Method and system for detecting and centroid correcting a target for cross-pixel broadening

Abstract

The present invention relates to the technical field of spatial infrared camera information processing. The present invention discloses a method and system for detecting and centroid correction of cross-pixel broadened targets, which are used for the cross-pixel broadened point target splitting phenomenon that occurs in the multi-tap readout mode of a high-frame-rate infrared focal plane detector. A target cross-pixel broadened splitting degradation model is constructed, and by using the strong correlation between the splitting points, a target detection process of pre-detection, energy dimension iteration, model update, and post-detection is designed. Through single-frame morphological filtering and multi-hypothesis tracking association, the detection and independent tracking of each observation of the splitting points are realized, and the statistical information of the splitting points is obtained. Then, the standardized and corrected degradation model is mapped to the current detection image, and the model parameters are iteratively updated in the energy dimension, so as to realize target centroid correction. The technical solution of the present invention can improve the positioning accuracy and target identification ability of spatial infrared remote sensing instruments, enabling them to effectively distinguish multi-targets and false targets caused by splitting points.

Description

Technical Field

[0001] The present invention relates to the technical field of spatial infrared camera information processing, and in particular to a method for detecting and centroid correcting a cross-pixel broadened target. Background Art

[0002] In the fields of airborne infrared detection, satellite remote sensing, etc., detectors often use a multi-tap alternating interval readout method to increase the imaging frame rate several times to meet the detection requirements of high temporal resolution. However, due to the insufficient bandwidth of the readout circuit, the point spread of the optical system, and the influence of the cross-pixel effect of point targets, when imaging point targets mainly composed of stars, the target is cross-pixel broadened. Under the multi-tap interval readout and fixed pixel arrangement, the target presents a split point phenomenon. Under the high-gain response and high-frequency readout of high-frame-rate infrared focal plane detectors, the split point phenomenon is more prominent, which reduces the positioning accuracy and target identification ability of infrared remote sensing instruments.

[0003] Existing vision saliency algorithms utilize the characteristics of the target's Gaussian-like distribution and the background consistency features, construct a saliency map by building multiple windows and introducing multiple local contrast measures, and the overall detection effect is relatively good. However, for the cross-pixel broadened "split" target, the gray distribution no longer satisfies the Gaussian-like characteristics, the windows used to calculate the saliency map fail, and it is easy to miss the "tails" of the split.

[0004] In addition, relying solely on the spatial domain information obtained from a single frame cannot meet the detection requirements of high detection probability and low false alarm rate for split point targets. It is necessary to further combine the temporal domain information and perform multi-frame data association to achieve independent tracking of multiple observations of split point targets. Commonly used data association tracking techniques include global nearest neighbor association, joint probabilistic data association, and multiple hypothesis tracking algorithms. Among them, both the global nearest neighbor association and joint probabilistic data association algorithms are single-scan type methods. The trajectory association of the current frame only depends on the association decision of the previous frame and the observation data of the current frame. When neighboring among associated targets, trajectory mis-tracking is likely to occur. Summary of the Invention

[0005] There is a gap in the research algorithms for detecting the morphological split point targets of the above-mentioned existing high-frame-rate infrared focal plane detectors, and the identification ability for this target morphology is poor; the existing technology uses single-scan type methods, and when neighboring among associated targets, trajectory mis-tracking is likely to occur. The present invention provides a method for detecting and centroid correcting a cross-pixel broadened target. By using the multiple hypothesis tracking technology to combine the observation data of multiple scans and delaying the decision of trajectory association, it can effectively reduce the interlacing of neighboring trajectories and improve the multi-target identification ability of high-frame-rate infrared focal plane detectors; the positioning accuracy of high-frame-rate infrared focal plane detectors is improved by correcting the centroid positioning error.

