A method and system for processing multi-target tracks of infrared panoramic images based on LMB

Abstract

The present invention relates to a multi-target track processing method and system for infrared panoramic images based on LMB (labeled multi-Bernoulli algorithm), the method comprising: dividing the infrared panoramic image into sectors, reading measurement tracks according to each sector, and using SVM to classify the tracks; establishing a track list, including new tracks and existing (surviving) tracks after LMB prediction filtering; updating the tracks under different association conditions across adjacent sectors for tracks falling into any track gate based on LMB; calculating track weights, pruning and merging the tracks, outputting the track numbers if there are already track numbers; starting the tracks and compiling them if there are no track numbers; and regenerating the input tracks that fail to successfully associate with the tracks, for use in establishing a track list in the next cycle. The multi-target tracking algorithm LMB-SVM can significantly reduce false alarms while significantly improving the track initiation and track maintenance capabilities.

Description

Technical Field

[0001] The present invention relates to the field of optoelectronic system information processing, and in particular to a LMB-based infrared panoramic image multi-target track processing method and system. Background Art

[0002] Infrared detection equipment is widely used in military and civilian fields. Infrared detection equipment mainly detects images through infrared, extracts infrared target points, and processes the points to output stable track information. The multi-target track processing algorithm directly affects the detection accuracy of infrared detection equipment and plays a vital role. However, the existing multi-target tracking algorithm of infrared detection equipment has the following problems: (1) It is difficult to start the track of targets that appear and disappear from time to time; (2) The track batches of intersecting targets are interchangeable; (3) The target tracking anti-interference ability is insufficient.

[0003] There are many reasons why the infrared detection and tracking capabilities of weak targets are insufficient at this stage:

[0004] (1) Deficiencies of the existing information processing framework for small and weak target detection and tracking

[0005] The existing information processing framework for small target detection and tracking is to detect first and then track. This information processing framework is a one-way transmission of information, and information cannot flow in the opposite direction. The detection stage only uses the limited information of a single frame instead of multi-frame information, which makes the success rate of target judgment low, causing multiple false alarms in complex backgrounds and the problem of missed detection of small targets with low signal-to-noise ratio.

[0006] (2) Insufficient single-frame detection capability

[0007] The lack of single-frame detection capability is mainly reflected in two aspects: insufficient background suppression capability and insufficient target screening capability. In the strong clutter background, in order to achieve the purpose of extracting long-range weak targets, the segmentation threshold must be greatly reduced. When the background suppression capability is insufficient, a large amount of clutter and weak targets are segmented at the same time. The target screening capability of the original technology is limited, and it is impossible to identify weak targets from a large amount of clutter, resulting in false alarms and missed detections.

[0008] (3) Insufficient target tracking capabilities

[0009] Our existing multi-target track processing uses the nearest neighbor multi-target allocation and association method, which has the disadvantage of track combination explosion. Under strong clutter, only some of the detected suspected targets can be effectively associated. In the track association process, only the coordinate information of the target is used, without the characteristic information of the target, resulting in insufficient anti-interference ability. In the track extraction process, only the uniform motion model is considered, and the adaptability to strong maneuvering targets is insufficient. The corresponding space between the track and the batch number is not effectively established, which is affected by the change of heading angle and the shaking of the ship. The track prediction capability is insufficient, and the ability to handle the occasional target is limited, resulting in the instability of the track batch number. Summary of the invention

