated mask to obtain a non-uniform correction frame.

[0034] In this embodiment, the response of each pixel of the infrared focal plane array to incident radiation follows a linear model, meaning that the output value of each pixel is equal to the product of its gain coefficient and the actual incident radiation, plus a bias coefficient. Because the gain coefficients and bias coefficients of each pixel differ, significant fixed-pattern fringe noise is superimposed on the directly output raw frame. The goal of the correction is to apply a linear compensation transform to each pixel of the raw frame, estimating and continuously updating the gain correction coefficients and bias correction coefficients of each pixel to make the responses of each pixel more consistent. The processing flow of this embodiment revolves around how to accurately complete the above estimation and updating under the low-contrast thermal field conditions of a tunnel.

[0035] In the pre-enhancement stage, the system first recursively calculates the gain correction coefficients using the median ratio constant constraint method on the original frame sequence. This method is based on the assumption that the gain ratio between adjacent pixels tends to be constant across multiple frames. Starting from the top left corner of the image, it recursively calculates the coefficients pixel-by-pixel along the row and column directions. The gain correction coefficients at subsequent positions depend on the parameters of the adjacent positions that have already been calculated. It should be noted that this recursive process has some cumulative error, but this error is acceptable for subsequent image registration. Since the median ratio method is sensitive to discrete defect pixels, after the recursive calculation of the gain correction coefficients, a Gaussian low-pass spatial filter is applied to the results to suppress residual discrete bright spot noise and spatial white noise, resulting in a pre-enhanced frame sequence. For example, in tunnel scenes, the thermal distribution texture, originally obscured by stripe noise, becomes visible after pre-enhancement, creating conditions for subsequent feature extraction.

[0036] In the feature extraction and matching stage, the system extracts SIFT and SURF feature points from adjacent frames in the pre-enhanced frame sequence, respectively. These two feature extraction methods are chosen because tunnel thermal field images lack strong corner features, while SIFT and SURF are both speckle-type feature extractors, exhibiting strong responsiveness to cloud-like thermal distribution regions at different gray levels. After feature point extraction, the MLESAC algorithm is used to match feature points between two frames and estimate the inter-frame translation displacement.

[0037] Based on the above explanation, to balance registration efficiency and accuracy, the system introduces a dynamic matching threshold adjustment strategy. The correct matching rate is calculated as the ratio of the number of correct matches to the total number of matches, and its judgment threshold is dynamically adjusted with the number of correction iterations: In the initial stage of correction, the residual noise in the fixed pattern is relatively large, making registration more difficult, so the system appropriately lowers the correct matching rate threshold to ensure the registration success rate; as the correction iterations progress and the residual noise gradually decreases, the threshold is gradually increased, and the registration accuracy tightens accordingly. Experiments show that when the signal-to-noise ratio is not less than 3, the inter-frame displacement estimation error of the above feature registration method can be controlled within 0.5 pixels, and the matching success rate exceeds 90%.

[0038] In the parameter estimation phase, the system estimates the gain correction coefficient and offset correction coefficient in two steps, based on the inter-frame displacement. The first step uses Kalman filtering to perform a fast recursive estimation of the offset correction coefficient. The update increment of the offset correction coefficient is set as the state variable of the Kalman filter, the pixel residuals in the overlapping area of ​​adjacent correction frames are set as the observations, the observation matrix is ​​determined by the inter-frame registration results, and the observation noise is set as zero-mean white noise. The system updates the state estimate frame by frame according to the standard Kalman recursive formula. After each state estimation, the estimated update increment is superimposed on the current offset correction coefficient to complete the update of the offset correction coefficient. It should be noted that the direct inversion of the Kalman filter gain matrix is ​​computationally intensive. In this embodiment, it is transformed into a system of linear equations and solved iteratively using the conjugate gradient method, achieving efficient computation with the help of sparse matrix tools. After the first 25 iterations, the offset correction coefficient estimated by the Kalman filter is used as the initial value for the second step.

