Dynamic tracking method, system and equipment for explosion fragments and storage medium

Abstract

The invention provides a dynamic tracking method, system and device for an explosion fragment and a storage medium, and belongs to the field of optical tests.The method comprises the steps that a Bouguet epipolar correction algorithm is used for processing a binocular high-speed image, and matching search is reduced from two dimensions to one dimension; thirdly, providing a Rank-Census mixed binocular matching algorithm, and realizing high-precision matching and three-dimensional reconstruction of fragments with the same name in a complex environment by fusing the statistical robustness of Rank transformation and the texture sensitivity of Census transformation; secondly, a fragment diffusion physical model is introduced, the explosion center distance and the diffusion angle range of fragments are defined according to the charging structure, and a candidate space area of the fragments at the next moment is predicted; and finally, in the prediction area, adopting a speed consistency screening strategy, calculating an arithmetic mean value of the instantaneous speed of the candidate fragments, and selecting the fragments with the speed closest to the mean value for track association, thereby realizing stable and continuous tracking of the target fragments.

Description

Technical Field

[0001] This invention belongs to the field of optical testing, and specifically relates to a method, system, device and storage medium for dynamic tracking of explosive fragments. Background Technology

[0002] Explosive fragments are the core carriers of warhead damage effects. Their spatial distribution, trajectory, and velocity attenuation characteristics are of great significance for weapon power assessment, protective structure design, and target damage mechanism research. Achieving accurate, stable, and full-field dynamic tracking and measurement of explosive fragments is a key technological requirement in this field.

[0003] Currently, high-speed imaging-based methods for tracking explosive fragments rely on stereo vision-based matching and reconstruction. This method recovers the three-dimensional position of the fragments by matching features from the left and right images, thus tracking their trajectory. However, stereo vision-based matching and reconstruction methods face two significant bottlenecks when applied to explosive fragment tracking: First, the fragment targets are typically small, have weak texture, and similar shapes, resulting in low feature discrimination under complex interference, leading to poor matching accuracy and a high mismatch rate. Second, finding corresponding points in the left and right images requires searching in two-dimensional image space, which is computationally complex and difficult to meet the real-time requirements of high-speed fragment tracking. Therefore, achieving both high efficiency and high accuracy in binocular stereo matching within high-interference, high-density explosive fragment fields has become a key challenge restricting explosive fragment tracking. Summary of the Invention

[0004] To address the aforementioned problems, this invention provides a method, system, device, and storage medium for dynamic tracking of explosive fragments.

[0005] To achieve the above objectives, the present invention provides a dynamic tracking method for explosive fragments, comprising: High-speed images of the explosion field are acquired simultaneously by binocular cameras at continuous moments to form a binocular image sequence; the binocular image sequence includes a left view and a right view.

[0006] Based on the current moment, a target fragment is selected from the left view of the binocular image sequence. The feature points of the target fragment are extracted as local neighborhood windows. Rank transformation and Census transformation are performed on the local neighborhood windows respectively, and the two transformation results are fused to obtain the first composite feature. From the right view of the binocular image sequence, all fragments on the same scan line as the local neighborhood windows of the left view are selected. Local neighborhood windows of each fragment are selected, and Rank transformation and Census transformation are performed on each local neighborhood window respectively to obtain multiple second composite features. The difference value between the first composite feature and each second composite feature is calculated. The fragment corresponding to the smallest difference value is selected from all the difference values ​​to form a fragment pair with the target fragment.

[0007] Based on the pixel coordinates of the same fragment pair in the binocular image sequence, the spatial coordinates of the target fragment at the current moment are determined; based on the spatial coordinates of the target fragment at the current moment and the preset fragment motion constraint range, the candidate spatial region of the target fragment in the binocular image sequence at the next moment is determined; the spatial coordinates of each fragment in the candidate spatial region at the previous and subsequent moments are calculated, the instantaneous velocity of each fragment is determined by the coordinates at the previous and subsequent moments, and the target fragment at each moment is determined by the instantaneous velocity; based on the spatial coordinates of the target fragment at different moments, the dynamic tracking of the target fragment is completed.

[0008] Preferably, the Rank transformation includes: calculating the number of pixels in the local neighborhood window whose gray values ​​are less than the gray value of the center pixel, as a local statistical feature; the Census transformation includes: generating a binary feature by calculating the gray difference between the neighboring pixels and the center pixel in the local neighborhood window; and fusing the local statistical feature with the binary feature to obtain a first composite feature.