[0010] The present invention aims at the technical problems existing in the prior art and provides a method and system for processing multi-target tracks of infrared panoramic images based on LMB. The method and system for processing multi-target tracks of infrared panoramic images based on LMB provided by the present invention divide the infrared panoramic image into sectors, send point tracks by sector, and consider the information of multiple adjacent sectors for track initiation and track update, so as to solve the problem of time asynchrony caused by the long scanning cycle of infrared detection equipment; use SVM for pre-classification processing, and use standard gating technology to divide the LMB form multi-target prediction density function into different groups before continuing the track, so as to make the track continue. The computational complexity of the subsequent process is significantly reduced; the multi-target tracking algorithm based on labeled multi-Bernoulli has better track initiation and track maintenance capabilities; with the support of SVM classifier, the number of false tracks generated by cloud background can be greatly suppressed; the complexity of the multi-target tracking algorithm based on labeled multi-Bernoulli is higher than that of the nearest neighbor association algorithm. Under appropriate clutter conditions, the required computing time is about twice that of the nearest neighbor association algorithm, and it has the possibility of engineering application; the multi-target tracking algorithm LMB-SVM, which is composed of labeled multi-Bernoulli filter and SVM classifier, can greatly reduce false alarms while greatly improving the track initiation and track maintenance capabilities.

[0011] According to a first aspect of the present invention, a method for processing multi-target tracks of an infrared panoramic image based on LMB is provided, comprising: step 1, dividing the infrared panoramic image into sectors, reading measurement points according to each sector, and classifying the points using SVM;

[0012] Step 2, establish a track list, including new point tracks and surviving tracks after LMB prediction filtering;

[0013] Step 3, grouping and processing the established track list based on standard gating technology; executing step 4 for the points that do not fall into any track gate, and starting or updating the tracks across adjacent sectors based on LMB for the points that fall into any track gate; and updating the surviving tracks in the track list in the next cycle based on the started or updated tracks.

[0014] Step 4: The input points that have not been successfully associated with the track are regenerated. If there is a new point, it is used to establish the new point in the track list in the next cycle.

[0015] Based on the above technical solution, the present invention can also make the following improvements.

[0016] Optionally, the process of starting a new track across adjacent sectors based on LMB in step 3 includes:

[0017] When the current sector number of the trace is N+1 and the current scanning cycle number is M+1, the trace of the Nth sector in the Mth scanning cycle and the traces of the N-1th, Nth and N+1th sectors in the M+1th scanning cycle are collected;

[0018] The target is screened for LMB filtering and updating. When it is determined that the point track is within the track starting gate, the target is batched and the track is output.

[0019] Optionally, in step 3, comparing the point track with the predicted track of the track based on the standard gating technology and performing grouping processing includes:

[0020] The predicted coordinates of the track in the predicted track are taken out in sequence, step 4 is executed for the track points that do not fall into any track gate, and the track starting or updating across adjacent sectors is performed based on LMB for the track points that fall into any track gate.

[0021] Optionally, the process of updating the track across adjacent sectors based on LMB for the point track falling into any track gate in step 4 includes:

[0022] When the current sector number of the trace is N+1 and the current scanning cycle number is M+1, the traces of the sectors N-1, N and N+1 of the M+1th scanning cycle are collected;

[0023] When the point track data of the first sector of any M+1th scanning cycle is collected, the track of the Mth scanning cycle is predicted by LMB filtering;

[0024] The point tracks of the M+1th scanning cycle are grouped according to the track information of the Mth scanning cycle to generate various track groups, and the tracks of each track group are subjected to LMB filtering and updating processing.

[0025] Optionally, performing LMB filtering and updating processing on the tracks of each track grouping includes: performing LMB filtering and updating on the tracks in different association situations, and calculating track weights in different association situations.

[0026] Optionally, after performing LMB filtering update processing on the tracks of each track grouping, the method further includes: normalizing the track weights of each track grouping, clipping and merging the updated tracks, and then outputting the updated tracks.

[0027] Optionally, the clipping of the updated track includes: clipping the track whose weight value is less than a set threshold.