[0006] The method for detecting and centroid correcting a cross-pixel broadened target includes the following steps:

[0007] S1. Receive an infrared image frame, establish a potential trajectory hypothesis tree based on suspicious targets, predict the state of the next frame; receive the next infrared image frame, and determine the gating relationship between each observation and the predicted state of each trajectory's current frame;

[0008] S2. Generate a corresponding weighted undirected graph based on the trajectory hypothesis tree, generate the global optimal hypothesis for each trajectory by solving the maximum weighted independent set of the weighted undirected graph, according to the global optimal hypothesis solved at the scanning time k, trace back to the node at the k - N 1 moment, prune the branches deviating from the current optimal hypothesis, confirm the association decision at the k - N 1 +1 moment, continuously associate the N 2 frame images, and screen the target trajectories through the trajectory length and the proportion of true associated points;

[0009] S3. Statistically confirm the coordinate differences between the trajectories, compare with the number p of detector taps, record the pairs of trajectories and the satisfaction conditions, when the N 3 frame determines that the pair of trajectories satisfies this condition, obtain the split point statistical information;

[0010] S4. Construct a target cross - pixel broadening split degradation model, based on the split point statistical information, determine the rationality of the target cross - pixel broadening split degradation model, when the deviation value is greater than the threshold, perform iterative correction of the target cross - pixel broadening split degradation model, update the model parameters, and re - use it for detecting images.

[0011] Further, the step S1 includes:

[0012] S101. Receive an infrared image frame, perform single - frame morphological filtering, obtain an enhanced image by enhancing the target saliency of the infrared image, perform threshold segmentation on the enhanced image, and screen out suspicious targets;

[0013] S102. Establish a potential trajectory hypothesis tree for each suspicious target, initialize the trajectory state, construct a Kalman filter, predict the state of the next frame of the suspicious target, and initialize the trajectory score;

[0014] S103. Receive a new frame of detection image, determine the gating relationship between each observation and the predicted state of each trajectory's current frame, for the observations within the association gate of a certain trajectory, perform trajectory association; for the observations not within any trajectory association gate, perform trajectory initiation, and re - execute the step S102.

[0015] Further, the step S2 includes: Generate a corresponding weighted undirected graph G=(V, E, W) according to the constructed trajectory hypothesis tree, and the vertex vi∈V in G is defined as a track node The weight wi ∈ W of vertex vi is defined as the track score For two incompatible trajectories and corresponding vertex v i Vertex vi is connected to vj by an edge (i, j) ∈ E; the globally optimal hypothesis of each trajectory is generated by solving the maximum weighted independent set of the undirected graph, and the expression of the maximum weighted independent set is:

[0016]

[0017]

[0018] where is a binary variable indicating whether the ith track is included in the optimal solution.

[0019] Furthermore, in step S3, the recorded track pair and The satisfaction conditions are as follows, and the expression for the satisfaction conditions is:

[0020]

[0021] The coordinates of the track pair are (X i , Y i ) and (X j , Y j ). Continuously judge whether the track pair satisfies this condition for N 3 frames. If not, it is determined that the track is generated by multiple targets; if so, the split point statistical information is obtained.

[0022] Furthermore, in step S4, the signal of the target cross-pixel broadening splitting degradation model is broadened, which is represented by a low-pass filter, and the sampling deviation and the filtering bandwidth are set;

[0023] Furthermore, in step S4, according to the split point statistical information, the constructed target cross-pixel broadening splitting degradation model is mapped to the current detection image, the centroid of the model solution is calculated, and the centroid deviation is compared with the ideal centroid of the energy concentration synthesis. By comparing with the set deviation threshold, the rationality of the target cross-pixel broadening splitting degradation model is verified. When the deviation exceeds the threshold, it is unreasonable, and the target cross-pixel broadening splitting degradation model is iteratively corrected, the model parameters are updated, and the model parameters are updated and used again for the detection image.

[0024] A system for detecting and centroid correction of cross-pixel broadened targets includes:

[0025] The first unit is used to receive a frame of infrared image, establish a potential trajectory hypothesis tree based on the suspicious target, and predict the state of the next frame; receive the next frame of infrared image, and determine the gating relationship between each observation and the predicted state of the current frame of each trajectory;

[0026] The second unit generates a corresponding weighted undirected graph based on the trajectory hypothesis tree, generates the global optimal hypothesis of each trajectory by solving the maximum weighted independent set of the weighted undirected graph, and traces back to the kNth trajectory according to the global optimal hypothesis solved at the scanning time k. 1 At the node at the moment, prune the branches that deviate from the current optimal hypothesis and confirm kN 1 +1 moment association decision, continuous association N 2 Frame image, filter the target trajectory by trajectory length and the proportion of real associated points;

[0027] The third unit is used to statistically confirm the coordinate difference between the trajectories, compare it with the number of detector taps p, and record the trajectory pair. and The conditions are met when N 3 When the frame judgment trajectory pair meets this condition, the split point statistics are obtained;

[0028] The fourth unit is used to construct a target cross-pixel stretch splitting degradation model, determine the rationality of the target cross-pixel stretch splitting degradation model based on splitting point statistical information, and when the deviation value is greater than a threshold, iteratively correct the target cross-pixel stretch splitting degradation model, update the model parameters, and reuse it to detect the image.