[0009] Preferably, before selecting a target fragment from the left view of the binocular image sequence based on the current time and extracting the feature points of the target fragment as a local neighborhood window, the method further includes correcting the binocular image sequence using the Bouguet epipolar correction algorithm. Specifically, this includes: acquiring the relative angle data and relative distance data of the left and right cameras of the binocular image sequence acquisition device; determining the rotation matrix R and translation vector T of the left and right cameras using the relative angle data and relative distance data; and correcting the binocular high-speed image sequence based on the rotation matrix R and translation vector T to obtain a row-aligned corrected image pair sequence.

[0010] Preferably, the absolute difference of the Rank changes between the first composite feature and the second composite feature is calculated, and the Hamming distance of the Census changes between the first composite feature and the second composite feature is calculated. The difference value between the first composite feature and each second composite feature is determined by the absolute difference of the Rank changes and the Hamming distance of the Census changes.

[0011] Preferably, the step of calculating the spatial coordinates of each fragment at the preceding and following moments within the candidate spatial region, determining the instantaneous velocity of each fragment from the coordinates at the preceding and following moments, and determining the target fragment at each moment from the instantaneous velocity, specifically includes: using the spatial coordinates of the preceding and following moments to determine the instantaneous velocity of each fragment, determining the average velocity of the entire explosion field from the instantaneous velocity of each fragment, and selecting the fragment corresponding to the instantaneous velocity that differs the least from the average velocity as the target fragment for the next moment.

[0012] Preferably, the preset fragment motion constraint range is determined based on the physical structural parameters of the explosive charge and the fragment diffusion model.

[0013] Preferably, determining the candidate spatial region of the target fragment in the binocular image sequence at the next moment further includes: introducing a dynamic compensation factor to adaptively adjust the diffusion angle range of each fragment to compensate for trajectory deviation caused by aerodynamic disturbance.

[0014] This invention also provides a dynamic tracking method system for explosive fragments, comprising: The data acquisition module is used to synchronously acquire high-speed images of the explosion field at continuous moments using binoculars, forming a binocular image sequence; the binocular image sequence includes a left view and a right view.

[0015] The object finding module is used to select a target fragment from the left view of the binocular image sequence based on the current time, extract the feature points of the target fragment as a local neighborhood window, perform Rank transformation and Census transformation on the local neighborhood window respectively, and fuse the two transformation results to obtain a first composite feature; from the right view of the binocular image sequence, select all fragments that are in the same scan line as the local neighborhood window of the left view, select the local neighborhood window of each fragment, and perform Rank transformation and Census transformation on each local neighborhood window in turn to obtain multiple second composite features; calculate the difference value between the first composite feature and each second composite feature, and select the fragment corresponding to the smallest difference value from all the difference values ​​to form a fragment pair with the target fragment.

[0016] The positioning module is used to determine the spatial coordinates of the target fragment at the current moment based on the pixel coordinates of the same fragment pair in the binocular image sequence; based on the spatial coordinates of the target fragment at the current moment and the preset fragment motion constraint range, it determines the candidate spatial region of the target fragment in the binocular image sequence at the next moment; it calculates the spatial coordinates of each fragment in the candidate spatial region at the previous and subsequent moments, determines the instantaneous velocity of each fragment from the coordinates at the previous and subsequent moments, and determines the target fragment at each moment from the instantaneous velocity; and completes the dynamic tracking of the target fragment based on the spatial coordinates of the target fragment at different moments.

[0017] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement any of the steps in the dynamic tracking method for explosive fragments.

[0018] The present invention also provides a computer-readable storage medium storing a computer program that, when loaded by a processor, can execute any of the steps in the dynamic tracking method for explosive fragments.

[0019] The dynamic tracking method for explosive fragments provided by this invention has the following beneficial effects: Firstly, this invention constrains the search range in the right image to the same scan line as the target in the left image, reducing the traditional two-dimensional full-image search to a one-dimensional search, significantly reducing computational load and ensuring efficiency in processing high-speed image sequences. Secondly, addressing the matching ambiguity problem caused by weak fragment texture and similar morphology, this invention innovatively integrates Rank transform and Census transform into the target window, achieving robustness to illumination and noise and sensitivity to local structures. This ultimately generates a composite feature descriptor, requiring both statistical characteristics and texture structure similarity to be satisfied during matching, thereby achieving high-precision differentiation among numerous similar candidate points and significantly reducing the false matching rate. Furthermore, subsequent constraints on the predicted tracking area through a motion model further ensure the stability of long-term tracking. This invention significantly improves stereo matching efficiency and enhances matching robustness in complex environments. Through dual guarantees of spatial constraints and velocity screening, it achieves stable tracking of ultra-high-speed fragments. Fragment spatial trajectory and velocity data can be obtained solely through algorithm optimization, achieving both high-efficiency and high-precision tracking processing in high-speed, high-interference explosive fragment fields. Attached Figure Description

[0020] To more clearly illustrate the embodiments and design schemes of the present invention, the accompanying drawings required for this embodiment will be briefly described below. 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.

