Suspension track projection target thermal imaging processing method and system

Abstract

The application provides a suspension track projection target thermal imaging processing method and system, and belongs to the technical field of thermal imaging detection. The pixel mapping strategy combining offline calibration and online optical flow fine adjustment effectively overcomes the spatial mapping dynamic drift problem caused by the continuous movement of the suspension track, significantly improves the frame-by-frame alignment accuracy of the thermal imaging image and the projection pattern. Meanwhile, the application can accurately identify the hit hot spot from the complex background temperature rise interference, and further iteratively correct the hot spot boundary by using the heat diffusion direction, so that the reconstructed hot spot shape is more consistent with the real heat conduction law, thereby generating a projection calibration result highly consistent with the hot spot diffusion trend.

Description

Technical Field

[0001] This invention belongs to the field of thermal imaging detection technology, specifically relating to a method and system for thermal imaging processing of suspended track projection targets. Background Technology

[0002] In automated shooting training and target reporting systems, thermal imaging technology has become an important means of detecting bullet impact points. When a bullet impacts a metal target, kinetic energy is converted into heat energy, forming a localized transient hot spot in the impact area. By acquiring the temperature distribution of the target surface using a thermal imager, the impact location can be determined. To further enhance the intuitiveness of human-computer interaction, enhanced target reporting schemes combining thermal imagers and projectors have emerged in recent years: after the thermal imager detects the hot spot, the projector projects information such as the impact location and ring number back onto the target surface, achieving visual feedback. In fixed target scenarios, this technology has made some progress, mainly relying on offline calibration to establish a static spatial mapping relationship between the thermal imager and the projector, and using temperature thresholds or simple frame difference methods to extract the hot spot.

[0003] However, existing methods face significant challenges when the target is suspended on a moving track (such as reciprocating or swinging). First, the spatial mapping relationship between the thermal imager and the projector changes dynamically with the target's movement, rendering static calibration results inapplicable. Recalibrating the target in three dimensions frame by frame would be computationally intensive and fail to meet real-time requirements. Second, the background temperature distribution on the target surface during movement may experience non-hit temperature rise interference due to factors such as airflow and friction, making traditional threshold segmentation prone to false detections.

[0004] More importantly, the hot spot at the point of impact is not an ideal rigid region; its temperature distribution exhibits a diffusion pattern with a high center and gradually decreasing edges. Furthermore, as heat is conducted to the surrounding area, the hot spot boundary dynamically expands. Existing hot spot recognition algorithms mostly rely on connected component analysis or simple morphological processing, failing to effectively model the boundary evolution during heat diffusion. This leads to discrepancies between the extracted hot spot shape and the actual heat distribution area, thus affecting the accuracy and visual effect of the projection calibration results. In addition, there is a lack of efficient and robust methods for frame-by-frame spatial alignment between thermal imaging images and projected images in motion scenarios. Existing technologies either rely on additional structured light measurements or assume that the camera and target are relatively stationary, both of which are ill-suited to the continuous motion characteristics of suspended tracks. Summary of the Invention

[0005] The purpose of this section is to outline some aspects of the embodiments of the present invention and to briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section, as well as in the abstract and title of the present application, to avoid obscuring the purpose of this section, the abstract and title of the invention. Such simplifications or omissions shall not be used to limit the scope of the present invention.

[0006] In view of the aforementioned existing problems, the present invention is proposed.

[0007] Therefore, the technical problem solved by this invention is: how to achieve accurate detection of hot spots, reasonable reconstruction of diffusion morphology, and adaptive spatial alignment of projected content under motion conditions.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0009] A method for thermal imaging processing of a suspended track projection target includes: synchronously acquiring a projection image and a corresponding thermal image during the movement of the suspended track; constructing a pixel mapping table based on the spatial mapping relationship between the projection image and the thermal image; and performing spatial alignment processing on the thermal image using the pixel mapping table to obtain initial registered thermal image data; dividing the initial registered thermal image data into a temperature change time series to form multiple time segments, and extracting a temperature rise change pattern encoding sequence from each time segment to obtain candidate hot spot region data; extracting hot spot center features and edge diffusion features from the candidate hot spot region data to form a feature pair; reconstructing the diffusion morphology of the hot spot boundary based on the spatial offset relationship between the feature pairs to obtain corrected hot spot region data; and generating a mask from the corrected hot spot region data and fusing it with the projection pattern to form a projection calibration result consistent with the hot spot diffusion morphology.

[0010] As a preferred embodiment of the present invention, the initial registration thermal image data includes: during the continuous operation of the suspended track according to predetermined motion parameters, synchronously acquiring, at a fixed frame rate, a projection image of the structured light or feature pattern projected by the projector and modulated by the object surface, and a corresponding thermal image; using the pre-calibrated relative pose parameters between the projector and the thermal imager, combined with the coordinates of known feature points in the projection image and the coordinates of corresponding feature points in the thermal image, establishing an initial correspondence set from the two-dimensional coordinates of the thermal image to the two-dimensional coordinates of the projection image; based on the initial correspondence set, dividing the thermal image and the projection image into multiple local sub-regions, calculating the local mapping relationship in each sub-region using spatial interpolation based on the distribution of feature points, and fusing all local mapping relationships according to the continuity and smoothness requirements of the boundaries of adjacent regions to form a pixel mapping table for each frame acquisition time; spatially resampling the same frame of thermal image according to the pixel mapping table to obtain initial registration thermal image data with the same size as the projection image and located in the projection space coordinate system.

[0011] As a preferred embodiment of the present invention, the process of obtaining candidate hot spot region data includes: organizing the initial registered thermal image data obtained from each frame into a temperature value sequence for each pixel position according to the acquisition time sequence; dividing the entire time axis into multiple continuous time segments according to a preset longest time interval and a fixed number of consecutive frames; calculating the rate of temperature change and the trend of the rate change for each pixel position within each time segment; generating a temperature rise change pattern encoding sequence for the corresponding pixel within the corresponding time period based on preset heating rate thresholds, cooling rate thresholds, and the direction of rate change; identifying pixels in the pattern encoding string that continuously exhibit accelerated heating or uniform heating for a duration exceeding a preset number of frames as abnormal heating pixels; performing connected component analysis on all abnormal heating pixels, removing isolated noise regions with an area smaller than a preset minimum hot spot area, and outputting candidate hot spot region data located in the projection space coordinate system.