[0029] A computer-readable storage medium stores a computer program, which, when executed by a processor of a computer, enables the computer to execute the method for detecting and correcting a target with cross-pixel stretching.

[0030] An electronic device, characterized in that it includes: a memory storing a computer program; a processor reading the computer program stored in the memory to execute the detection and centroid correction method for cross-pixel widening targets.

[0031] A computer program product, characterized in that the computer program product comprises a computer program, and the computer program is executed by a processor to implement the detection and centroid correction method for cross-pixel widening targets.

[0032] Compared with the prior art, the advantages and beneficial effects of the present invention are as follows: Through the technical solution of the present invention, the gap in the research algorithm for detecting the target morphology of the imaging split point of high-frame-rate infrared focal plane detectors is filled, and the identification ability of the detection instrument for this target morphology is improved; by analyzing the causes of split point targets, a degradation model is constructed and mapped to the detection image, improving the target positioning accuracy; to a certain extent, it compensates for the signal broadening caused by the degradation of the optical system modulation function and insufficient readout bandwidth, etc., and provides algorithm support for removing the optical focusing mechanism from the on-orbit observation infrared camera; it is of great significance for improving the multi-target identification ability and positioning accuracy of high-frame-rate infrared focal plane detectors. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a flowchart of the method of the present invention;

[0034] Figure 2 It is a flowchart of step S1 of the method of the present invention;

[0035] Figure 3 It is a block diagram of the system structure of the present invention;

[0036] Figure 4 It is a schematic diagram of the multi-tap readout logic and the split point formation process. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0037] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present disclosure. In the following detailed description, for the sake of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, obviously, one or more embodiments can also be implemented without these specific details. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessarily confusing the concepts of the present disclosure.

[0038] The terms used herein are merely for describing specific embodiments and are not intended to limit the present disclosure. The terms "including", "comprising", etc. used herein indicate the presence of features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0039] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0040] In the case of using expressions such as "at least one of A, B, and C, etc.", generally, it should be interpreted according to the meaning that those skilled in the art usually understand this expression (for example, "a system having at least one of A, B, and C" should include but not be limited to a system having only A, only B, only C, having A and B, having A and C, having B and C, and / or having A, B, and C, etc.). In the case of using expressions such as "at least one of A, B, or C, etc.", generally, it should be interpreted according to the meaning that those skilled in the art usually understand this expression (for example, "a system having at least one of A, B, or C" should include but not be limited to a system having only A, only B, only C, having A and B, having A and C, having B and C, and / or having A, B, and C, etc.).

[0041] Some block diagrams and / or flowcharts are shown in the drawings. It should be understood that some blocks or combinations thereof in the block diagrams and / or flowcharts can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing devices, so that when executed by the processor, these instructions can create a device for implementing the functions / operations illustrated in these block diagrams and / or flowcharts. The technology of the present disclosure can be implemented in the form of hardware and / or software (including firmware, microcode, etc.). Additionally, the technology of the present disclosure can take the form of a computer program product on a computer-readable storage medium storing instructions, which can be used by or in conjunction with an instruction execution system.

[0042] A method for detecting and centroid correction of cross-pixel spreading targets, comprising the following steps:

[0043] S1. Receive an infrared image frame, establish a potential trajectory hypothesis tree based on suspicious targets, and predict the state of the next frame; receive the next infrared image frame, and determine the gating relationship between each observation and the predicted state of each trajectory in the current frame;

[0044] Specifically, step S1 includes:

[0045] S101. Receive an infrared image frame, perform single-frame morphological filtering, obtain an enhanced image by enhancing the target saliency of the infrared image, and perform threshold segmentation on the enhanced image to screen out suspicious targets;

[0046] The bright top-hat transformation formula is:

[0047] V o = a * WTH - b * BTH

[0048] where V ois the target enhanced result image. WTH and BTH represent the bright top-hat transform and the dark top-hat transform respectively, and a and b are ratio adjustment parameters. For the enhanced image V o perform threshold segmentation, and screen out suspicious targets according to the threshold segmentation;