[0021] Figure 1 This is a flowchart illustrating a dynamic tracking method for explosive fragments according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating an embodiment of the present invention. Figure 3 This is an enhanced grayscale image of the left and right explosion fragments from an embodiment of the present invention. Figure 3 Image (a) is the left fragment image. Figure 3 (b) is the image of the right fragment; Figure 4 This is a schematic diagram illustrating the polar correction process according to an embodiment of the present invention. Figure 5 These are calibration images before and after Bouguet epipolar correction, based on an embodiment of the present invention. Figure 6 This is a preliminary fragment matching effect diagram according to an embodiment of the present invention. Figure 6 (a) is the left target fragment. Figure 6 (b) is the right candidate target fragment; Figure 7 This is a diagram showing the distribution of target pixel values ​​in the left and right images according to an embodiment of the present invention. Figure 7(a) shows the distribution of pixel values ​​of the left target fragment. Figure 7 (b) shows the distribution of pixel values ​​for the right candidate detected target fragment. Figure 7 (c) represents the distribution of pixel values ​​of the right candidate detected target fragment. Figure 7 (d) represents the distribution of pixel values ​​for the right candidate detected target fragment; Figure 8 This is a schematic diagram of the predicted fragment distribution area after dynamic compensation according to an embodiment of the present invention. Figure 8 (a) is the left fragment. Figure 8 (b) is the right fragment; Figure 9 Embodiments of the present invention Fragments of Time Position in the left and right images, Figure 9 Image (a) is the left fragment image. Figure 9 (b) is the image of the right fragment; Figure 10 This is a schematic diagram showing the farthest and closest distances from fragments in different regions to the epicenter, according to an embodiment of the present invention. Figure 11 Embodiments of the present invention A schematic diagram illustrating the range of movement positions of the fragment at any given moment; Figure 12 Embodiments of the present invention The positions of the two valid fragments in the left and right images at any given time. Figure 12 Image (a) is the left fragment image. Figure 12 (b) is the image of the right fragment; Figure 13 This is a diagram showing the trajectory of the fragment movement in the left and right images of an embodiment of the present invention. Figure 13 (a) is the left fragment. Figure 13 (b) is the right fragment; Figure 14 These represent the spatial positions of the fragments at different times in embodiments of the present invention. Figure 15 This is a graph showing the fragmentation velocity-time relationship in an embodiment of the present invention. Detailed Implementation

[0022] To enable those skilled in the art to better understand and implement the technical solutions of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be construed as limiting the scope of protection of the present invention.

[0023] This invention proposes a dynamic tracking method for explosive fragments based on high-speed binocular images. By employing the Bouguet epipolar correction algorithm, the Rank-Census hybrid binocular matching algorithm, and a fragment diffusion model, the tracking and measurement of explosive fragments are achieved. The core idea is to apply the Bouguet epipolar correction algorithm to binocular image processing, reducing the matching point search dimension from two dimensions to one dimension, significantly improving stereo matching efficiency. In the epipolar-corrected image, the Rank-Census hybrid binocular matching algorithm is used to accurately match multiple candidate targets at the same height in the right image where a target in the left image may correspond to multiple candidate targets. Based on the fragment diffusion model, by defining the fragment position and the range of the explosion center distance, the average velocity of multiple fragments is calculated, and the fragment closest to this value is selected as the trajectory, effectively enhancing tracking robustness.

[0024] like Figure 2 As shown in the flowchart, this system demonstrates a complete process for binocular vision-based explosive fragment tracking and parameter extraction. First, the acquired fragment scattering images undergo grayscale conversion and contrast enhancement preprocessing. Then, Bouquet epipolar correction aligns corresponding targets in the left and right images, followed by coarse matching and fine matching based on Rank-Census transform composite features. Weight adjustments optimize the matching results to achieve accurate fragment pairing. Based on this, the system combines explosive charge structure information and a dynamic compensation algorithm to predict the fragment distribution area at the next moment. By calculating the distance from the detonation center and the fragment velocity within the area, the fragment closest to the average velocity is selected as the continuous tracking target. Finally, the system outputs multi-dimensional parameters such as the fragment's spatial position, velocity, scattering angle, and trajectory, thus completing the entire process from image processing and stereo matching to trajectory tracking and motion analysis.