[0039] The second step employs the inter-frame registration least mean square method to jointly iteratively update the gain correction coefficients and bias correction coefficients. For each pixel position in the overlapping region of adjacent frames, the pixel residual between the current correction frame and the previous correction frame after displacement compensation is calculated. Aiming to minimize the sum of squares of the residuals in the overlapping region, partial derivatives are calculated with respect to the gain correction coefficients and bias correction coefficients to obtain their respective update increments. The update increment of the gain correction coefficient is equal to the product of the negative residual value and the original pixel response value multiplied by the learning rate; the update increment of the bias correction coefficient is equal to the negative residual value multiplied by the learning rate. These two are then superimposed on the corresponding coefficients of the previous frame to complete the update. Using the estimated bias correction coefficient output from the Kalman filter as the starting point for iteration allows the iteration to begin from a position closer to the true value, significantly accelerating the convergence speed. After the above two estimation steps, the estimated non-uniformity correction parameters for each pixel are obtained, including the gain correction coefficients and bias correction coefficients.

[0040] During the linear correction phase, the system substitutes the estimated non-uniformity correction parameters into the correction model and performs a linear affine transformation on the response values ​​of each pixel in the original frame sequence. This involves multiplying the original output value of each pixel by the corresponding gain correction coefficient and adding the bias correction coefficient to obtain the corrected output value for that pixel. After performing the above transformation on all pixels in the entire frame, the initial corrected frame sequence is obtained. At this point, the fixed-pattern stripe noise has been initially eliminated, but residual errors caused by discrete defect pixels or target motion may still exist in some areas.

[0041] During the gated mask generation stage, the system calculates the difference pixel by pixel in the overlapping areas of adjacent frames in the initial calibration frame sequence to obtain a difference matrix, and calculates the global standard deviation of the difference matrix as a reference. The learning rate gating coefficients at each pixel position are determined according to the following three strategies.

[0042] The first strategy targets overly bright or dark pixels: For each pixel position in the difference matrix, if the absolute value of the difference at that position exceeds three times the global standard deviation, the gating coefficient at that position is set to 0, thus blocking the update of the correction parameters at that position and eliminating the interference of discrete defective pixels on the iteration process.

[0043] The second strategy targets target edge pixels: For each pixel position in the difference matrix, calculate the local standard deviation of the difference in a 3×3 neighborhood centered on that position. If the local standard deviation exceeds three times the global standard deviation, set the gating coefficient at that position to 0. This is used to identify and mask target edge regions, preventing ghost artifacts from being generated near the edges when the target's motion direction is inconsistent with the overall inter-frame displacement direction.

[0044] The third strategy targets the extended target area: Pixel locations that failed the above two strategies are marked as candidate masking locations, and the candidate masking status of each pixel location is recorded within 5 consecutive frames. For locations marked as candidate masking within 5 consecutive frames and whose corresponding connected region area is less than 20 pixels, they are determined to be discrete defect pixels rather than real moving targets, and their gate coefficient is restored to 1, thus not being protected; conversely, candidate masking locations with a connected region area of ​​not less than 20 pixels are determined to be real target areas, and their gate coefficient is kept at 0.

[0045] By combining the results of the three strategies, a logical OR operation is performed on each pixel position to obtain the final gated mask.

[0046] During the selective iterative update phase, the system replaces the learning rate at each pixel location with the product of the original learning rate and the gating coefficient, performing gated iterative updates on the gain and bias correction coefficients of the initial corrected frame sequence. For pixel locations with a gating coefficient of 0, the learning rate is set to zero, and the correction parameters at that location remain unchanged in this iteration, thus protecting the target area from miscorrection. For pixel locations with a gating coefficient of 1, the iteration proceeds at the normal learning rate, and residual non-uniform noise is continuously eliminated. For example, in a scene where a pedestrian target exists within a tunnel, the pixel location corresponding to the pedestrian area is set to 0 in the gating mask, and the correction parameter update pauses in that area, preventing ghosting of the pedestrian silhouette due to correction iterations. Meanwhile, the gating coefficient for background areas such as the tunnel wall is 1, allowing correction iterations to proceed normally in the background area, and stripe noise continues to decrease. After the above selective iterative update, a non-uniform corrected frame is obtained.