[0012] As a preferred embodiment of the present invention, the generation of the corrected hot spot region data includes: for each region in the candidate hot spot regions, finding the pixel with the highest temperature in the region in the initial registered thermal image data as the hot spot center; expanding the region boundary outwards by a preset neighborhood range, extracting the boundary neighbor pixels located outside the region within the neighborhood range, calculating the temperature difference between each boundary neighbor pixel and the average temperature inside the region, and the temperature gradient direction at the boundary neighbor pixel, to form an edge diffusion feature set; forming a feature pair with the hot spot center and the edge diffusion feature set, calculating the unit direction vector from the hot spot center to each boundary neighbor pixel, comparing it with the heat diffusion direction unit vector at the boundary neighbor pixel, and determining whether the boundary point diffuses outwards; statistically analyzing the diffusion directions of all boundary neighbor pixels, taking the direction with the highest frequency as the main trend direction of heat diffusion; starting from the hot spot center, iteratively adjusting the region boundary along the main trend direction and the vertical direction until the number of boundary pixels whose state changes in two consecutive iterations is less than a preset tolerance threshold, thereby obtaining the corrected hot spot region data located in the projection space coordinate system.

[0013] As a preferred embodiment of the present invention, the process of forming a projection calibration result consistent with the hot spot diffusion morphology includes: using the corrected hot spot region data as the target calibration position in the projection space coordinate system, extracting the vertex coordinate set of the smallest circumscribed polygon or contour of the corrected hot spot region; generating a corresponding binary mask image from the vertex coordinate set according to the output resolution of the projector, wherein the region corresponding to the hot spot is marked as the effective value, and the remaining regions are marked as the background value; performing a morphological closing operation on the mask image to remove contour noise; fusing the smoothed mask image with the original projection pattern, adjusting the projection content according to the projection calibration target within the effective region corresponding to the hot spot, forming a projection calibration result consistent with the hot spot diffusion morphology, and outputting the projection calibration result to the projector for display.

[0014] In a preferred embodiment of the present invention, when establishing the initial pixel-level correspondence set from the two-dimensional coordinates of the thermal imaging image to the two-dimensional coordinates of the projected image, the mapping of the three-dimensional spatial coordinates of the object surface is not performed; the correspondence is established only using the two-dimensional position information of the known feature point pairs that are commonly visible in the projected image and the thermal imaging image.

[0015] As a preferred embodiment of the present invention, the specific method of iteratively adjusting the boundary of the region is as follows: if the heat diffusion direction at the boundary point points outward and the angle between the boundary point and the outward normal is less than a preset angle threshold, then the boundary point is expanded outward by one pixel; if the heat diffusion direction points inward, then the boundary point is contracted inward by one pixel; the above expansion and contraction process is repeated until the stopping condition is met.

[0016] As a preferred embodiment of the present invention, the temperature rise change mode encoding sequence includes at least the following encoding types: accelerated heating, uniform heating, decelerated heating, stable temperature, and cooling.

[0017] As a preferred embodiment of the present invention, the pixel mapping table is constructed once during the offline calibration stage, and can be finely adjusted frame by frame during the online monitoring stage based on optical flow or Kalman filtering, but the basic mapping relationship remains unchanged, so that each frame of thermal imaging image is spatially aligned using the pixel mapping table corresponding to its respective time during the continuous movement of the suspended track.

[0018] On the other hand, the present invention also provides a suspended track projection target thermal imaging processing system, comprising:

[0019] The mapping and alignment module synchronously acquires the projected image and the corresponding thermal image during the movement of the suspended track. It constructs a pixel mapping table based on the spatial mapping relationship between the projected image and the thermal image, and uses the pixel mapping table to perform spatial alignment processing on the thermal image to obtain the initial registered thermal image data.

[0020] The temperature rise identification module divides the initial registered thermal image data into temperature change time series to form multiple time segments, and extracts the temperature rise change pattern encoding sequence from each time segment to obtain candidate hot spot region data.

[0021] The diffusion reconstruction module extracts hot spot center features and edge diffusion features from the candidate hot spot region data to form a feature pair. Based on the spatial offset relationship between the feature pairs, it reconstructs the diffusion morphology of the hot spot boundary to obtain the corrected hot spot region data.

[0022] The mask fusion module generates a mask from the corrected hot spot region data and fuses it with the projection pattern to form a projection calibration result consistent with the hot spot diffusion pattern.

[0023] The beneficial effects of this invention are as follows: Compared with the prior art, the technical effects of this invention are as follows: This invention effectively overcomes the problem of dynamic drift in spatial mapping caused by the continuous movement of the suspended track by using a pixel mapping strategy that combines offline calibration and online optical flow fine-tuning, and significantly improves the frame-by-frame alignment accuracy between thermal imaging images and projection patterns; at the same time, this invention can accurately identify the hot spot from complex background temperature rise interference, and further use the heat diffusion direction to iteratively correct the hot spot boundary, so that the reconstructed hot spot morphology is more in line with the real heat conduction law, thereby generating a projection calibration result that is highly consistent with the hot spot diffusion trend.

[0024] Furthermore, the processing flow of this invention fully meets the requirements for real-time operation, and realizes automatic target reporting and visual feedback of moving targets without relying on external structured light measurement, which greatly improves the hit recognition rate, calibration stability and scene adaptability of the target reporting system. Attached Figure Description

[0025] Figure 1 This is a flowchart of the suspended track projection target thermal imaging processing method described in this invention.

[0026] Figure 2 This is a structural diagram of the suspended track projection target thermal imaging processing system described in this invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this invention.

[0028] like Figure 1As shown, the suspended track projection target thermal imaging processing method of the present invention is used in shooting training scenarios: the target (such as a metal target plate) is suspended below the track. After the shooter fires, the bullet impact area generates heat due to friction, forming a local hot spot. The system collects the target surface temperature distribution in real time through a thermal imager, processes the data, and projects the impact position and highlighting results back to the target surface through a projector, realizing visualized automatic target reporting.

[0029] The execution sequence of this method is as follows:

[0030] During the warm-up phase, the suspended track is stationary, and the system performs offline calibration to obtain a pixel space mapping table. During the motion monitoring phase, the suspended track moves at a constant speed of 0.2 m / s, the projector continuously projects conventional navigation patterns (non-structured light), the thermal imager continuously acquires thermal imaging images at 30 fps, and the visible light camera synchronously acquires projected images at 30 fps (optional, used to verify mapping stability).

[0031] For each frame of thermal imaging, spatial alignment is performed using an offline calibrated mapping table to obtain initial registered thermal image data. After accumulating N consecutive frames (e.g., 15 frames) of initial registered thermal image data, temperature time series analysis (S2) is performed (the analysis window sliding step size is 5 frames). S3 and S4 are only executed when a new candidate hotspot region is detected, and only frame-by-frame within 30 frames after a hit (approximately 1 second), then stopping until a new candidate hotspot appears in the next hit. If candidate hotspots exist in multiple consecutive frames (multiple hits or persistent hotspots), then S3 and S4 are executed for each frame.

[0032] In this embodiment of the invention, the specific operations are as follows:

[0033] S1: During the movement of the suspended track, the projected image and the corresponding thermal image are acquired synchronously. A pixel mapping table is constructed based on the spatial mapping relationship between the projected image and the thermal image. The thermal image is then spatially aligned using the pixel mapping table to obtain the initial registered thermal image data.