[0025] Based on this, the present invention provides a dynamic tracking method for explosive fragments, specifically as follows: Figure 1 As shown, it includes: S1. Synchronously acquire high-speed images of the explosion field at continuous moments using binoculars to form a binocular image sequence; the binocular image sequence includes a left view and a right view.

[0026] Based on the Bouguet epipolar correction algorithm, Rank-Census hybrid binocular matching, and fragment diffusion model, the tracking and measurement of explosive fragments was achieved. The specific implementation method is as follows:

[0027] Matching fragments of the same name S101. The acquired RGB three-channel image is converted into a grayscale image using a weighted average method, and the contrast between the target and the background is enhanced using histogram equalization technology. The enhancement effect of the grayscale image of the explosion fragments is as follows. Figure 3 As shown, Figure 3 Image (a) is the left fragment image. Figure 3 (b) is the image of the right fragment.

[0028] S102. The Bouguet epipolar correction algorithm adjusts the geometry of the binocular camera system to make the optical axes of the left and right cameras parallel, while aligning the baseline (the straight line connecting the optical centers of the left and right cameras) with the x-axis of the image plane. This eliminates vertical parallax in the original image, allowing matching points in the corrected image to be searched only on the same horizontal line, thus reducing the matching search space. The principle diagram of the correction process is shown below. Figure 4 As shown.

[0029] The calibration of binocular cameras is accomplished by detecting the coordinates of preset landmarks or natural feature points within their common field of view. After calibration, the intrinsic and extrinsic parameters of each camera, as well as the spatial pose relationship (rotation matrix) between the binocular systems, can be obtained. Translation vector Based on these parameters, the Bouguet epipolar correction algorithm is used for image correction. First, R is decomposed to obtain... and :

[0030] ; In equation (1), yes The inverse matrix, and This is for synthesizing rotation matrices.

[0031] Secondly, through the translation vector of the right camera relative to the left camera. T Implement and construct the transformation matrix. : ; In equation (2), ; ; ; .

[0032] Will , respectively with Multiplying them together yields the stereo correction matrix for the binocular camera. and : ; The overall rotation matrix calculated using formula (3) is used to transform the coordinate systems of the left and right cameras respectively, so that the baseline is completely parallel to the two imaging planes, ultimately obtaining a stereo image pair that meets the requirements of line alignment. The calibration images before and after Bouguet epipolar correction are shown below. Figure 5 As shown.

[0033] Based on the binocular image pair that has undergone epipolar geometric correction, target fragment feature points are selected in the left view image. For reference, then based on the epipolar constraint conditions, coarse matching and filtering of feature points of the same fragment are performed within the corresponding epipolar region in the right view, as illustrated in the diagram below. Figure 6 As shown, Figure 6 (a) is the left target fragment. Figure 6 (b) is the right candidate detection target fragment.

[0034] S2. Based on the current time, select a target fragment from the left view of the binocular image sequence, extract the feature points of the target fragment as a local neighborhood window, perform Rank transformation and Census transformation on the local neighborhood window respectively, and fuse the two transformation results to obtain the first composite feature; select all fragments in the same scan line as the local neighborhood window of the left view from the right view of the binocular image sequence, select the local neighborhood window of each fragment, and perform Rank transformation and Census transformation on each local neighborhood window in turn to obtain multiple second composite features; calculate the difference value between the first composite feature and each second composite feature, and select the fragment corresponding to the smallest difference value from all the difference values ​​to form a fragment pair with the target fragment.

[0035] S201. In the left image, using the centroid coordinates of the target fragment ( x,y A local neighborhood window is constructed centered on the center pixel; Rank transform and Census transform are then performed on this window. The Rank transform calculates the number of pixels within the window whose gray values ​​are less than the gray value of the center pixel. As a local statistical feature; the Census transform generates a binary feature descriptor by comparing the gray levels of neighboring pixels with the center pixel. In the right image matching stage, for the initially screened candidate fragments with the same name, the same Rank-Census feature extraction process is performed at the corresponding centroid position using a window of the same size, so as to establish the feature correspondence for subsequent accurate matching.

[0036] S202. In the feature matching stage, a dual-metric criterion is used for similarity evaluation; firstly, the Rank feature of the candidate region in the right image is calculated. Rank features of the reference region in the left image absolute difference First, as a statistical feature similarity measure; second, perform a bitwise XOR operation on the Census binary descriptors of corresponding regions in the left and right images, and the number of 1s in the statistical result is... , as a measure of local texture similarity. These two difference measures and Together, they constitute the composite discrimination criterion for feature matching.