[0028] According to a second aspect of the present invention, there is provided an infrared panoramic image multi-target track processing system based on LMB, comprising: a point track classification module, a track link list establishment module, a point track generation module, a track start module and a track continuation module;

[0029] The point trace classification module is used to divide the infrared panoramic image into sectors, read the measurement point traces according to each sector, and classify the point traces using SVM;

[0030] The track list establishment module is used to establish a track list, determine whether there is a new point track, and store the new point track in the track list; determine whether there is a surviving track, if so, perform LMB prediction filtering on the surviving track, and save the surviving track that has passed the prediction filtering in the track list summary.

[0031] The point track regeneration module performs point track regeneration on the points that cannot be associated with the track chain list, so as to enter the track chain list in the next cycle;

[0032] The track initiation module is used to compare the point track with the predicted track of the track based on the standard gating technology and then perform grouping processing; execute step 4 for the point track that does not fall into any track wave gate, and start the track across adjacent sectors based on LMB for the point track that falls into any track wave gate; update the track chain list based on the track after initiation, and batch the track.

[0033] The track continuation module is used to compare the point track with the predicted track of the track based on the standard gating technology and then perform grouping processing; execute step 4 for the point track that does not fall into any track gate, and update the track across adjacent sectors based on LMB for the point track that falls into any track gate; update the track list based on the updated track. If the track is not labeled, a new batch number is given to the track, and the track continues successfully. The original batch number is kept unchanged.

[0034] According to a third aspect of the present invention, there is provided an electronic device comprising a memory and a processor, wherein the processor is used to implement the steps of a method for processing multi-target tracks of infrared panoramic images based on LMB when executing a computer management program stored in the memory.

[0035] According to a fourth aspect of the present invention, there is provided a computer-readable storage medium having a computer management program stored thereon, wherein when the computer management program is executed by a processor, the steps of a method for processing multi-target tracks of infrared panoramic images based on LMB are implemented.

[0036] The present invention provides a multi-target track processing method, system, electronic device and storage medium for infrared panoramic images based on LMB, which establishes the state of weak targets and sensor measurements in complex scenes as a random finite set model, and uses the random finite set statistics theory to perform multi-target tracking task processing; random finite set statistics provides a theoretical optimal solution and an engineering implementation method for the multi-target track processing task in infrared panoramic images; the multi-target track processing method for infrared panoramic images based on label multi-Bernoulli filtering does not introduce data association, avoids track combination explosion, has low computational complexity, is engineering-friendly, can naturally complete the start, maintenance and termination of the track, can realize strong maneuvering target tracking by combining with interactive multi-models, and can realize detection and tracking of extended targets such as missile tail flames and blurred ships by adding an extended target model, which can effectively solve the problems faced by infrared detection multi-target track processing; two-frame navigation, anti-interference and motion noise suppression are realized simultaneously, which greatly improves the detection capability of infrared detection equipment; the anti-interference success rate of weak target tracks is increased by more than 50% compared with the original technology; the navigation success rate of weak targets with only 50% detection probability is improved by more than 50%; The power is increased by more than 45%. To solve the problem of asynchronous measurement data, the 360° image area scanned by the infrared detection equipment is evenly divided into 16 sectors, and the track information is sent by sector. At the same time, processing by sector brings another problem. The target is in the Nth sector during the Mth circle, and the target may fly to the N-1 or N+1 sector during the M+1th circle. This makes data processing difficult. Therefore, when designing the algorithm, multiple adjacent sector information is considered for track initiation and track update. When inputting the measurement point track data, SVM is first used for pre-classification. processing; at the same time, before continuing the track, the standard gating technology is used to divide the LMB form multi-target prediction density function into different groups; each group of measurement points is adjacent to the track space position; group processing significantly reduces the computational complexity of the track continuation process; consider introducing the multi-model idea into the labeled multi-Bernoulli filter, and use a combination of multiple motion models to handle the multi-target track processing problem of infrared detection equipment; using the "label technology", the Gaussian component inherits the corresponding historical track label in each iteration, and the established stable track is sent out according to the batch number. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 A flowchart of multi-target track processing of infrared panoramic images based on LMB provided by the present invention;