[0049] S102. Establish a potential trajectory hypothesis tree for each suspicious target, initialize the trajectory state, construct a Kalman filter, predict the state of the next frame of the suspicious target, and initialize the trajectory score;

[0050] The initialized trajectory score expression is:

[0051]

[0052] where λ new and λ F respectively represent the new clutter density and the false alarm clutter density, with the unit of m -2 ;

[0053] S103. Receive a new frame of detection image, determine the gating relationship between each observation and the predicted state of the current frame of each trajectory, perform trajectory association on the observations within the association gate of a certain trajectory; for the observations not within the association gate of any trajectory, perform trajectory initiation, and re-execute step S102.

[0054] Considering the situation of missed detection in the current frame of the trajectory, add a virtual missed detection branch to each trajectory, set the continuous missed detection threshold, and all trajectory hypothesis branches take missed detection into account, which is achieved by adding a missed detection virtual branch. The score increment is ln(1 - P D ). Different from the real-time update of the Kalman filter parameters with associated observations in trajectory association, the Kalman filter parameters at the previous moment will be maintained to predict the target state of the next frame during missed detection. Both the virtual branch and the hypothesis association branch need to participate in the generation of the global optimal hypothesis, and at the same time record the situation of the trajectory association virtual branch. There are two purposes: 1. When the continuous missed detection reaches the threshold, delete the current trajectory hypothesis; 2. For subsequent trajectory screening (corresponding to step S2), calculate the proportion of the actual associated points of each trajectory and compare it with the set threshold rth, and retain the trajectories with a proportion greater than the threshold as the real moving target trajectories.

[0055] When the number of consecutive missed detection frames reaches this threshold, delete the current trajectory hypothesis. Update the trajectory The trajectory score at the kth moment where represents the score increment, and its calculation formula is:

[0056]

[0057] where i k ≠0 and i k= 0 corresponds to the two cases of trajectory association and trajectory undetected, respectively. Among them, represents the likelihood density function of the observation at time k originating from the trajectory The associated observation is the score contribution of the trajectory D to the trajectory, ln(1 - P ) represents the score contribution of the trajectory i not associated with any observation at time k, and P D is the target detection probability;

[0058] S2. Generate a corresponding weighted undirected graph based on the trajectory hypothesis tree, generate the global optimal hypothesis of each trajectory by solving the maximum weighted independent set of the weighted undirected graph, and according to the global optimal hypothesis solved at the scanning time k, trace back to the node at time k - N 1 , prune the branches deviating from the current optimal hypothesis, confirm the association decision at time k - N 1 + 1, continuously associate the N 2 frame images, and screen the target trajectories by the trajectory length and the proportion of true associated points;

[0059] Specifically, step S2 includes: generating a corresponding weighted undirected graph G=(V, E, W) according to the constructed trajectory hypothesis tree. The vertex vi ∈ V in G is defined as the track node The weight wi ∈ W of the vertex vi is defined as the track score For two incompatible trajectories and the corresponding vertices v i and vj are connected by an edge (i, j) ∈ E; generate the global optimal hypothesis of each trajectory by solving the maximum weighted independent set of the undirected graph. The expression of the maximum weighted independent set is:

[0060]

[0061]

[0062] Among them, is a binary variable, indicating whether the i-th track is included in the optimal solution.

[0063] Set the trajectory length threshold l th and the proportion r th of true associated points, and retain the trajectories and satisfying the conditions as the confirmed trajectories, where represents the trajectory length of the trajectory at time k, Represents the total associated point count in the trajectory, Represents the virtual associated point count, corresponding to the case of missed detection of the trajectory in step S1;

[0064] S3. Statistically confirm the coordinate difference between the trajectories, compare it with the number of detector taps p, and record the trajectory pair and of the satisfied conditions. When the N 3 frame determines that the trajectory pair satisfies this condition, obtain the split point statistical information;

[0065] Specifically, in step S3, record the trajectory pair and of the satisfied conditions. The expression for the satisfied conditions is:

[0066]

[0067] Compare the coordinate difference with the number of detector taps p (the split point spacing corresponds to the number of detector taps), and the unit of p is pixel;

[0068] The coordinates of the trajectory pair are (X i , Y i ) and (X j , Y j ). Continuously judge whether the trajectory pair satisfies this condition for N 3 frames. If not satisfied, it is determined as the trajectory generated by multiple targets; if satisfied, obtain the split point statistical information.