[0037] S203. In the feature matching decision stage, a multi-criteria optimization strategy is adopted; when the Rank feature difference measurement... The minimum location and the Hamming distance of the Census feature When the minimum values ​​of the two metrics coincide, the window region corresponding to that position is determined to be the optimal matching target. When the spatial positions of the two optimal metrics are inconsistent, a weighted fusion strategy is introduced. By adaptively adjusting the weight coefficients of the Rank transform and the Census transform, a composite matching cost function is constructed. Finally, the window that makes the composite cost function C achieve the global minimum value is selected as the correct matching result.

[0038] This matching strategy effectively balances the contributions of statistical and structural features. Its composite matching cost function is:

[0039] ; In equation (4), and To integrate weights, .

[0040] With fragmentation target For example, the target fragment in the left image. And the candidate target fragment in the right image Regional pixel value distribution as follows Figure 7 As shown, Figure 7 (a) shows the distribution of pixel values ​​of the left target fragment. Figure 7 (b) shows the distribution of pixel values ​​for the right candidate detected target fragment. Figure 7 (c) represents the distribution of pixel values ​​of the right candidate detected target fragment. Figure 7 (d) represents the distribution of pixel values ​​of the right candidate detected target fragment.

[0041] Feature extraction is performed using Rank and Census transforms on the pixel values ​​of the target region of the fragment in the left and right images. Considering the limitations of fragment surface characteristics, motion speed, and imaging conditions, fragments often exhibit weak texture features. Therefore, a weighted fusion strategy is adopted ( α =0.6, β =0.4) enhances the robustness contribution of Rank transformation. Based on the hybrid matching cost calculation model of formula (6), the weights of Rank and Census transformations are adjusted ( α =0.6, β =0.4), resulting in the target fragment in the left image. Candidate targets in the right image The matching costs are 9, 13, and 25 respectively. Based on the minimum cost criterion, the matching costs are determined as follows: for correct matching point.

[0042] S3. Based on the pixel coordinates of the same fragment pair in the binocular image sequence, determine the spatial coordinates of the target fragment at the current moment; based on the spatial coordinates of the target fragment at the current moment and the preset fragment motion constraint range, determine the candidate spatial region of the target fragment in the binocular image sequence at the next moment; calculate the spatial coordinates of each fragment in the candidate spatial region at the previous and subsequent moments, determine the instantaneous velocity of each fragment from the coordinates at the previous and subsequent moments, and determine the target fragment at each moment from the instantaneous velocity; based on the spatial coordinates of the target fragment at different moments, complete the dynamic tracking of the target fragment.

[0043] Same-name tracking S301, Fragments obtained based on binocular vision matching and Based on the initial position and the dimensions of the cylindrical charged explosive, the predicted distribution area of ​​fragments in the left and right images at the next sampling time can be derived. However, in actual testing, because the height of the cylindrical charged explosive is much smaller than the detonation distance, the fragment distribution angle is very small. Even with a high-sampling-rate binocular high-speed system, the fragments are affected by aerodynamic disturbances during flight, and their trajectory is not straight, which may cause their actual position to deviate from the predicted area. Therefore, a dynamic compensation algorithm is proposed to adaptively adjust the predicted distribution angle. The schematic diagram of the predicted fragment distribution area after dynamic compensation is shown below. Figure 8 As shown, Figure 8 (a) is the left fragment. Figure 8 (b) is the right fragment.

[0044] The distribution angle of the same fragments in the left and right images at the next time step and for: ; ; In equations (5) and (6), It is a dynamic compensation factor; The fragments in the left and right images are respectively and The angle between the maximum and minimum distance between the burst centers; , These represent the distances from the top and bottom of the cylindrical charge in the left image to the fragment, respectively. , These represent the distances from the top and bottom of the cylindrical charge in the right image to the fragment, respectively. This refers to the height of the explosive charge with its casing.

[0045] Target fragments For example, based on the relative positional relationship between the cylindrical loaded charge and the fragments, the angle of the charge in the left and right images is calculated to be 8.1°. Fragments of Time The positions in the left and right images are as follows Figure 9 As shown, Figure 9 Image (a) is the left fragment image. Figure 9 (b) is the image of the right fragment.