[0038] Figure 2A flowchart of establishing a track link list based on LMB cross-sector processing of infrared panoramic images provided by an embodiment of the present invention;

[0039] Figure 3 A flowchart of starting a new track based on LMB cross-sector processing of infrared panoramic images provided by an embodiment of the present invention;

[0040] Figure 4 A flow chart of track continuation based on LMB cross-sector processing of infrared panoramic images provided by an embodiment of the present invention;

[0041] Figure 5 A structural block diagram of an infrared panoramic image multi-target track processing system based on LMB provided by the present invention;

[0042] Figure 6 A schematic diagram of the hardware structure of a possible electronic device provided by the present invention;

[0043] Figure 7 A schematic diagram of the hardware structure of a possible computer-readable storage medium provided by the present invention. DETAILED DESCRIPTION

[0044] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0045] Mahler proposed to apply the theory of random finite sets to the field of multi-target tracking. Random finite sets can identify the state model and observation model of the target in the form of random sets; it is based on the Bayesian filtering theory and directly estimates the number and state of the target. In recent years, scholars have conducted in-depth research in the field of random finite sets. Among them, the more classic algorithms are the probability hypothesis density filter algorithm (PHD) and the potential probability hypothesis density filter algorithm (CPHD). The label multi-Bernoulli algorithm LMB can more accurately estimate the target's track information and shows obvious advantages. LMB is mainly divided into the following two parts: prediction process and update process.

[0046] (1) Prediction process

[0047] In the prediction process, it is assumed that both the prior density and the new target density are LMB, based on the state space and label space The prior density of is expressed by the parameter set as Based on state space and label space The new target density can be expressed as the parameter set The label space of the new target and the label space of the surviving target must be different, that is,

[0048] If the prior density and the new target density are given as above, then based on the target state space and label space The predicted density parameter set for is:

[0049]

[0050] Among them, the existence probability and spatial distribution of the surviving target can be calculated by the following formula:

[0051]

[0052] η s (l)= <p s (·,l),p(·,l)> Formula 4

[0053] In the above formula, the survival probability of trajectory l is p s (·, l), f(x|·, l) is the single target Markov transition probability. The existence probability and spatial distribution of the new target are given according to the new target model. The prediction process is basically the same as the traditional multi-Bernoulli filter.

[0054] (2) Update process

[0055] In space The results after the above prediction are uniformly written in the form of the following LMB parameter set:

[0056] In order to achieve measurement update, the predicted LMB needs to be converted into the equivalent δ-GLMB form:

[0057]

[0058]

[0059] In space The LMB RFS of the updated multi-objective posterior density can be approximated by the following parameter set:

[0060] The existence probability and spatial distribution of a single trajectory after update can be calculated by the following formula:

[0061]

[0062]

[0063] in:

[0064]

[0065] Indicates track I+ When the point distribution θ is satisfied, the density function of its existence probability is: (θ) (x,l|Z) represents the track space distribution function when the point track is assigned θ.

[0066] Figure 1 A flowchart of a multi-target track processing method for infrared panoramic images based on LMB is provided in an embodiment of the present invention, such as Figure 1 As shown, the multi-target track processing method includes: step 1, dividing the infrared panoramic image into sectors, reading measurement points according to each sector, and using SVM (Support Vector Machine, SVM, support vector machine) to classify the points.

[0067] In a specific implementation, the 360° image area scanned by the infrared detection device can be evenly divided into 16 sectors.

[0068] Step 2: Create a track list, including new track points and surviving tracks after LMB prediction filtering

[0069] Step 3, grouping and processing the established track list based on standard gating technology; executing step 4 for the points that do not fall into any track gate, and updating the tracks across adjacent sectors based on LMB for the points that fall into any track gate; and updating the surviving tracks in the track list in the next cycle based on the updated tracks.

[0070] Step 4: Create new points for the input points that have not been successfully associated with the track. If there are new points, they are used to establish new points in the track list in the next cycle.