[0069] S4. Construct a target cross-pixel broadening split degradation model. Based on the split point statistical information, determine the rationality of the target cross-pixel broadening split degradation model. When the deviation value is greater than the threshold, perform iterative correction of the target cross-pixel broadening split degradation model, update the model parameters, and re-use them for detecting images.

[0070] Specifically, in step S4, the signal broadening of the target cross-pixel broadening split degradation model is represented by a low-pass filter, and the sampling deviation and filtering bandwidth are set;

[0071] Specifically, in step S4, according to the split point statistical information, map the constructed target cross-pixel broadening split degradation model to the current detection image, calculate the centroid of the model solution, compare the centroid deviation with the ideal centroid of the energy concentration synthesis, and verify the rationality of the target cross-pixel broadening split degradation model by comparing with the set deviation threshold. When the deviation exceeds the threshold, it is unreasonable, perform iterative correction of the target cross-pixel broadening split degradation model, update the model parameters, and update the model parameters, and re-use them for detecting images.

[0072] The present invention conducts single-frame target search and multi-frame association of observations of split-point targets through single-frame morphological filtering and multi-hypothesis association techniques; utilizes the strong correlation between split points to achieve trailing detection and obtain split-point statistical information; constructs a degradation model based on the process of point target "splitting" across pixels and applies it to the detection image, and iteratively updates the model parameters through the energy dimension to achieve centroid correction of targets with cross-pixel broadening, thereby improving the positioning accuracy of high-frame-rate infrared focal plane detectors.

[0073] The present invention is applied to imaging split-point targets of a high-frame-rate infrared focal plane 4-tap detector. The scanning period T is set to 1 s, the target moves at a constant speed of 300 m / s, and the new clutter density λ new = 10 -9 m -2 The false-alarm clutter density λ F = 3×10 -8 m -2 , the detection probability P D = 0.97. For the simulation scenarios of moving targets in two typical horizontal and vertical directions, the multi-hypothesis tracking algorithm is verified. The experimental results show that the algorithm can better achieve independent tracking of multiple observations generated by split-point targets; the degradation model parameters are set with an equivalent bandwidth w = 30 MHz and a sampling deviation k = -0.42. Applying this model to the sequence detection image, the results show that in the two moving directions of horizontal and vertical, the corrected centroid positioning error is stably controlled within 0.3 pixels, and the average positioning accuracy after applying the model is improved from more than 0.5 pixels to within 0.15 pixels. Based on the above experimental results, the technical solution of the present invention is of great significance for improving the multi-target identification ability and positioning accuracy of high-frame-rate infrared focal plane detectors.

[0074] A detection and centroid correction system for targets with cross-pixel broadening, comprising:

[0075] The first unit is used to receive an infrared image frame, establish a potential trajectory hypothesis tree based on suspicious targets, and predict the state of the next frame; receive the next infrared image frame and determine the gating relationship between each observation and the predicted state of each trajectory's current frame.

[0076] The second unit generates a corresponding weighted undirected graph based on the trajectory hypothesis tree, generates the global optimal hypothesis of each trajectory by solving the maximum weighted independent set of the weighted undirected graph, traces back to the node at the k - N 1 moment according to the global optimal hypothesis solved at the scanning moment k, prunes the branches deviating from the current optimal hypothesis, confirms the association decision at the k - N 1 +1 moment, continuously associates the N 2 frame images, and screens the target trajectories by the trajectory length and the proportion of true associated points.

[0077] The third unit is used to statistically confirm the coordinate differences between trajectories, compare them with the number p of detector taps, and record the trajectory pairs and the satisfaction conditions of 3 When the N-frame determines that the trajectory pair satisfies this condition, split point statistical information is obtained;

[0078] The fourth unit is used to construct a target cross-pixel broadening split degradation model. Based on the split point statistical information, the rationality of the target cross-pixel broadening split degradation model is determined. When the deviation value is greater than the threshold, iterative correction of the target cross-pixel broadening split degradation model is performed, and the model parameters are updated and reused for detecting images.

[0079] The system for detecting and centroid correcting a cross-pixel broadening target includes a processor and a memory. The above-mentioned first unit, second unit, third unit, and fourth unit are all stored in the memory as program units, and the corresponding functions are implemented by the processor executing the above program units stored in the memory.