[0046] S302. Since the initial generation position of fragments within a cylindrical encased charge cannot be precisely located, the possible range of values ​​for the distance from the detonation center is calculated by jointly using the geometric parameters of the cylindrical encased charge and the spatial distribution data of the fragments. A schematic diagram illustrating the farthest and closest distances from fragments to the detonation center in different regions is shown below. Figure 10 As shown. The distance from the tip of the cylindrical charge to the fragment; The distance from the bottom of the cylindrical, encased charge to the fragment; The vertical distance from the fragment to the cylindrical encased charge; The angle between the line connecting the tip of the cylindrical casing charge to the fragment and the generatrix of the cylindrical casing charge; The angle between the line connecting the tip of the cylindrical casing charge to the fragment and the generatrix of the cylindrical casing charge; The distance from the tip of the cylindrical encased charge to the fragment. Distance from the bottom of the cylindrical encased charge to the fragment The angle between them.

[0047] The distance from the burst center of the fragment varies in different regions. R The range is: (1) When the fragment is in region 1, i.e. hour, ; (2) When the fragment is in region 2, i.e. hour, ; (3) When the fragment is in region 3, i.e. hour, .

[0048] After determining the initial fragment burst center distance range, the derivation process of the fragment burst center distance at the next sampling time is as follows: From And fragment velocity decay formula We can obtain: ; Integrating both sides of the above equation, we get: ; From the above formula, we can obtain: ; In equation (9), t 1 and t 2 These represent the time it takes for the fragment to reach its initial position and its next position from the moment of detonation, respectively. This is the initial velocity of the fragment; The fragment attenuation coefficient; and They represent fragments in t 1 and t 2 The distance from the center of the explosion at any given moment.

[0049] Initial position Substituting the maximum and minimum burst center distances of the fragments at each moment into equation (9), we obtain... The maximum and minimum burst distances of the fragments at any given time; according to The range of the explosion center distance at any given time and the corrected distribution angle are used to determine... The location and range of the fragment at any given time, such as Figure 11 As shown. , For respectively The minimum and maximum distance between the burst centers of the fragments at any given moment; , For respectively The minimum and maximum distance between the burst centers of the fragments at any given moment. for Minimum burst distance of fragments at any given moment and maximum center-blast distance The angle between them.

[0050] Taking fragment 1 as an example, combining equation (9) with t 1 Fragments of Time The range of values ​​for is used to derive the distance from the center of the explosion. The range is 3.34m≤ ≤3.67m; The fragment distribution angle is determined by iterative adjustment of the dynamic compensation factor K: when K When the radius is 3.4, the fragment distribution angle is 21.06°, indicating that there are two valid fragments in this region, which meets the fragment radius constraint. The positions of the two valid fragments in the left and right images at any given time are as follows: Figure 12 As shown, Figure 12 Image (a) is the left fragment image. Figure 12 (b) is the image of the right fragment.

[0051] S303, according to the above Given a fixed distance between the fragment burst centers and their distribution angles, the instantaneous velocities and arithmetic mean velocities of all fragments within the target area are calculated. Next, the minimum deviation method is used to select target fragments for tracking and velocity calculation at the next moment. The trajectory tracking and velocity calculation process is consistent with the aforementioned method and is therefore omitted. The final output shows the fragment trajectories in the left and right images at different times, the spatial positions of fragments at different times, and the fragment velocity-time relationship curves, as shown below. Figure 13, Figure 14 , Figure 15 As shown, Figure 13 (a) is the left fragment. Figure 13 (b) is the right fragment.

[0052] In summary, this invention proposes a dynamic tracking method for explosive fragments based on binocular high-speed images. Tracking and measurement are achieved through the Bouguet epipolar correction algorithm, Rank-Census hybrid binocular matching, and fragment diffusion model. The specific details are as follows:

[0053] 1. Based on the intrinsic and extrinsic parameters calibrated by the binocular camera, the Bouguet epipolar correction algorithm is used for binocular image correction. After correction, the vertical parallax in the original image can be eliminated, so that the matching points in the corrected image only need to be searched on the same horizontal row.

[0054] 2. Based on the statistical robustness of the Rank transform and the texture sensitivity of the Census transform, this invention proposes a Rank-Census hybrid binocular matching algorithm. This method integrates the feature description capabilities of both transforms, constructing a more robust matching cost function under epipolar geometric constraints, thereby achieving high-precision matching of fragment target features in the left and right images.

[0055] 3. Based on the fragment diffusion model, by defining the fragment location and the range of the explosion center distance, the average velocity of multiple fragments is calculated and the fragment closest to this value is selected as the trajectory, which effectively enhances the tracking robustness.

[0056] The core advantage of the dynamic tracking method for explosive fragments based on binocular high-speed images proposed in this invention lies in the construction of a multi-level robust system, which systematically overcomes many traditional challenges in explosive fragment field measurement. This method does not merely improve a single step, but rather achieves end-to-end enhancement from data acquisition to result output through the deep integration of algorithms, models, and strategies.