[0071] The invention provides an infrared panoramic image multi-target track processing method based on LMB, which divides the infrared panoramic image into sectors, sends point tracks by sector, and considers the information of multiple adjacent sectors to start and update the track, so as to solve the problem of time asynchronism caused by the long scanning cycle of infrared detection equipment; uses SVM for pre-classification processing, and before continuing the track, uses standard gating technology to divide the LMB form multi-target prediction density function into different groups, so that the calculation complexity of the track continuing process is significantly reduced; the multi-target tracking algorithm based on label multi-Bernoulli has better track starting and track maintenance capabilities; with the support of SVM classifier, the number of false tracks generated by cloud background can be greatly suppressed; the complexity of the multi-target tracking algorithm based on label multi-Bernoulli is higher than that of the nearest neighbor association algorithm, and under appropriate clutter conditions, the required operation time is about twice that of the nearest neighbor association algorithm, so as to have the possibility of engineering application; the multi-target tracking algorithm LMB-SVM composed of label multi-Bernoulli filter and SVM classifier can greatly reduce false alarms while greatly improving the track starting and track maintenance capabilities.

[0072] Example 1

[0073] Embodiment 1 of the present invention is an embodiment of a multi-target track processing method for infrared panoramic images based on LMB provided by the present invention, combined with Figure 1 It can be seen that the embodiment of the multi-target track processing method includes:

[0074] Step 1: Divide the infrared panoramic image into sectors, read the measurement points according to each sector, and use SVM to classify the points.

[0075] Step 2: Create a track list, including new track points and surviving tracks after LMB prediction filtering

[0076] Step 3, grouping and processing the established track list based on standard gating technology; executing step 4 for the track points that do not fall into any track gate, and starting or updating the track across adjacent sectors based on LMB for the track points that fall into any track gate; and updating the surviving tracks in the track list in the next cycle based on the updated tracks.

[0077] Step 4: The input points that have not been successfully associated with the track are regenerated. If there is a new point, it is used to establish the new point in the track list in the next cycle.

[0078] When starting and continuing the infrared panoramic image track, it is necessary to focus on the cross-sector processing of the target. Figure 2 A flowchart of establishing a track link list based on LMB cross-sector processing of infrared panoramic images provided by an embodiment of the present invention; Figure 3 FIG. 1 is a flowchart of a method for processing the initiation of a new track of an infrared panoramic image based on LMB cross-sector processing according to an embodiment of the present invention, combined with FIG. Figure 1 , Figure 2 and Figure 3 It can be seen that, in a possible implementation manner, the process of starting a new track across adjacent sectors based on LMB in step 3 includes:

[0079] When the current sector number of the trace is N+1 and the current scanning cycle number is M+1, the trace of the Nth sector in the Mth scanning cycle and the traces of the N-1th, Nth and N+1th sectors in the M+1th scanning cycle are collected.

[0080] The target is screened for LMB filtering and updating. When it is determined that the point track is within the track starting gate, the target is batched and the track is output.

[0081] When the current sector number is N+1, the Mth scanning cycle collects the N sector traces after SVM classification processing; in the M+1th scanning cycle, the N-1, N, N+1 sector traces are collected; the Mth scanning cycle is filtered and updated; if the trace is in the starting wave gate, the trace is batched; the track is output; the new track is started successfully.

[0082] Step 4, compare the point track with the predicted track of the track based on the standard gating technology and then perform grouping processing; execute step 3 for the point track that does not fall into any track gate, and update the track across adjacent sectors based on LMB for the point track that falls into any track gate; update the track list based on the updated track.