[0080] The processor contains a kernel, and the kernel retrieves the corresponding program units from the memory. One or more kernels can be set. By adjusting the kernel parameters, the observation data of multiple scans are combined through multi-hypothesis tracking technology, and the delayed decision trajectory association can effectively reduce the interlacing of adjacent trajectories and improve the multi-target identification ability of the high-frame-rate infrared focal plane detector; the positioning accuracy of the high-frame-rate infrared focal plane detector is improved by correcting the centroid positioning error.

[0081] The memory may include non-permanent memory in a computer-readable medium, forms such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash memory (flash RAM), and the memory includes at least one memory chip.

[0082] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by the processor of the computer, it enables the computer to execute a method for detecting and centroid correcting a cross-pixel broadening target.

[0083] An embodiment of the present invention provides an electronic device, which is characterized in that it includes: a memory storing a computer program; a processor reading the computer program stored in the memory, and when the processor executes the program, the following steps are implemented:

[0084] S1. Receive an infrared image, establish a potential trajectory hypothesis tree based on suspicious targets, and predict the state of the next frame; receive the next infrared image, and determine the gating relationship between each observation and the predicted state of each trajectory in the current frame;

[0085] S101. Receive an infrared image frame, perform single-frame morphological filtering, enhance the target saliency of the infrared image to obtain an enhanced image, and perform threshold segmentation on the enhanced image to screen out suspicious targets;

[0086] S102. Establish a potential trajectory hypothesis tree for each suspicious target, initialize the trajectory state, construct a Kalman filter, predict the state of the next frame of the suspicious target, and initialize the trajectory score;

[0087] S103. Receive a new frame of detection image, determine the gating relationship between each observation and the predicted state of each trajectory in the current frame, perform trajectory association on the observations within the association gate of a certain trajectory; for the observations not within any trajectory association gate, perform trajectory initiation, and re-execute step S102;

[0088] S2. Generate a corresponding weighted undirected graph based on the trajectory hypothesis tree, generate the globally optimal hypothesis for each trajectory by solving the maximum weighted independent set of the weighted undirected graph, according to the globally optimal hypothesis solved at the scanning time k, trace back to the node at the k - N 1 moment, prune the branches deviating from the current optimal hypothesis, confirm the association decision at the k - N 1 +1 moment, continuously associate the N 2 frame images, and screen the target trajectories by the ratio of the trajectory length to the proportion of true association points;

[0089] S3. Statistically confirm the coordinate differences between the trajectories, compare with the number p of detector taps, record the conditions satisfied by the trajectory pairs and , when the N 3 frame determines that the trajectory pair satisfies this condition, obtain the split point statistical information;

[0090] S4. Construct a target cross-pixel broadening split degradation model, based on the split point statistical information, determine the rationality of the target cross-pixel broadening split degradation model, when the deviation value is greater than the threshold, perform iterative correction of the target cross-pixel broadening split degradation model, update the model parameters, and re-use them for detecting images. The device in this article can be a server, a PC, a PAD, a mobile phone, etc.

[0091] The embodiment of the present invention provides a computer program product, characterized in that the computer program product includes a computer program, and the computer program is adapted to be executed by a processor to initialize a program with the following method steps:

[0092] S1. Receive an infrared image frame, establish a potential trajectory hypothesis tree based on suspicious targets, and predict the state of the next frame; receive the next frame of infrared image, and determine the gating relationship between each observation and the predicted state of each trajectory in the current frame;

[0093] S101. Receive an infrared image frame, perform single-frame morphological filtering, enhance the target saliency of the infrared image to obtain an enhanced image, and perform threshold segmentation on the enhanced image to screen out suspicious targets;

[0094] S102. Establish a potential trajectory hypothesis tree for each suspicious target, initialize the trajectory state, construct a Kalman filter, predict the state of the next frame of the suspicious target, and initialize the trajectory score;

[0095] S103. Receive a new frame of detection image, determine the gating relationship between each observation and the predicted state of the current frame of each trajectory, perform trajectory association on the observations within the association gate of a certain trajectory; for the observations not within any trajectory association gate, perform trajectory initiation, and re-execute step S102;