[0047] In one embodiment, such as Figure 4 As shown, step S30 specifically includes the following steps: S31: For each pixel in the non-uniformity correction frame, count the number of pixels whose gray value difference with the pixel exceeds a preset threshold within a preset neighborhood range, obtain the neighborhood condition frequency of each gray level, and summarize the neighborhood condition frequency of all pixels in each sub-block to obtain the neighborhood condition histogram of each sub-block. S32: Perform cumulative summation mapping calculation based on the neighborhood condition histogram of each sub-block to obtain the local gray-level mapping relationship of each sub-block; S33: Sum the neighborhood condition histograms of each sub-block of the non-uniformity correction frame full image, and perform cumulative summation mapping calculation on the summation results to obtain the global grayscale mapping relationship.

[0048] In this embodiment, the corrected frame after non-uniformity correction has eliminated fixed-pattern stripe noise. However, due to the highly uniform thermal field distribution within the tunnel, the overall grayscale dynamic range of the image is compressed into a narrow range. If the traditional histogram equalization method is directly applied to the corrected frame, the grayscale level of the uniform thermal field background region accounts for a very high proportion in the original histogram. After equalization, a large grayscale range will be allocated to the background region, resulting in excessive amplification of background noise, while the target detail regions that truly need enhancement do not receive sufficient grayscale stretching. To solve this problem, this embodiment introduces a neighborhood conditional histogram to replace the traditional histogram as the basis for mapping calculation.

[0049] It's important to note that the fundamental difference between neighborhood-conditional histograms and traditional histograms lies in their statistical methods. Traditional histograms count the total number of pixels at each gray level in an image. Gray levels in uniform background regions, due to their large pixel count, dominate the histogram, causing subsequent mapping to be biased towards the background. Neighborhood-conditional histograms, however, introduce an additional constraint of neighborhood difference when counting the frequency of a gray level. Only when the gray value of a pixel differs from the gray values ​​of its neighbors exceeds a preset threshold does that pixel contribute to the frequency count of the corresponding gray level. As a result, pixels located in uniform background regions, with generally small gray-value differences from their neighbors, contribute very little to the histogram frequency count; while pixels located at the edges of objects or in detail areas, with larger gray-value differences from their neighbors, contribute significantly more. In the final neighborhood conditional histogram, the gray level frequency of the detail region is higher and that of the background region is lower. Based on this, the mapping calculation can allocate more gray level range to the detail region, thereby effectively stretching the target contrast without amplifying the background noise.

[0050] In the neighborhood conditional frequency statistics stage, the system divides the non-uniformity correction frame into several non-overlapping sub-blocks of the same size, with each sub-block measuring 64×64 pixels. For each pixel within each sub-block, a square neighborhood with a side length of 2r+1 is taken centered on that pixel. The number of pixels in the neighborhood whose absolute value of the difference between their grayscale value and that of the center pixel exceeds a preset threshold t is counted. This number is divided by the total number of pixels in the neighborhood minus 1 and then added to the frequency corresponding to the grayscale level of the center pixel. After performing the above statistics for each pixel in the sub-block, the neighborhood conditional histogram of the sub-block is obtained. Based on the above explanation, for sub-blocks with highly concentrated grayscale uniform hot field backgrounds, the grayscale difference between each pixel and its neighboring pixels is generally less than the threshold t, and the overall frequency value of the neighborhood conditional histogram is very low. However, for sub-blocks containing targets such as pedestrians or vehicles, the grayscale difference between pixels at the target edges and their neighboring pixels is large, and the frequency value of the corresponding grayscale level is significantly higher. For example, in a tunnel scene, the pedestrian's body temperature is higher than the background wall temperature. The difference between the pixel grayscale value at the pedestrian's outline edge and the adjacent background pixels is large, resulting in a large number of neighboring pixels exceeding the threshold t. Consequently, this grayscale level has a higher frequency in the neighborhood conditional histogram, and subsequent mapping calculations will allocate a larger output grayscale range to it. In this embodiment, the neighborhood radius r is set to 2, and the threshold t is set to 10 for a 14-bit infrared image.

[0051] In the local gray-level mapping calculation stage, the system performs cumulative summation mapping calculations on the neighborhood conditional histogram of each sub-block. Specifically, the frequency values ​​of each gray level in the neighborhood conditional histogram of the sub-block are accumulated sequentially from low gray level to high gray level to obtain the cumulative frequency value of each gray level. Then, the cumulative frequency value is normalized to the output gray level range to obtain the gray level mapping table of the sub-block, i.e., the local gray-level mapping relationship. The local gray-level mapping relationship describes the output gray level corresponding to each input gray level in the sub-block. Different sub-blocks correspond to different mapping tables due to the different distribution of their neighborhood conditional histograms, enabling differentiated gray-level stretching based on the local content characteristics of each sub-block.