[0083] like Figure 4 A flowchart of a track continuation based on LMB cross-sector processing of infrared panoramic images provided by an embodiment of the present invention, combined with Figure 1 , Figure 2 and Figure 4 It can be seen that in a possible embodiment, the grouping process of comparing the point track with the predicted track of the track based on the standard gating technology in step 4 includes:

[0084] Take out the predicted coordinates of the track in the predicted track in turn, execute step 3 for the point tracks that do not fall into any track gate, and update the track across adjacent sectors based on LMB for the point tracks that fall into any track gate.

[0085] In a possible embodiment, the process of updating the track across adjacent sectors based on LMB for the point track falling into any track gate in step 4 includes:

[0086] When the current sector number of the trace is N+1 and the current scanning cycle number is M+1, the traces of the sectors N-1, N and N+1 of the M+1th scanning cycle are collected.

[0087] When the point track data of the first sector of any M+1th scanning cycle is collected, LMB filtering prediction is performed on the track of the Mth scanning cycle.

[0088] According to the track information of the Mth scanning cycle, the point tracks of the M+1th scanning cycle are grouped to generate various track groups, and the tracks of each track group are updated by LMB filtering.

[0089] In a possible implementation manner, performing LMB filtering and updating processing on the tracks of each track group includes: performing LMB filtering and updating on the tracks in different association situations, and calculating the track weights in different association situations.

[0090] In a possible implementation manner, after performing LMB filtering update processing on the tracks of each track group, the method further includes: normalizing the track weights of each track group, clipping and merging the updated tracks, and then outputting the updated tracks.

[0091] In a possible implementation manner, clipping the updated track includes: clipping the track whose weight value is less than a set threshold.

[0092] In an embodiment of a method for processing multi-target tracks of infrared panoramic images based on LMB provided by the present invention, the real position motion characteristics of the target and the position noise error of the target in the infrared image are fully considered; in the process of multi-target track processing, the historical position information of the multiple targets is fully utilized, while the possible situations of multiple tracks are retained, and a stronger anti-interference ability is possessed; theoretically, the problems of track intersection, target track batch number interchange and insufficient target tracking anti-interference ability in the existing track association algorithm are effectively solved; at the same time, the algorithm designed in this article does not rigidly require three frames to start, and has a stronger navigation capability for targets that appear and disappear from time to time. In specific implementation, the following problems exist when applying random finite sets to the actual multi-target track processing problem of infrared panoramic images:

[0093] 1) The measurement data is not synchronized;

[0094] The sensor data processed by the multi-target tracking algorithm based on multi-Bernoulli filtering are all assumed to be time synchronized; however, in fact, the infrared detection equipment has a long scanning cycle, which often leads to time asynchrony, which objectively reduces the output track accuracy.

[0095] In order to solve the problem of asynchronous measurement data, the 360° image area scanned by the infrared detection equipment is evenly divided into 16 sectors, and the track information is sent by sector. At the same time, processing by sector brings another problem: the target is in the Nth sector during the Mth circle, and in the M+1th circle, the target may fly to the N-1 or N+1 sector, which makes data processing difficult. Therefore, when designing the algorithm implementation, multiple adjacent sector information is considered for track initiation and track update.

[0096] 2) The large number of infrared targets puts great pressure on algorithm calculation.

[0097] Under the complex background of sea and sky, the extracted infrared targets are dense and numerous, which brings great difficulties to the algorithm. In view of this problem, we first use SVM for pre-classification when inputting the measurement point data; at the same time, before continuing the track, we use standard gating technology to divide the LMB form multi-target prediction density function into different groups; each group of measurement points is adjacent to the track space position. Grouping processing significantly reduces the computational complexity of the track continuation process.

[0098] 3) The maneuverability is not strong.

[0099] At present, the existing infrared detection equipment multi-target track processing algorithm uses binary extrapolation method, which reduces the practicality of the algorithm when dealing with maneuvering targets. This algorithm considers introducing the multi-model idea into the label multi-Bernoulli filter and uses a combination of multiple motion models to deal with the infrared detection equipment multi-target track processing problem.