[0096] S2. Generate a corresponding weighted undirected graph based on the trajectory hypothesis tree, generate the globally optimal hypothesis for each trajectory by solving the maximum weighted independent set of the weighted undirected graph, according to the globally optimal hypothesis solved at the scanning time k, trace back to the node at the k - N 1 moment, prune the branches deviating from the current optimal hypothesis, confirm the association decision at the k - N 1 +1 moment, continuously associate the N 2 frame images, and screen the target trajectories through the ratio of the trajectory length to the proportion of true association points;

[0097] S3. Statistically confirm the coordinate differences between the trajectories, compare them with the number p of detector taps, record the pairs of trajectories and the conditions that are met. When the N 3 frame determines that the pair of trajectories meets this condition, obtain the split point statistical information;

[0098] S4. Construct a target cross-pixel broadening, splitting, and degradation model. Based on the split point statistical information, determine the rationality of the target cross-pixel broadening, splitting, and degradation model. When the deviation value is greater than the threshold, perform iterative correction of the target cross-pixel broadening, splitting, and degradation model, update the model parameters, and re-use them for detecting images. The device in this article can be a server, a PC, a PAD, a mobile phone, etc.

[0099] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0100] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It should be understood that each flow and / or block in the flowchart and / or block diagram, and combinations of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processors of general purpose computers, special purpose computers, embedded processors, or other programmable data processing devices to produce a machine, such that the instructions executed by the processors of the computer or other programmable data processing devices generate means for implementing the functions specified in the flow Figure 1 one or more flows and / or blocks Figure 1 or means for implementing the functions specified in a block or multiple blocks.

[0101] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory produce a manufactured article including instruction means that implement the functions specified in the flow Figure 1 one or more flows and / or blocks Figure 1 or means for implementing the functions specified in a block or multiple blocks.

[0102] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in the flow Figure 1 one or more flows and / or blocks Figure 1 or means for implementing the functions specified in a block or multiple blocks.

[0103] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.

[0104] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM), and / or non-volatile memory such as read-only memory (ROM) or flash memory. The memory is an example of computer-readable media.

[0105] A computer-readable medium includes both permanent and non-permanent, removable and non-removable media and can implement information storage by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transitory medium that can be used to store information accessible by a computing device. As defined herein, a computer-readable medium does not include transitory computer-readable media such as modulated data signals and carrier waves.

[0106] It should also be noted that the term "comprising", "including" or any other variation thereof is intended to cover non-exclusive inclusion, such that a process, method, article or apparatus comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or apparatus. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or apparatus comprising the element.

[0107] The above are only embodiments of the present application and are not used to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims

1. A method for detecting and centroid correction of cross-pixel broadening targets, characterized in that, it includes the following steps: S1. Receive an infrared image frame, establish a potential trajectory hypothesis tree based on suspicious targets, and predict the state of the next frame; Receive the next infrared image frame, and determine the gating relationship between each observation and the predicted state of each trajectory in the current frame; S2. Generate a corresponding weighted undirected graph based on the trajectory hypothesis tree, generate the global optimal hypothesis of each trajectory by solving the maximum weighted independent set of the weighted undirected graph, and according to the global optimal hypothesis solved at the scanning time k, trace back to the node at the k - N 1 moment, prune the branches deviating from the current optimal hypothesis, and confirm the association decision at the k - N 1 +1 moment, continuously associate the N 2 frames of images, and screen the target trajectories by the trajectory length and the proportion of true associated points; S3. Statistically confirm the coordinate differences between the trajectories, compare them with the number of detector taps p, and record the trajectory pairs and the satisfaction conditions. When the N 3 frame determines that the trajectory pair satisfies this condition, obtain the split point statistical information; S4. Construct a target cross-pixel broadening splitting and degradation model, based on the splitting point statistical information, determine the rationality of the target cross-pixel broadening splitting and degradation model. When the deviation value is greater than the threshold, perform iterative correction of the target cross-pixel broadening splitting and degradation model, update the model parameters, and re-use them for detecting images.

2. The method for detecting and centroid correction of cross-pixel broadening targets according to claim 1, characterized in that, the step S1 includes: S101. Receive an infrared image frame, perform single-frame morphological filtering, enhance the target saliency of the infrared image to obtain an enhanced image, perform threshold segmentation on the enhanced image, and screen out suspicious targets; S102. Establish a potential trajectory hypothesis tree for each suspicious target, initialize the trajectory state, construct a Kalman filter, predict the state of the next frame of the suspicious target, and initialize the trajectory score; S103. Receive a new frame of detection image, determine the gating relationship between each observation and the predicted state of each trajectory in the current frame. For the observations within the correlation gating of a certain trajectory, perform trajectory association; for the observations not within any trajectory correlation gating, perform trajectory initiation, and re-execute the step S102.