[0052] During the global grayscale mapping calculation phase, the system sums up the neighborhood condition histograms of all sub-blocks in the entire image, block by block. Specifically, for each grayscale level, the frequency values ​​of the neighborhood conditions of all sub-blocks at that grayscale level are summed to obtain the summarized neighborhood condition histogram result for the entire image. This summarized result is then calculated using the same cumulative summation mapping method as the local grayscale mapping to obtain the global grayscale mapping relationship.

[0053] In one embodiment, the step of weighted contrast adjustment of the non-uniformity correction frame based on the content features of each local region using the local grayscale mapping relationship and the global grayscale mapping relationship to obtain a contrast-enhanced frame includes: The Sobel operator is applied to the non-uniformity correction frame to obtain the edge information of each pixel position; Based on the ratio of the sum of frequencies of the neighborhood condition histograms of each sub-block to the total number of pixels in the sub-block, a weighting coefficient for the local and global gray-level mapping relationship is determined for each sub-block. The local and global gray-level mapping relationships are then weighted using the weighting coefficients, and gray-level remapping is performed on each sub-block of the non-uniformity correction frame to obtain a locally enhanced frame. Based on the edge information, the contrast adjustment parameters for each pixel position are determined. The Weber contrast of the local enhanced frame and the global grayscale mapping result are weighted and superimposed using the contrast adjustment parameters to obtain the contrast enhanced frame.

[0054] In this embodiment, after calculating the local and global grayscale mapping relationships, the system first applies the Sobel operator to the non-uniformity correction frame to extract edge information. Specifically, Sobel convolution is performed on the non-uniformity correction frame in both the horizontal and vertical directions to obtain gradient maps in two directions. The composite result of the gradient magnitudes in the two directions is taken as the edge response value of each pixel location, and the edge response value is linearly stretched to the same dynamic range as the subsequent enhancement result to obtain the edge information of each pixel location. The edge information is used as a metric for region features in the subsequent determination of contrast adjustment parameters. Pixel locations with higher edge response values ​​correspond to target contours or detail regions, while pixel locations with lower edge response values ​​correspond to uniform thermal field background regions.

[0055] It should be noted that edge information extraction is applied to the non-uniformity correction frame rather than the subsequent enhancement result. This is because the correction frame retains the original pixel grayscale distribution relationship. Using the gradient information of the correction frame as the measurement basis for regional features can truly reflect the richness of detail of each pixel position in the original scene, avoiding the introduction of additional errors due to changes in grayscale distribution during the enhancement process.

[0056] During the grayscale remapping stage, the system determines a weighting coefficient between the local and global grayscale mapping relationships for each sub-block. The weighting coefficient is calculated as the ratio of the sum of frequencies in the neighborhood conditional histogram of the sub-block to the total number of pixels in the sub-block. Since the frequencies in the neighborhood conditional histogram are only accumulated when there is a sufficient grayscale difference between a pixel and its neighborhood, a larger ratio of the sum of frequencies to the total number of pixels indicates a higher proportion of detailed areas within the sub-block, resulting in a larger weighting coefficient and a higher weight for the local grayscale mapping relationship during subsequent remapping. Conversely, for sub-blocks with a uniform hot field background, a smaller ratio of the sum of frequencies to the total number of pixels results in a smaller weighting coefficient and a higher weight for the global grayscale mapping relationship during remapping.