[0034] In this embodiment, the suspended track moves at a constant speed of 0.2 m / s. A projector and a thermal imager are mounted below the track, both fixed by a rigid bracket. Their relative poses are calibrated at the factory (relative rotation matrix and translation vector). The projector has a resolution of 1280×720, and the thermal imager has a resolution of 640×512, with a frame rate of 30 frames per second for both. The object being measured is a target (a metal plate, 320 mm × 256 mm in size, 8 mm thick, with a matte black paint coating to enhance thermal radiation absorption). Twenty-five circular reflective markers (5 mm in diameter) are pre-attached to the surface. These markers are highly reflective under visible light but invisible in thermal imaging due to their uniform temperature with the environment. The spatial positions of these markers are known in advance and are used to establish a mapping relationship between the target surface coordinate system and the projection coordinate system, enabling accurate target reporting coordinate transformation of the hit point.

[0035] To address the issue of thermal imagers being unable to directly identify visible light markers, a high thermal emissivity black paint (emissivity 0.95) is applied to the marker, and a heating patch is used to uniformly heat the marker to 5°C above the ambient temperature (maintained through closed-loop control), making it appear as a prominent high-temperature bright spot in thermal imaging. For example, a PID temperature controller combined with a patch-type PT100 temperature sensor (attached to the back of the marker) can be used for closed-loop control, with the ambient temperature measured in real-time by a built-in independent temperature sensor. Those skilled in the art can also employ other equivalent constant-temperature control methods.

[0036] In this embodiment of the invention, the specific implementation of this step is as follows:

[0037] S1.1: This invention is divided into two independent stages:

[0038] During the offline calibration phase, with the track stationary or in a known pose, the projector sequentially projects Gray code + phase-shift structured light patterns, which are then captured by a visible light camera to reconstruct the 3D point cloud of the object's surface. Simultaneously, the thermal imager acquires thermal images at the same time. By establishing a two-dimensional correspondence through joint marker points, the transformation relationship between the projector coordinate system, the thermal imager coordinate system, and the 3D spatial coordinates is calculated in one go, generating a pixel mapping table from the thermal image to the projected image (unchanged under rigid installation).

[0039] During the online monitoring phase, as the suspended track moves, the projector continuously projects conventional background patterns (such as ring number zones, bullseye indicators, or reference images to be calibrated). The thermal imager acquires thermal images frame by frame, and spatial alignment is performed using the pixel mapping table obtained during the offline calibration phase.

[0040] S1.2: Since thermal imagers cannot directly perceive visible light structured light, this embodiment uses pre-made marker points as cross-modal features.

[0041] During the continuous movement of the suspended track, a frame-by-frame dynamic calibration strategy is adopted:

[0042] At each frame acquisition moment, multiple joint marker points (e.g., 9 dots distributed at the four corners and center of the target surface) are pre-set on the object surface. These marker points possess the following characteristics: high reflectivity or unique color under visible light, and bright spots with temperatures higher than ambient temperature under thermal imaging due to additional heating or a high thermal emissivity coating. The actual target surface coordinates of these marker points (e.g., millimeter coordinates with the target center as the origin) are pre-measured using laser ranging or a calibration plate. This is used to convert the hot spot pixel coordinates in the projection space into physical coordinates on the target surface, thereby calculating the hit count and deviation, and outputting the target reporting data.

[0043] In the projected image, the coordinates of the virtual marker pattern can be identified through image processing, and in the thermal imaging image, the coordinates of the marker point can be identified through temperature threshold segmentation. Thus, the two-dimensional pixel coordinate pairs of the same physical point in the two images are obtained, and an initial set of correspondences is established.

[0044] Optionally, if the current frame has fewer than 9 feature point pairs due to occlusion or excessively fast movement, Kalman filtering or optical flow is used to predict the pixel mapping table of the previous frame, and the measured data of the next frame is used for correction.

[0045] S1.3: Due to the continuous movement of the suspended track, the relative pose of the object and the camera changes slowly over time. This invention uses a combination of offline calibration and online fine-tuning to obtain the pixel mapping table for each frame.

[0046] Specifically, in the offline calibration phase: after system installation is complete or the track is stationary, an initial set of feature point pairs (e.g., 25 pairs) is obtained using the S1.2 method. The thermal imaging image is divided into 36 local sub-regions of 6×6 (each sub-region is approximately 107×85 pixels). For each sub-region, if the number of feature point pairs is greater than or equal to 4, bilinear interpolation (or barycentric coordinate interpolation) is used for local mapping; if the number of feature point pairs is less than 4, the mapping is obtained by inverse distance weighted extrapolation using the established mappings of adjacent sub-regions. Weighted fusion of each local mapping is performed to form a continuous mapping across the entire field of view. This traverses all 640×512 pixels to obtain a basic pixel mapping table, which is a three-dimensional array of size 640×512×2, with each element storing a pair of floating-point coordinates. , which represents the initial corresponding position of the thermal imaging pixel in the projection space.

[0047] During the online monitoring phase, for each frame of thermal imaging acquired during the movement of the suspended track, the entire mapping table is not reconstructed; instead, fine-tuning is performed on the basic mapping table. Specifically, optical flow (or Kalman filtering) is used to track the positions of several stable feature points from the previous frame in the current frame. The displacement deviations of these feature points are calculated, and a low-order spatial transformation (such as affine transformation or local weighted averaging) is applied to the basic mapping table to obtain the pixel mapping table corresponding to the current frame. .

[0048] For example, taking the optical flow method, in the previous frame of the thermal imaging image, several stable feature points are selected, such as corner points located in the non-hotspot region of the target surface or joint marker points in S1.2, and their mapped coordinates in the projection space are recorded; in the current frame of the thermal imaging image, the pixel displacement of these feature points is calculated using the Lucas-Kanade optical flow method. Due to the underlying mapping table The static mapping from thermal imaging coordinates to projected coordinates is given. The actual projected coordinates of the i-th feature point in the current frame are estimated as follows: ;in, Let be the pixel coordinates of the i-th feature point in the previous frame of the thermal imaging image. In the above calculation, This means that the input is any sub-pixel coordinate in the thermal image, and the output is the corresponding projection space coordinate (floating point value).

[0049] Calculate the average displacement of all feature points in the projection space. And the affine transformation parameter A (least square fitting).

[0050] The pixel mapping table for the current frame can be corrected using the following formula: ; in, The center coordinates (floating-point values) of the thermal image.

[0051] To reduce computation, feature points are re-detected and the mapping parameters of the current frame are updated only when the cumulative feature point tracking error exceeds 3 pixels.