[0057] Specifically, firstly, by employing Bouguet epipolar correction, the complex two-dimensional global matching problem is simplified into an efficient one-dimensional row search, significantly improving the computational efficiency and real-time potential of stereo matching. Secondly, an innovative Rank-Census hybrid matching algorithm is proposed, cleverly combining the stability of statistical features with the sensitivity of texture features. This allows the algorithm to maintain high-precision matching of corresponding points even under extremely harsh imaging conditions such as weak texture of the fragment target, motion blur in the image, and smoke noise, laying a reliable foundation for subsequent 3D reconstruction. Finally, and most notably, this method creatively combines visual observation with prior physical knowledge. By introducing a fragment diffusion model to constrain the prediction range and utilizing velocity consistency statistical decision criteria for trajectory association, this strategy enables the system to effectively handle trajectory intersection and occlusion problems caused by high-density fragment groups, significantly enhancing the continuity and stability of long-term tracking. Ultimately, it achieves accurate and robust measurement of the 3D trajectory and velocity parameters of high-speed moving fragments in a real and complex explosion environment.

[0058] Based on the same inventive concept, this invention also provides a dynamic tracking method system for explosive fragments, comprising: The data acquisition module is used to synchronously acquire high-speed images of the explosion field at continuous moments using binoculars, forming a binocular image sequence; the binocular image sequence includes a left view and a right view.

[0059] The object finding module is used to select a target fragment from the left view of the binocular image sequence based on the current time, extract the feature points of the target fragment as a local neighborhood window, perform Rank transformation and Census transformation on the local neighborhood window respectively, and fuse the two transformation results to obtain a first composite feature; from the right view of the binocular image sequence, select all fragments that are in the same scan line as the local neighborhood window of the left view, select the local neighborhood window of each fragment, and perform Rank transformation and Census transformation on each local neighborhood window in turn to obtain multiple second composite features; calculate the difference value between the first composite feature and each second composite feature, and select the fragment corresponding to the smallest difference value from all the difference values ​​to form a fragment pair with the target fragment.

[0060] The positioning module is used to determine the spatial coordinates of the target fragment at the current moment based on the pixel coordinates of the same fragment pair in the binocular image sequence; based on the spatial coordinates of the target fragment at the current moment and the preset fragment motion constraint range, it determines the candidate spatial region of the target fragment in the binocular image sequence at the next moment; it calculates the spatial coordinates of each fragment in the candidate spatial region at the previous and subsequent moments, determines the instantaneous velocity of each fragment from the coordinates at the previous and subsequent moments, and determines the target fragment at each moment from the instantaneous velocity; and completes the dynamic tracking of the target fragment based on the spatial coordinates of the target fragment at different moments.

[0061] This invention also provides a computer device, which, at the hardware level, includes a processor, an internal bus, a network interface, memory, and non-volatile memory, and may also include other hardware required for business operations. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to implement the dynamic tracking method for explosive fragments provided above.

[0062] The present invention also provides a computer-readable storage medium storing a computer program that can be used to execute the above-described dynamic tracking method for explosive fragments.

[0063] Specific limitations on the computational system for the dynamic tracking method of explosive fragments can be found in the limitations of the dynamic tracking method for explosive fragments described above, and will not be repeated here. Each module in the above-described dynamic tracking method system for explosive fragments can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0064] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification. Furthermore, the above embodiments only illustrate several implementation methods of this application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for dynamic tracking of explosive fragments, characterized in that, include: The binocular system simultaneously acquires high-speed images of the explosion field at continuous moments, forming a binocular image sequence; the binocular image sequence includes a left view and a right view; Based on the current moment, a target fragment is selected from the left view of the binocular image sequence, and the feature points of the target fragment are extracted as local neighborhood windows. Rank transformation and Census transformation are performed on the local neighborhood windows respectively, and the two transformation results are fused to obtain the first composite feature. From the right view of the binocular image sequence, all fragments on the same scan line as the local neighborhood windows of the left view are selected, and the local neighborhood windows of each fragment are selected. Rank transformation and Census transformation are then performed on each local neighborhood window respectively to obtain multiple second composite features. Calculate the difference value between the first composite feature and each second composite feature, and select the fragment corresponding to the smallest difference value from all the difference values ​​to form a fragment pair with the target fragment; Based on the pixel coordinates of the same fragment pair in the binocular image sequence, the spatial coordinates of the target fragment at the current moment are determined; based on the spatial coordinates of the target fragment at the current moment and the preset fragment motion constraint range, the candidate spatial region of the target fragment in the binocular image sequence at the next moment is determined; the spatial coordinates of each fragment in the candidate spatial region at the previous and subsequent moments are calculated, the instantaneous velocity of each fragment is determined by the coordinates at the previous and subsequent moments, and the target fragment at each moment is determined by the instantaneous velocity; based on the spatial coordinates of the target fragment at different moments, the dynamic tracking of the target fragment is completed.