[0057] Based on the above explanation, the system linearly weights the local and global grayscale mapping relationships of each sub-block using weighted coefficients to obtain the fused mapping relationship for that sub-block. The fused mapping relationship is equal to the weighted average of the global and local grayscale mapping relationships using the weighted coefficients. The closer the weighted coefficient is to 1, the more the fused mapping relationship tends to be a local grayscale mapping relationship; the closer the weighted coefficient is to 0, the more the fused mapping relationship tends to be a global grayscale mapping relationship. Then, the grayscale values ​​of each pixel within the sub-block are remapped using the fused mapping relationship, and bilinear interpolation is performed on the mapping results between adjacent sub-blocks to eliminate stitching artifacts at the sub-block boundaries. After completing the above processing for all sub-blocks in the entire image, a locally enhanced frame is obtained. For example, in a tunnel scene, sub-blocks containing pedestrian targets have larger weighted coefficients and higher weights in the local grayscale mapping relationship, resulting in sufficient local contrast stretching of the pedestrian outline; while sub-blocks corresponding to large-area uniformly heated walls have smaller weighted coefficients and higher weights in the global grayscale mapping relationship, preventing excessive amplification of noise in the wall area.

[0058] During the contrast adjustment stage, the system determines two adaptive parameters for each pixel location based on the edge information: a regularization parameter and a contrast adjustment parameter. The regularization parameter is obtained by multiplying a preset base regularization coefficient by the edge information of that pixel location; the higher the edge response value, the larger the regularization parameter corresponding to the pixel location. The contrast adjustment parameter is determined by the regularization parameter and a preset base contrast coefficient, with its value ranging from 1 to 1 plus the base contrast coefficient. Pixel locations with higher edge response values ​​have larger contrast adjustment parameters, while pixel locations with lower edge response values ​​have contrast adjustment parameters that are closer to 1.

[0059] Based on this, the system calculates the Weber contrast at each pixel location for both the local enhancement frame and the global grayscale mapping result. Weber contrast is defined as the difference between the pixel's grayscale value and the average grayscale value of its local neighborhood, divided by the average grayscale value of the neighborhood, reflecting the contrast degree of the pixel relative to its local background. The system uses a regularization parameter to perform a weighted average of the Weber contrast of the local enhancement frame and the global grayscale mapping result to obtain a baseline contrast. Then, it scales the baseline contrast using a contrast adjustment parameter to obtain the target contrast. Finally, based on the target contrast and the average grayscale value of the pixel's neighborhood, the final output grayscale value of the pixel is calculated. After performing the above processing on all pixels in the entire image, a contrast-enhanced frame is obtained.

[0060] For example, in a scenario where pedestrians and distant vehicles coexist in a tunnel, the pixel edge response values ​​at the edges of pedestrian outlines are high, resulting in a larger contrast adjustment parameter. Weber contrast is fully amplified in this area, making the pedestrian outlines clearly discernible. Distant vehicles, due to their greater distance, have blurred outlines in thermal imaging. The corresponding pixel edge response values ​​are moderate, so the contrast adjustment parameter is set to a middle value, resulting in a moderate enhancement of the vehicle outlines. The edge response values ​​of large-area uniformly heated wall regions are close to zero, and the contrast adjustment parameter approaches 1. The grayscale output in this region remains largely unchanged, without introducing additional noise amplification. In the final contrast-enhanced frame, the grayscale difference between the target area and the background area is reasonably distributed, making it suitable for providing the driver with a clear tunnel thermal imaging auxiliary image on the central control screen.

[0061] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0062] In one embodiment, an automatic thermal imaging visual assistance device for automobile tunnels is provided, which corresponds one-to-one with the automatic thermal imaging visual assistance method for automobile tunnels described in the above embodiments. For example... Figure 5 As shown, the automatic thermal imaging visual assistance device for automobile tunnels includes a trigger judgment module 501, a uniformity correction module 502, a mapping determination module 503, and a contrast enhancement module 504. Detailed descriptions of each functional module are as follows: The trigger judgment module 501 is used to fuse and judge the location information and ambient light intensity information collected during vehicle movement to obtain the tunnel trigger signal; The uniformity correction module 502 is used to estimate the non-uniformity correction parameters of the original frame sequence of the infrared focal plane array output by the thermal imaging sensor according to the tunnel trigger signal, and use the parameter estimate as the initial value for inter-frame iterative correction to correct the non-uniformity of the focal plane array response of the original frame sequence to obtain a non-uniformity correction frame. The mapping determination module 503 is used to determine the local grayscale mapping relationship and the global grayscale mapping relationship for the non-uniformity correction frame based on the pixel neighborhood content statistics of each local region in the non-uniformity correction frame. The contrast enhancement module 504 is used to perform weighted contrast adjustment on the non-uniformity correction frame according to the content characteristics of each local area using the local grayscale mapping relationship and the global grayscale mapping relationship, to obtain a contrast enhancement frame, and to display and output the contrast enhancement frame to realize automatic thermal imaging visual assistance for automobile tunnels.