[0052] In this way, each frame of thermal imaging image undergoes spatial alignment using its corresponding pixel mapping table. For simplicity, the following steps will use the same notation. This represents the pixel mapping table used for spatial alignment operations in the current frame after offline calibration and online fine-tuning. Unless otherwise specified, subsequent pixel mapping tables refer to the corrected mapping table for that frame.

[0053] S1.4: For the thermal imaging image of the current frame (Temperature value matrix, data type uint16, minimum unit of measurement 0.01℃, i.e., storing a value of 100 represents 1.00℃), using the pixel mapping table constructed in step S1.3. Perform spatial alignment.

[0054] It should be noted that the raw temperature values ​​output by the thermal imager are stored as integers with a minimum unit of measurement of 0.01℃. When calculating the heating rate, cooling rate, and gradient, this stored value should first be divided by 100 to convert it to floating-point Celsius before participating in all subsequent calculations. All preset thresholds (such as 0.8℃ / s) are based on Celsius units.

[0055] Specifically, create a blank image with the same dimensions as the projected image (1280×720). For thermal imaging images... Each integer pixel coordinate in ,according to Obtain the corresponding floating-point coordinates in the projection space. The temperature value of the thermal imaging pixel is assigned to the four nearest integer pixel positions in the projected image using bilinear interpolation.

[0056] After traversing all thermal imaging pixels, the average of multiple contribution values ​​at each integer pixel location in the projected image is taken (if there is no contribution, it is marked as invalid). The final result is the initial registered thermal image data located in the projection space coordinate system. The dimensions are 1280×720, and each pixel value is a temperature value (unit: 0.01℃).

[0057] The data is spatially aligned with the projected pattern of the projector, providing a coordinate-consistent input for subsequent time series analysis.

[0058] It should be noted that this invention avoids the technical problem of thermal imagers lacking depth information by directly establishing a two-dimensional to two-dimensional correspondence. Furthermore, it ensures the accuracy of spatial alignment during the movement of the suspended track by finely adjusting the pixel mapping table frame by frame based on offline calibration and using optical flow method for tracking and updating.

[0059] S2: Divide the initial registered thermal image data into temperature change time series to form multiple time segments, and extract the temperature rise change pattern encoding sequence from each time segment to obtain candidate hot spot region data.

[0060] S2.1: For each pixel position in the projection space The temperature values ​​collected from each frame are arranged in chronological order according to their timestamps to form a temperature sequence. ,in The acquisition time of the corresponding frame (unit: seconds, precision 0.001 seconds). These are the pixel coordinates in the projection space.

[0061] To prevent sequence breaks caused by data frame loss, a combination of sliding window and gap checking is used to divide the time into segments:

[0062] The system slides across the entire timeline with a fixed number of frames (15 frames) and a sliding step size (5 frames). Frames within each window are sorted by timestamp. If the time interval between two adjacent frames within a window is greater than 0.1 seconds (i.e., frame loss or communication interruption), the window is discarded and not used for temperature rise analysis. If the number of valid frames within a window is ≥12 (which can be set according to actual needs), the window is retained as a time segment. The overlapping portion (5 frames) between adjacent windows is allowed for smooth transition, but sequences spanning windows are not jointly analyzed.

[0063] The above operations avoid false segmentation caused by brief interruptions, while ensuring that the temperature changes within each segment are continuous in time.

[0064] S2.2: For each pixel within each time segment Extract the temperature values ​​of all frames within this segment, along with their corresponding times. The rate of temperature change over time was calculated using the central difference method. :

[0065] For internal frames ( =2,…,n-1, where n is the number of frames in the segment). ; For the first frame and the last frame, forward difference and backward difference are used for calculation, respectively.

[0066] Next, the trend of the rate of change (i.e., the second time derivative of temperature) is calculated: for the inner frame: ; The second derivatives of the first and last frames are set to 0.

[0067] In addition, to suppress noise, the temperature sequence can be smoothed by a 3-point moving average (window width of 3 frames) before the difference calculation is performed.

[0068] S2.3: For each pixel in each frame within each time segment, based on the calculated... and Generate pattern encoding.

[0069] Among them, the heating rate threshold The effective temperature rise rate threshold is 0.8℃ / s (based on live-fire tests, recording the temperature change curve of the center pixel in the bullet impact area, calculating the maximum temperature rise rate within 1 second after impact, and taking the lower quartile of multiple tests as the effective temperature rise rate threshold); the cooling rate threshold is... -0.5℃ / s (absolute value less than) (To avoid misjudging an anomaly due to natural cooling); small acceleration. It is 0.2℃ / s² (used for judgment) Whether it is close to 0), its typical value is 0.2°C / s², and the calibration method is as follows: when there is no hit on the target surface, 100 frames of static thermal images are continuously acquired, the absolute value of the second derivative of the temperature sequence of each pixel is calculated, and the 95th percentile is taken as .

[0070] Furthermore, in this embodiment, the encoding rule is as follows: like and The code is then set to accelerate heating. like and The code is then coded as uniform heating. like and The code is then set to slow down the temperature rise. like If so, it is encoded as temperature stable; like ( If the value is negative, then the code is cooling.

[0071] If multiple conditions are met simultaneously, they are matched in the order described above. After each time segment, each pixel receives an encoded sequence with a length equal to the number of frames in that segment (e.g., 12 to 15 codes).

[0072] S2.4: For each pixel, find the longest consecutive segment in the encoded sequence where accelerated or uniform heating occurs continuously. If the number of frames in this consecutive segment is greater than or equal to a preset duration threshold, sufficient to exclude single-frame noise, then the pixel is marked as an abnormal heating pixel. In this embodiment, the preset duration threshold is set to 5 frames, corresponding to a continuous heating time ≥ 0.167 seconds.

[0073] If a pixel in a segment is judged to be abnormal, its mark is recorded; subsequent segments are recalculated without cross-segment accumulation (to avoid misjudgment due to reheating after a brief interruption).

[0074] S2.5: Perform spatial connectivity analysis on all pixels marked as abnormally heated.

[0075] Specifically, an 8-neighbor connectivity rule is used (i.e., adjacent pixels in the top, bottom, left, right, and four diagonal directions are considered connected). Abnormally heated pixels that are connected together are merged into a single connected region. For each connected region, the total number of pixels it contains is calculated as the region area.

[0076] The preset minimum hotspot area is 25 pixels (corresponding to a physical size of approximately 2.5mm × 2.5mm, determined based on the thermal imager's spatial resolution of 0.5mm / pixel and the minimum detectable hotspot size). If the area of ​​a connected region is less than 25 pixels, it is judged as isolated noise or minor interference and is discarded; if the area is greater than or equal to 25 pixels, it is retained as a candidate hotspot region.

[0077] The final output is candidate hotspot region data. Each candidate region is represented by a set of pixel coordinates in the projected coordinate space. Simultaneously, to facilitate subsequent processing, the coordinates of the minimum bounding rectangle of each candidate region and the location of the highest temperature point within that region are recorded.