3. The method for detecting and centroid correction of cross-pixel broadening targets according to claim 1, characterized in that, The said step S2 includes: generating a corresponding weighted undirected graph G=(V, E, W) according to the constructed trajectory hypothesis tree, where the vertex vi∈V in G is defined as a track node The weight wi∈W of the vertex vi is defined as the track score For two incompatible trajectories and the corresponding vertices v i and vj are connected by an edge (i, j)∈E; generating the global optimal hypothesis of each trajectory by solving the maximum weighted independent set of the undirected graph, and the expression of the maximum weighted independent set is: Among them, is a binary variable indicating whether the i-th track is included in the optimal solution.

4. The method for detecting and centroid correction of cross-pixel broadening targets according to claim 1, characterized in that, In the step S3, the recording track pair and meet the conditions, and the expression for representing the conditions met is: The coordinates of the trajectory pair are respectively (X i , Y i ) and (X j , Y j ). Continuously judge whether the trajectory pair satisfies this condition for N 3 frames. If it does not satisfy, it is determined as the trajectory generated by multiple targets; if it satisfies, the split point statistical information is obtained.

5. The method for detecting and centroid correction of cross-pixel broadening targets according to claim 1, characterized in that, In the step S4, the signal broadening of the target cross-pixel broadening splitting and degradation model is represented by a low-pass filter, and the sampling deviation and the filtering bandwidth are set.

6. The method for detecting and centroid correction of cross-pixel broadening targets according to claim 5, characterized in that, In the step S4, according to the splitting point statistical information, map the constructed target cross-pixel broadening splitting and degradation model to the current detection image, calculate the centroid solved by the model, and compare the centroid deviation with the ideal centroid of the energy accumulation synthesis. By comparing with the set deviation threshold, verify the rationality of the target cross-pixel broadening splitting and degradation model. When the deviation exceeds the threshold, it is unreasonable, perform iterative correction of the target cross-pixel broadening splitting and degradation model, update the model parameters, and update the model parameters, and re-use them for detecting images.

7. A system for detecting and centroid correction of cross-pixel broadening targets, characterized in that, it includes: The first unit, which is used to receive an infrared image frame, establish a potential trajectory hypothesis tree based on suspicious targets, and predict the state of the next frame; Receive the next infrared image frame, and determine the gating relationship between each observation and the predicted state of each trajectory in the current frame; The second unit, which generates a corresponding weighted undirected graph based on the trajectory hypothesis tree, generates the global optimal hypothesis for each trajectory by solving the maximum weighted independent set of the weighted undirected graph, and traces back to the node at the k - N 1 moment according to the global optimal hypothesis solved at the scanning moment k, prunes the branches deviating from the current optimal hypothesis, and confirms the association decision at the k - N 1 +1 moment, continuously associates the N 2 frames of images, and screens the target trajectories by the trajectory length and the proportion of true associated points; The third unit is used to statistically confirm the coordinate differences between the trajectories, compare them with the number p of detector taps, record the pairs of trajectories and that meet the conditions. When the N 3 frame determines that the pair of trajectories meets this condition, split point statistical information is obtained; The fourth unit is configured to construct a target cross-pixel broadening splitting degradation model, determine the rationality of the target cross-pixel broadening splitting degradation model based on the splitting point statistical information. When the deviation value is greater than the threshold, iterative correction of the target cross-pixel broadening splitting degradation model is performed, and the model parameters are updated and reused for image detection.

8. A computer-readable storage medium, characterized in that, a computer program is stored thereon, and when the computer program is executed by a processor of the computer, the computer is caused to execute the method for detecting and centroid correction of a cross-pixel broadening target according to any one of claims 1 to 6.

9. An electronic device, characterized in that, comprising: a memory storing a computer program; a processor configured to read the computer program stored in the memory to execute the method for detecting and centroid correction of a cross-pixel broadening target according to any one of claims 1 to 6.

10. A computer program product, characterized in that, the computer program product includes a computer program, and the computer program is executed by a processor to implement the method for detecting and centroid correction of a cross-pixel broadening target according to any one of claims 1 to 6.

Images