[0063] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0064] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0065] This application also provides a computer device, such as... Figure 6 As shown, the computer device includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor. When the processor executes the computer program, it implements the steps in any of the above method embodiments, or when the processor executes the computer program, it implements the functions of each module / unit in the above device embodiments.

[0066] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete this application. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the computer device.

[0067] Those skilled in the art will understand that Figure 6 The computer device described is merely an example and does not constitute a limitation on the computer device. It may include more or fewer components than shown, or combine certain components, or different components. For example, the computer device may also include input / output devices, network access devices, buses, etc.

[0068] The aforementioned processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0069] The memory can be an internal storage unit of the computer device, such as a hard drive or RAM. The memory can also be an external storage device of the computer device, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card. Furthermore, the memory can include both internal and external storage units of the computer device.

[0070] This application also provides a readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the various method embodiments above.

[0071] This application provides a computer program product that, when run on an electronic device, enables the electronic device to perform the steps described in the various method embodiments above.

[0072] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a photographing device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0073] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0074] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0075] In the embodiments provided in this application, it should be understood that the disclosed apparatus / devices and methods can be implemented in other ways. For example, the apparatus / device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0076] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0077] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for automatic thermal imaging visual assistance in vehicle tunnels, characterized in that, The automatic thermal imaging visual assistance method for vehicle tunnels includes: The location information and ambient light intensity information collected during vehicle movement are fused and judged to obtain the tunnel trigger signal; Based on the tunnel trigger signal, the non-uniformity correction parameters of the original frame sequence of the infrared focal plane array output by the thermal imaging sensor are estimated, and the estimated parameter values ​​are used as the initial values ​​for inter-frame iterative correction to correct the non-uniformity of the focal plane array response of the original frame sequence, thereby obtaining the non-uniformity corrected frame. Based on the pixel neighborhood content statistics of each local region in the non-uniformity correction frame, the local grayscale mapping relationship and the global grayscale mapping relationship are determined for the non-uniformity correction frame respectively. Based on the content characteristics of each local area, the non-uniformity correction frame is weighted and its contrast is adjusted according to the local grayscale mapping relationship and the global grayscale mapping relationship to obtain a contrast-enhanced frame. The contrast-enhanced frame is then displayed and output to achieve visual assistance for automatic thermal imaging of automobile tunnels.

2. The automatic thermal imaging visual assistance method for vehicle tunnels according to claim 1, characterized in that, The process of fusing and judging the location information and ambient light intensity information collected during vehicle movement to obtain the tunnel trigger signal includes: The location information collected during vehicle movement is compared with pre-stored tunnel geographical location data to obtain the location determination result; Based on the position determination result, the brightness statistics of the visible light frames collected by the vehicle image sensor are compared with thresholds to obtain the illumination change judgment result. The tunnel trigger signal is obtained by logically fusing the location determination result with the illumination change determination result.

3. The automatic thermal imaging visual assistance method for vehicle tunnels according to claim 1, characterized in that, The step of estimating the non-uniformity correction parameters of the original frame sequence of the infrared focal plane array output by the thermal imaging sensor based on the tunnel trigger signal includes: Gain correction coefficients are recursively calculated for the pixel response values ​​of each frame in the original frame sequence under the constraint of constant median ratio, and the calculation results are subjected to low-pass spatial filtering to obtain the pre-enhanced frame sequence. Feature points are extracted and matched from adjacent frames in the pre-enhanced frame sequence to obtain the inter-frame displacement. Based on the inter-frame displacement, the bias correction coefficients of each pixel are recursively estimated using Kalman filtering. The estimated values ​​are used as the initial bias correction coefficients. Based on the inter-frame displacement, the gain correction coefficients and bias correction coefficients of each pixel are iteratively updated using the least mean square criterion to obtain the estimated values ​​of the non-uniformity correction parameters.