[0078] The above thresholds are all based on the given criteria or example values, and those skilled in the art can adjust them according to the actual scenario.

[0079] It should be noted that this invention solves the problem of discontinuous time series caused by frame loss or brief occlusion during motion by combining a fixed-frame-number sliding window with interruption segmentation. In a target reporting scenario, a bullet hitting the target surface causes the local temperature to rise sharply within 0.1 to 0.3 seconds, and then slowly decrease due to heat diffusion. The temperature rise change pattern encoding in the above operation is based on this physical characteristic, identifying pixels that heat up rapidly and maintain that temperature for a certain period of time as hit hotspots, avoiding misjudging environmental noise or slow temperature rises as hits.

[0080] S3: Extract hot spot center features and edge diffusion features from the candidate hot spot region data to form a feature pair. Based on the spatial offset relationship between the feature pairs, reconstruct the diffusion morphology of the hot spot boundary to obtain the corrected hot spot region data.

[0081] It should be noted that, after processing by S2, candidate hotspot region data located in the projected spatial coordinate system were obtained. Each region is represented by a set of continuous pixel coordinates. However, since the temperature rise pattern encoding is based solely on temporal thresholds, the boundaries of candidate regions often appear jagged or excessively contracted / expanded, failing to accurately reflect the true spatial morphology of the hot spot formed by heat diffusion (typically characterized by a high temperature at the center, gradually decreasing temperature towards the edges, with a continuous transition at the boundaries). Further utilization of the spatial distribution characteristics of the temperature field is needed. By extracting the diffusion features at the center and edges of the hot spot and iteratively adjusting the boundaries based on the temperature gradient direction, a corrected hot spot region consistent with the heat diffusion trend can be obtained.

[0082] To achieve the above objectives, step S3 is implemented as follows:

[0083] S3.1: For candidate hot spot regions For each connected region in the array, perform the following sub-steps: (1) Initial registration thermal image data corresponding to this area (Located in the projected coordinate system, unit 0.01℃), find the temperature value of all pixels within the region, and take the coordinate of the pixel with the highest temperature as the hotspot center c. If multiple pixels have the same temperature and are all at the highest, then take the geometric center of these pixels as the hotspot center.

[0084] (2) Preset boundary expansion range Set to 5 pixels (based on the thermal imager's spatial resolution of 0.5mm / pixel and typical half-width of thermal spread of 2.5mm; can be linearly increased if the hotspot size is larger). Expand the original boundary of the current area outwards. 1 pixel, to obtain the expanded region Extraction located at Pixels that are inside the original region but not inside it are considered as the boundary neighborhood pixel set B.

[0085] (3) For each boundary neighborhood pixel in set B Calculate the following two characteristic temperature differences and temperature gradient directions.

[0086] in, The temperature gradient is calculated by subtracting the average temperature of all pixels within the original region (excluding boundary pixels) from the temperature value of this pixel; temperature gradient direction. The horizontal gradient at this pixel is calculated using the Sobel operator. and vertical gradient Then, the gradient direction angle is obtained: ; ; in, The unit is radians, and the range is from -π to π. and These are the horizontal and vertical gradients calculated by the Sobel operator, respectively.

[0087] The gradient direction points in the direction of the fastest temperature increase.

[0088] The above calculation results for each boundary neighboring pixel are combined into a record, and all records constitute the edge diffusion feature set E.

[0089] S3.2: Combine the hot spot center c with the edge diffusion feature set E to form a feature pair.

[0090] Calculate the distance from the hotspot center c to the boundary pixel. unit direction vector Simultaneously calculate the temperature gradient at that pixel. The direction of heat transfer is opposite to the direction of the temperature gradient (heat flows from high temperature to low temperature), therefore the direction of heat diffusion is a unit vector. .

[0091] Calculate the dot product of two vectors: The judgment rules are as follows: like A correlation >0.2 (positive correlation) indicates that heat diffuses outward from the center, and this boundary point should be considered as pushing outward; if A correlation of less than -0.2 (negative correlation) indicates that the heat is converging inward, and the boundary point should be considered as contracting inward; otherwise, it is considered as having an unclear direction and does not change the boundary state.

[0092] It should be noted that the threshold of 0.2 corresponds to the cosine of the included angle of approximately 78 degrees, which is used to filter out uncertain gradients that are close to the vertical direction.

[0093] S3.3: Statistics of all Boundary neighbor pixels with a value >0.2, relative to the direction angle Perform histogram statistics.

[0094] Among them, the direction angle . and These are the boundary neighbor pixels. and coordinate; and The centers of the hot spots and coordinate; Let be the angle between the vector pointing from the center of the hot spot c to the neighboring pixel p on the boundary and the positive direction of the horizontal axis.

[0095] The 360° angle was divided into 12 intervals (each interval being 30°), and the number of pixels satisfying the outward diffusion condition within each interval was counted. The center angle of the interval with the highest number of pixels was taken as the main trend direction of heat diffusion. (Unit: degrees, range 0° to 360°). Also record its orthogonal direction. (If it exceeds 360°, subtract 360°), used for anisotropic expansion.

[0096] If not satisfied For pixels >0.2, the main trend direction is set from the geometric center of the region to the second highest temperature point (excluding the center).

[0097] S3.4: Using the boundary pixel set of the current region as the processing object, iterate using an expansion / contraction rule based on the temperature gradient direction. Define the current region as... ,initial This represents the original candidate region. The iteration process is as follows:

[0098] (1) Identify all boundary pixels of the current region.

[0099] Wherein, boundary pixels are defined as: belonging to And at least one pixel in its 8-neighborhood does not belong to Pixels.

[0100] (2) For each boundary pixel Calculate the following data: Outward normal unit vector For the current region Generate the corresponding binary mask image. (A pixel value of 1 indicates the area within the region, and 0 indicates the area outside the region). For Perform a Euclidean distance transformation to obtain the distance field. , whose value is the Euclidean distance from each pixel to the nearest background pixel. In the region boundary pixels... Gradient of the field at that location Pointing inwards and perpendicular to the boundary. Therefore, the outward normal unit vector: ; The gradient is approximated by the Sobel operator during calculation.

[0101] Furthermore, calculate the cosine of the angle between the normal and the gradient direction: ; (3) Preset angle threshold The angle is set at 60° (corresponding to a cosine value of 0.5). This preset angle threshold is based on the isotropic assumption of thermal diffusion: in an ideal homogeneous medium, heat diffuses radially outward from the center of the hot spot, with the temperature gradient direction essentially aligned with the boundary normal, typically deviating by no more than 30°. Considering the non-uniformity of the actual target material and the influence of noise, the allowable deviation is increased to 60°. This threshold can be calibrated through thermal conduction simulation or measured temperature gradient distribution at the hot spot edge, with a typical range of 45° to 75°.