2. The dynamic tracking method for explosive fragments according to claim 1, characterized in that, The Rank transformation includes: calculating the number of pixels in the local neighborhood window whose gray values ​​are less than the gray value of the center pixel, as a local statistical feature; the Census transformation includes: generating a binary feature by calculating the gray difference between the neighboring pixels and the center pixel in the local neighborhood window; and fusing the local statistical feature and the binary feature to obtain a first composite feature.

3. The dynamic tracking method for explosive fragments according to claim 1, characterized in that, Before selecting a target fragment from the left view of the binocular image sequence based on the current time and extracting the feature points of the target fragment as a local neighborhood window, the method further includes using the Bouguet epipolar correction algorithm to correct the binocular image sequence. Specifically, this includes: acquiring the relative angle data and relative distance data of the left and right cameras of the binocular image sequence acquisition device; determining the rotation matrix R and translation vector T of the left and right cameras using the relative angle data and relative distance data; and correcting the binocular high-speed image sequence according to the rotation matrix R and translation vector T to obtain a row-aligned corrected image pair sequence.

4. The dynamic tracking method for explosive fragments according to claim 1, characterized in that, Calculate the absolute difference of the Rank changes between the first composite feature and the second composite feature, and calculate the Hamming distance of the Census changes between the first composite feature and the second composite feature. Determine the difference value between the first composite feature and each second composite feature based on the absolute difference of the Rank changes and the Hamming distance of the Census changes.

5. The dynamic tracking method for explosive fragments according to claim 1, characterized in that, The calculation of the spatial coordinates of each fragment within the candidate spatial region at previous and subsequent times, determining the instantaneous velocity of each fragment from the coordinates at previous and subsequent times, and determining the target fragment at each moment from the instantaneous velocity, specifically includes: Using the spatial coordinates of the two moments before and after, the instantaneous velocity of each fragment is determined, and the average velocity of the entire explosion field is determined by the instantaneous velocity of each fragment; the fragment corresponding to the instantaneous velocity with the smallest difference from the average velocity is selected as the target fragment for the next moment.

6. The dynamic tracking method for explosive fragments according to claim 1, characterized in that, The preset fragment motion constraint range is determined based on the physical structural parameters of the explosive charge and the fragment diffusion model.

7. The dynamic tracking method for explosive fragments according to claim 1, characterized in that, The method of determining the candidate spatial region of the target fragment in the binocular image sequence at the next moment also includes: introducing a dynamic compensation factor to adaptively adjust the diffusion angle range of each fragment in order to compensate for the trajectory deviation caused by aerodynamic disturbance.

8. A dynamic tracking method system for explosive fragments, characterized in that, include: The data acquisition module is used to synchronously acquire high-speed images of the explosion field at continuous moments using binoculars, forming a binocular image sequence; the binocular image sequence includes a left view and a right view; The object finding module is used to select a target fragment from the left view of the binocular image sequence based on the current time, extract the feature points of the target fragment as a local neighborhood window, perform Rank transformation and Census transformation on the local neighborhood window respectively, and fuse the two transformation results to obtain the first composite feature; from the right view of the binocular image sequence, select all fragments that are on the same scan line as the local neighborhood window of the left view, select the local neighborhood window of each fragment, and perform Rank transformation and Census transformation on each local neighborhood window in turn to obtain multiple second composite features; Calculate the difference value between the first composite feature and each second composite feature, and select the fragment corresponding to the smallest difference value from all the difference values ​​to form a fragment pair with the target fragment; The positioning module is used to determine the spatial coordinates of the target fragment at the current moment based on the pixel coordinates of the same fragment pair in the binocular image sequence; based on the spatial coordinates of the target fragment at the current moment and the preset fragment motion constraint range, it determines the candidate spatial region of the target fragment in the binocular image sequence at the next moment; it calculates the spatial coordinates of each fragment in the candidate spatial region at the previous and subsequent moments, determines the instantaneous velocity of each fragment from the coordinates at the previous and subsequent moments, and determines the target fragment at each moment from the instantaneous velocity; and completes the dynamic tracking of the target fragment based on the spatial coordinates of the target fragment at different moments.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is loaded by the processor, it is able to perform the steps of the method according to any one of claims 1 to 7.

Images