4. The automatic thermal imaging visual assistance method for vehicle tunnels according to claim 1, characterized in that, The step of using the parameter estimates as initial values ​​for inter-frame iterative correction to correct the non-uniformity of the focal plane array response of the original frame sequence, resulting in a non-uniformity corrected frame, includes: Based on the gain correction coefficient and bias correction coefficient in the estimated non-uniformity correction parameters, a linear affine transformation is performed on the response values ​​of each pixel in the original frame sequence to obtain the initial corrected frame sequence. Pixel residuals are calculated for the overlapping regions of adjacent frames in the initial correction frame sequence. Learning rate gating coefficients are determined for each pixel position based on the residual magnitude and the local residual standard deviation to obtain a gating mask. The gain correction coefficients and bias correction coefficients of the initial correction frame sequence are selectively iteratively updated according to the gated mask to obtain a non-uniform correction frame.

5. The automatic thermal imaging visual assistance method for vehicle tunnels according to claim 1, characterized in that, The step of determining the local grayscale mapping relationship and the global grayscale mapping relationship for the non-uniformity correction frame based on the pixel neighborhood content statistics of each local region in the non-uniformity correction frame includes: For each pixel in the non-uniformity correction frame, the number of pixels whose grayscale value differs from the pixel's grayscale value by more than a preset threshold within a preset neighborhood range is counted to obtain the neighborhood condition frequency of each grayscale level. The neighborhood condition frequencies of all pixels in each sub-block are then summarized to obtain the neighborhood condition histogram of each sub-block. The local grayscale mapping relationship of each sub-block is obtained by performing cumulative summation mapping calculation based on the neighborhood condition histogram of each sub-block. The neighborhood condition histograms of each sub-block of the non-uniformity correction frame are summed and aggregated on a sub-block basis, and the aggregated results are cumulatively summed and mapped to obtain the global grayscale mapping relationship.

6. The automatic thermal imaging visual assistance method for vehicle tunnels according to claim 1, characterized in that, The step of weighting the contrast of the non-uniformity correction frame according to the content features of each local region and the local gray-level mapping relationship and the global gray-level mapping relationship to obtain the contrast-enhanced frame includes: The Sobel operator is applied to the non-uniformity correction frame to obtain the edge information of each pixel position; Based on the ratio of the sum of frequencies of the neighborhood condition histograms of each sub-block to the total number of pixels in the sub-block, a weighting coefficient for the local and global gray-level mapping relationship is determined for each sub-block. The local and global gray-level mapping relationships are then weighted using the weighting coefficients, and gray-level remapping is performed on each sub-block of the non-uniformity correction frame to obtain a locally enhanced frame. Based on the edge information, the contrast adjustment parameters for each pixel position are determined. The Weber contrast of the local enhanced frame and the global grayscale mapping result are weighted and superimposed using the contrast adjustment parameters to obtain the contrast enhanced frame.

7. An automatic thermal imaging visual aid device for automobile tunnels, characterized in that, include: The trigger judgment module is used to fuse and judge the location information and ambient light intensity information collected during vehicle movement to obtain the tunnel trigger signal; The uniformity correction module is used to estimate the non-uniformity correction parameters of the original frame sequence of the infrared focal plane array output by the thermal imaging sensor according to the tunnel trigger signal, and use the parameter estimate as the initial value for inter-frame iterative correction to correct the non-uniformity of the focal plane array response of the original frame sequence to obtain the non-uniformity correction frame. The mapping determination module is used to determine the local grayscale mapping relationship and the global grayscale mapping relationship for the non-uniformity correction frame based on the pixel neighborhood content statistics of each local region in the non-uniformity correction frame. The contrast enhancement module is used to perform weighted contrast adjustment on the non-uniformity correction frame according to the content characteristics of each local area using the local grayscale mapping relationship and the global grayscale mapping relationship, to obtain a contrast enhancement frame, and to display the contrast enhancement frame to achieve visual assistance for automatic thermal imaging of automobile tunnels.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the automatic thermal imaging visual assistance method for vehicle tunnels as described in any one of claims 1 to 6.

9. A readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the automatic thermal imaging visual assistance method for vehicle tunnels as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, enables the implementation of the steps of the automatic thermal imaging visual assistance method for vehicle tunnels as described in any one of claims 1 to 6.

Images