[0102] Specifically, the adjustment rules are as follows: like >0.5 (i.e., the included angle is <60°) and If the boundary point points outward (i.e., it can enter the outer region by moving along the gradient direction), then the boundary point is expanded outward: Add the outer neighbor pixel pointed to by the point (the nearest pixel along the gradient direction that belongs to the outside). ;like <-0.5 (i.e., the included angle is >120°) and Pointing inwards, then Point from Delete (inward contraction); otherwise keep. The point state remains unchanged.

[0103] The expansion and contraction operations described above should be performed in batches based on the region boundaries at the start of the current iteration. That is, the proposed update states of all boundary points should be calculated first, and then updated uniformly. Avoid sequential dependencies.

[0104] S3.5: After each complete boundary scan and update (called an iteration), calculate the number of boundary pixels that have undergone state changes in this iteration. (Including pixels added during expansion and pixels removed during contraction). Preset tolerance threshold. The value is 10 pixels, and it is dynamically adjusted based on the area of ​​the region: if the area is less than 200 pixels, then... =max(5, 0.05 × area). In two consecutive iterations, All less than When the boundary is considered to have converged, the iteration stops.

[0105] If the number of iterations exceeds the maximum allowed number of iterations of 30, the loop will be forcibly stopped to avoid infinite loops.

[0106] Area at the time of stopping That is, to correct the hot spot area. The region remains within the projected spatial coordinate system, and its boundary follows the heat diffusion trend, exhibiting a continuous and smooth shape.

[0107] S3.6: The result obtained after iterative adjustment of each candidate hotspot region as described above. The output is used as corrected hot spot region data. If there are multiple original candidate regions, the corresponding corrected regions are output separately. Each corrected region is represented by a set of pixel coordinates, and its hot spot center coordinates are also recorded (the updated center may shift with boundary adjustments, requiring recalculation of the highest temperature pixel within the region).

[0108] As can be seen, this invention determines the boundary adjustment by using the angle between the temperature gradient direction and the boundary normal, directly integrating the physical laws of heat diffusion. This avoids the use of complex level set models, making the calculation simple and controllable. Furthermore, the stopping condition that the change amount is less than the tolerance threshold twice in a row ensures convergence and avoids infinite loops.

[0109] S4: Generate a mask from the corrected hot spot region data and fuse it with the projection pattern to form a projection calibration result consistent with the hot spot diffusion pattern.

[0110] S4.1: For correcting hot spot regions For each connected region in the dataset, extract its minimum bounding polygon or convex hull contour. Specific operations include:

[0111] If the number of pixels in the region is less than 500, the Graham scan algorithm is used to calculate the convex hull vertices. All pixels within the region are traversed, and the bottom-left corner is found as the base point. The pixels are then sorted by polar angle and scanned sequentially, removing points that would cause a right turn. Finally, the set of convex hull vertex coordinates is obtained. If the number of pixels in the region is large (≥500) and the shape is irregular, the minimum bounding rectangle is used as a substitute (the minimum rotated rectangle of the covered area is calculated), and the four corner points of the rectangle are used to form the boundary. If the frame contains multiple correction regions (e.g., multiple hot spots), then extract the contour for each region separately, and merge the vertex sets of all regions to form a multi-contour vertex set. .

[0112] S4.2: Based on the output resolution of the projector (1280×720 pixels in this embodiment), create a blank binary image with the same resolution and data type uint8, and initially set all pixel values ​​to 0 (background value).

[0113] for For each polygonal contour in the algorithm, a scan-line polygon filling algorithm is used to fill the pixels inside the closed contour with 1 (a valid value). The specific algorithm is as follows:

[0114] For each polygon, calculate the range of its ordinate. For each integer y-coordinate Calculate the x-coordinates of the intersection points of the polygon and the horizontal line when the polygon crosses the horizontal line (considering the parity rule of vertices), and obtain a series of intersection point intervals. .

[0115] Assign a value of 1 to all pixels in the blank binary image that satisfy (x,y) falling within any interval.

[0116] If multiple contours overlap, the overlapping area is still assigned a value of 1. After filling, a binary mask image is obtained. The region with a value of 1 precisely corresponds to the position and shape of the hot spot in the projection space.

[0117] S4.3: Because the boundaries of the corrected hotspot region may still have tiny jagged edges or isolated points after iterative adjustments, the boundaries of the directly generated mask image are not smooth enough, affecting the projection display quality. Therefore, [the following is addressed / addressed / implemented]: Perform morphological closing operations (dilation followed by erosion):

[0118] The dilation operation uses a 3×3 cross-shaped structural element (i.e., the center pixel and its four neighbors above, below, left, and right). Traversal... For each pixel, if at least one pixel within the area covered by the structuring element has a value of 1, then the corresponding position in the output image is set to 1; otherwise, it is set to 0. This operation can fill in small holes inside hotspot areas and connect small protrusions.

[0119] It should be noted that the use of isotropic cross-shaped structural elements can avoid boundary noise introduced by diagonal neighbors. Furthermore, for near-convex regions such as hot spots, the cross-shaped closing operation (expansion followed by erosion) can effectively eliminate boundary jaggedness and isolated holes while maintaining the area of ​​the region essentially unchanged.

[0120] The closing operation uses a 3×3 cross-shaped structuring element (center and its four adjacent neighbors above, below, left, and right) by default, repeated twice. If the ratio of the perimeter to the area (shape factor) of the hotspot region is greater than 0.6 (elongated shape), a 3×3 fully connected structuring element (including diagonal neighbors) is used instead. If the hotspot area is less than 100 pixels, the closing operation is repeated once. Regions with an area less than 25 pixels do not undergo morphological smoothing (the original mask is output directly). The above thresholds can be adjusted according to the actual image resolution; typical values ​​are given in this embodiment.

[0121] The erosion operation uses the same structuring element. The corresponding position in the output image is set to 1 only if all pixels covered by the structuring element are 1; otherwise, it is set to 0. This operation can recover slightly expanded boundaries after dilation and further smooth jagged edges.

[0122] The closing operation is repeated twice to obtain a smoother boundary. For small hotspots (e.g., less than 100 pixels), the number of iterations can be reduced or a smaller structuring element (e.g., a circle with a radius of 2 pixels) can be used. The final result is a smooth mask. .

[0123] Experimental tests show that two closing operations can eliminate more than 90% of single-pixel noise and isolated holes. Increasing the number of operations further will lead to excessive shrinkage of the hot spot area.

[0124] S4.4: Apply the smoothed mask It is blended with the current projected background pattern (i.e., the image that the projector originally planned to display, with a data type of RGB three channels, each channel ranging from 0 to 255).

[0125] The fusion process only occurs within the area where the mask value is 1, while the background area retains its original projection pattern.

[0126] In this embodiment of the invention, multiple fusion modes are preset, which can be selected according to the actual application scenario:

[0127] Edge Enhancement Mode: This mode extracts only the boundaries of the hotspot region (using Canny edge detection or directly taking the gradient of the mask) and overlays a highlight color (e.g., red, RGB value (255,0,0)) at the boundary locations. The region within the boundary remains uncolored. This mode is suitable for detection scenarios requiring precise hotspot boundary localization.

[0128] Color Overwrite Mode: Replaces the RGB values ​​of all pixels within the mask area with a semi-transparent pseudo-color (e.g., orange, with a blending factor of 0.6, meaning the output is 0.6 × pseudo-color + 0.4 × original pattern). This mode is suitable for monitoring scenarios that require rapid identification of hotspot areas.

[0129] Calibration overlay mode: A crosshair (30×30 pixels, 2 pixels wide, cyan color) is overlaid at the geometric center of the mask (obtained by calculating the average coordinates of all pixels within the mask), and the highest temperature value of the currently detected hotspot (obtained from the initial registration thermal image data in the first step) is overlaid nearby. This mode is suitable for interactive operation or situations requiring precise temperature readings.

[0130] For the automatic target reporting system, a target reporting data display mode can be added, which overlays the current hit number (e.g., 9 rings), the distance from the bullseye (e.g., 2.3 cm to the left), and the hit sequence number near the mask area (e.g., 10 pixels above the center of the hot spot) for real-time viewing.

[0131] The selection can be made according to actual needs to obtain the final projection calibration result image. The final projection calibration result is directly transmitted to the projector and displayed by replacing or overlaying the original background pattern.

[0132] During actual motion, steps S1 to S4 are repeated for each frame, updating the projection calibration results frame by frame. The entire processing flow is implemented on an embedded industrial computer, with a measured processing time of approximately 25 milliseconds per frame (less than the frame interval of 33 milliseconds), meeting real-time requirements.

[0133] It should be noted that this invention, through a frame-by-frame update mechanism combined with an offline calibration and frame-by-frame fine-tuning pixel mapping table, ensures that the projected calibration results remain synchronized with the actual hot spot position during the movement of the suspended track, thus solving the calibration offset problem caused by motion. In automatic target reporting applications, the system can also record and output the following data corresponding to each hit to the display terminal or training database: hit time, hot spot center target surface coordinates (converted to ring number and offset), maximum temperature, and hot spot area. This data can be used for shooting performance statistics and training analysis.

[0134] like Figure 2As shown, the present invention also includes a suspended track projection target thermal imaging processing system, comprising: a projector, a thermal imager, a visible light camera, and a suspended track; wherein, the visible light camera and the thermal imager are rigidly fixed and used to acquire the projected image of the object surface; the thermal imager is used to acquire thermal radiation images at the same time, and further includes:

[0135] The mapping and alignment module synchronously acquires the projected image and the corresponding thermal image during the movement of the suspended track. It constructs a pixel mapping table based on the spatial mapping relationship between the projected image and the thermal image, and uses the pixel mapping table to perform spatial alignment processing on the thermal image to obtain the initial registered thermal image data.

[0136] The temperature rise identification module divides the initial registered thermal image data into temperature change time series to form multiple time segments, and extracts the temperature rise change pattern encoding sequence from each time segment to obtain candidate hot spot region data.

[0137] The diffusion reconstruction module extracts hot spot center features and edge diffusion features from the candidate hot spot region data to form a feature pair. Based on the spatial offset relationship between the feature pairs, it reconstructs the diffusion morphology of the hot spot boundary to obtain the corrected hot spot region data.

[0138] The mask fusion module generates a mask from the corrected hot spot region data and fuses it with the projection pattern to form a projection calibration result consistent with the hot spot diffusion pattern.

[0139] The system also includes one or more processors and memory.

[0140] The memory is used to store operable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations, including the flow of the suspended track projection target thermal imaging processing method of the foregoing embodiments, in particular... Figure 1 The flowchart of the system is shown.

[0141] Other aspects disclosed in the embodiments of the present invention also propose a computer-readable medium for storing software including instructions executable by one or more computers, which, upon execution, cause the one or more computers to perform operations including the flow of the suspended track projection target thermal imaging processing method of the foregoing embodiments, particularly... Figure 1 The flowchart of the system is shown.

[0142] It should be recognized that embodiments of the present invention may be implemented or carried out by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable storage medium.

[0143] The system can be implemented using standard programming techniques, including a non-transitory computer-readable storage medium configured with computer programs, wherein such a storage medium enables the computer to operate in a specific and predefined manner.

[0144] Each program can be implemented in a high-level procedural or object-oriented programming language to communicate with the computer system; however, if required, the program can be implemented in assembly or machine language.

[0145] In any case, the language can be either compiled or interpreted.

[0146] Furthermore, for this purpose, the program can run on programmed application-specific integrated circuits.

[0147] The processes described herein (or variations and / or combinations thereof) can be executed under the control of one or more computer systems configured with executable instructions, and can be implemented by hardware or a combination thereof as code (e.g., executable instructions, one or more computer programs, or one or more applications) that commonly executes on one or more processors. The computer program includes a plurality of instructions executable by one or more processors.

[0148] Furthermore, the system can be operatively connected to any suitable type of computing platform, including but not limited to personal computers, minicomputers, mainframes, workstations, networked or distributed computing environments, standalone or integrated computer platforms, or communicate with charged particle tools or other imaging devices.

[0149] Various aspects of the present invention can be implemented in machine-readable code stored on a non-transitory storage medium or device, whether portable or integrated into a computing platform, such as a hard disk, optical read and / or write storage medium, RAM, ROM, etc., such that it can be read by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the processes described herein.

[0150] Furthermore, machine-readable code, or parts thereof, can be transmitted via wired or wireless networks.

[0151] When such media includes instructions or programs that combine with a microprocessor or other data processor to implement the steps described above, the invention described herein includes these and other different types of non-transitory computer-readable storage media.

[0152] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for thermal imaging processing of a suspended track projection target, characterized in that, include: During the movement of the suspended track, projection images and corresponding thermal imaging images are acquired simultaneously. A pixel mapping table is constructed based on the spatial mapping relationship between the projection images and thermal imaging images, and the thermal imaging images are spatially aligned using the pixel mapping table to obtain the initial registered thermal image data. The initial registered thermal image data is divided into temperature change time series to form multiple time segments, and the temperature rise change pattern encoding sequence is extracted from each time segment to obtain candidate hot spot region data. In the candidate hot spot region data, hot spot center features and edge diffusion features are extracted to form a feature pair. Based on the spatial offset relationship between the feature pairs, the diffusion morphology of the hot spot boundary is reconstructed to obtain the corrected hot spot region data. The corrected hot spot region data is used to generate a mask and fused with the projection pattern to form a projection calibration result consistent with the hot spot diffusion pattern.

2. The suspended track projection target thermal imaging processing method according to claim 1, characterized in that, The obtained initial registration thermal image data includes: During the continuous operation of the suspended track according to the predetermined motion parameters, the projected image of the structured light or feature pattern projected by the projector is modulated by the object surface and the corresponding thermal imaging image is acquired synchronously at a fixed frame rate. By using the pre-calibrated relative pose parameters between the projector and the thermal imager, and combining the coordinates of known feature points in the projected image with the coordinates of corresponding feature points in the thermal image, an initial set of correspondences from the two-dimensional coordinates of the thermal image to the two-dimensional coordinates of the projected image is established. Based on the initial correspondence set, the thermal imaging image and the projection image are divided into multiple local sub-regions. In each sub-region, the local mapping relationship is calculated by spatial interpolation based on the distribution of feature points. All local mapping relationships are then fused according to the continuity and smoothness requirements of the boundaries of adjacent regions to form a pixel mapping table for each frame acquisition time. Based on the pixel mapping table, the same frame of thermal imaging image is spatially resampled to obtain initial registered thermal image data with the same size as the projected image and located in the projection space coordinate system.

3. The suspended track projection target thermal imaging processing method according to claim 1, characterized in that, The obtained candidate hot spot region data includes: The initial registered thermal image data obtained from each frame are organized into a sequence of temperature values ​​at each pixel location according to the chronological order of acquisition time. Based on the preset longest time interval and a fixed number of consecutive frames, the entire timeline is divided into multiple consecutive time segments. Within each time segment, calculate the rate of temperature change over time and the trend of that rate for each pixel location; Based on preset heating rate thresholds, cooling rate thresholds, and the direction of rate change, a temperature rise change pattern encoding sequence for the corresponding pixel within the corresponding time period is generated. Pixels in the pattern encoding string that continuously exhibit accelerated or uniform heating for a duration exceeding a preset frame number threshold are identified as abnormal heating pixels. Perform connected component analysis on all abnormally heated pixels, remove isolated noise regions with an area smaller than the preset minimum hot spot area, and output candidate hot spot region data located in the projection space coordinate system.

4. The suspended track projection target thermal imaging processing method according to claim 1, characterized in that, The generation of the corrected hot spot region data includes: For each region in the candidate hot spot region, the pixel with the highest temperature in the initial registered thermal image data is selected as the hot spot center. The region boundary is extended outward to a preset neighborhood range. The boundary neighborhood pixels located outside the region within the neighborhood range are extracted. The temperature difference between each boundary neighborhood pixel and the average temperature inside the region, as well as the temperature gradient direction at the boundary neighborhood pixel, are calculated to form an edge diffusion feature set. The hot spot center and the edge diffusion feature set are combined into a feature pair. The unit direction vector from the hot spot center to each boundary neighboring pixel is calculated and compared with the heat diffusion direction unit vector at the boundary neighboring pixel to determine whether the boundary point diffuses outward. The diffusion directions of all boundary neighboring pixels are counted, and the direction with the highest frequency is taken as the main trend direction of heat diffusion. Starting from the center of the hot spot, the region boundary is iteratively adjusted along the main trend direction and the vertical direction until the number of boundary pixels that have undergone state changes in two consecutive iterations is less than the preset tolerance threshold, thus obtaining the corrected hot spot region data in the projection space coordinate system.

5. The suspended track projection target thermal imaging processing method according to claim 1, characterized in that, The projection calibration results that form a pattern consistent with the hot spot diffusion include: Using the corrected hotspot region data as the target calibration position in the projection space coordinate system, the vertex coordinate set of the smallest bounding polygon or contour of the corrected hotspot region is extracted. Based on the output resolution of the projector, a corresponding binary mask image is generated from the set of vertex coordinates, wherein the area corresponding to the hot spot is marked as the valid value, and the remaining area is marked as the background value; Perform morphological closing operations on the mask image to remove contour noise; The smoothed mask image is fused with the original projection pattern. The projection content is adjusted according to the projection calibration target in the effective area corresponding to the hot spot to form a projection calibration result consistent with the hot spot diffusion pattern. The projection calibration result is then output to the projector for display.

6. The method for thermal imaging processing of suspended track projection targets according to claim 2, characterized in that, When establishing the initial pixel-level correspondence set from the two-dimensional coordinates of the thermal imaging image to the two-dimensional coordinates of the projected image, the mapping of the three-dimensional spatial coordinates of the object surface is not performed. Instead, the correspondence is established using only the two-dimensional position information of the known feature point pairs that are commonly visible in the projected image and the thermal imaging image.

7. The suspended track projection target thermal imaging processing method according to claim 4, characterized in that, The specific method for iteratively adjusting the boundary of the region is as follows: If the heat diffusion direction at the boundary point points outward and the angle between the boundary point and the outward normal is less than a preset angle threshold, then the boundary point is expanded outward by one pixel; if the heat diffusion direction points inward, then the boundary point is contracted inward by one pixel. Repeat the expansion and contraction process as described above until the number of boundary pixels that undergo state changes in two consecutive iterations is less than the preset tolerance threshold.

8. The method for thermal imaging processing of suspended track projection targets according to claim 3, characterized in that, The temperature rise change pattern coding sequence includes at least the following coding types: accelerated heating, uniform heating, decelerated heating, stable temperature, and cooling.

9. The method for thermal imaging processing of suspended track projection targets according to claim 2, characterized in that, The pixel mapping table is constructed once during the offline calibration stage, and can be finely adjusted frame by frame during the online monitoring stage based on optical flow or Kalman filtering, but the basic mapping relationship remains unchanged, so that each frame of thermal imaging image is spatially aligned using the pixel mapping table corresponding to its time during the continuous movement of the suspended track.

10. A thermal imaging processing system for suspended track projection targets, based on the thermal imaging processing method for suspended track projection targets according to any one of claims 1 to 9, characterized in that: Also includes: The mapping and alignment module synchronously acquires the projected image and the corresponding thermal image during the movement of the suspended track. It constructs a pixel mapping table based on the spatial mapping relationship between the projected image and the thermal image, and uses the pixel mapping table to perform spatial alignment processing on the thermal image to obtain the initial registered thermal image data. The temperature rise identification module divides the initial registered thermal image data into temperature change time series to form multiple time segments, and extracts the temperature rise change pattern encoding sequence from each time segment to obtain candidate hot spot region data. The diffusion reconstruction module extracts hot spot center features and edge diffusion features from the candidate hot spot region data to form a feature pair. Based on the spatial offset relationship between the feature pairs, it reconstructs the diffusion morphology of the hot spot boundary to obtain the corrected hot spot region data. The mask fusion module generates a mask from the corrected hot spot region data and fuses it with the projection pattern to form a projection calibration result consistent with the hot spot diffusion pattern.

Images