[0100] 4) Uncertainty of target motion model, obtaining track estimation of multiple maneuvering targets and target track batching problem;

[0101] The last important step of infrared detection equipment data processing is to send out the established stable track according to the batch number. This paper adopts the "label technology", and the Gaussian component inherits the corresponding historical track label in each iteration.

[0102] Example 2

[0103] Embodiment 2 provided by the present invention is an embodiment of an infrared panoramic image multi-target track processing system based on LMB provided by the present invention. Figure 5 The structure diagram of a multi-target track processing system for infrared panoramic images based on LMB provided by the embodiment of the present invention is combined with Figure 5 It can be seen that this embodiment includes:

[0104] Point track classification module, track list establishment module, point track generation module, track start module and track continuation module.

[0105] The point trace classification module is used to divide the infrared panoramic image into sectors, read the measurement points according to each sector, and use SVM to classify the points.

[0106] The track list establishment module includes new point tracks and surviving tracks after LMB prediction filtering.

[0107] The track regeneration module performs track regeneration across adjacent sectors based on LMB. If the input track is not successfully associated with the surviving track, the track will be regenerated.

[0108] The track initiation module is used to compare the point track with the predicted track based on the standard gating technology and then perform grouping processing; execute step 3 for the point track that does not fall into any track wave gate, and start the track across adjacent sectors based on LMB for the point track that falls into any track wave gate; and batch the track; update the track list based on the track after initiation.

[0109] The track continuation module is used to compare the point track with the predicted track of the track based on the standard gating technology and then perform grouping processing; execute step 3 for the point track that does not fall into any track wave gate, and update the track across adjacent sectors based on LMB for the point track that falls into any track wave gate; update the track list based on the updated track.

[0110] It can be understood that the infrared panoramic image multi-target track processing system based on LMB provided by the present invention corresponds to the infrared panoramic image multi-target track processing method based on LMB provided in the aforementioned embodiments. The relevant technical features of the infrared panoramic image multi-target track processing system based on LMB can refer to the relevant technical features of the infrared panoramic image multi-target track processing method based on LMB, which will not be repeated here.

[0111] See also Figure 6 , Figure 6 Schematic diagram of an electronic device provided by an embodiment of the present invention. Figure 6 As shown, an embodiment of the present invention provides an electronic device, including a memory 1310, a processor 1320, and a computer program 1311 stored in the memory 1310 and executable on the processor 1320. When the processor 1320 executes the computer program 1311, the following steps are implemented: Step 1, the infrared panoramic image is divided into sectors, the measurement point traces are read according to each sector, and the point traces are classified using SVM; Step 2, a track list is established to determine whether there is a track of the point trace in the track list; if yes, step 3 is executed; otherwise, step 4 is executed; Step 3, a new track across adjacent sectors is started based on LMB, and the newly generated track is added to the track list; Step 4, the point trace is compared with the predicted track of the track based on the standard gating technology and then grouped; Step 3 is executed for the point trace that does not fall into any track wave gate, and the track across adjacent sectors is updated based on LMB for the point trace that falls into any track wave gate; and the track list is updated based on the updated track.

[0112] See also Figure 7 , Figure 7 Schematic diagram of an embodiment of a computer-readable storage medium provided by the present invention. Figure 7 As shown, this embodiment provides a computer-readable storage medium 1400, on which a computer program 1411 is stored. When the computer program 1411 is executed by a processor, the following steps are implemented:

[0113] Step 1, divide the infrared panoramic image into sectors, read the measurement points according to each sector, and use SVM to classify the points; Step 2, establish a track list, and determine whether there is a track of the point in the track list; if yes, execute step 3; otherwise, execute step 4; Step 3, start a new track across adjacent sectors based on LMB, and add the newly generated track to the track list; Step 4, compare the point track with the predicted track based on the standard gating technology and perform grouping processing; Execute step 3 for the point track that does not fall into any track gate, and update the track across adjacent sectors based on LMB for the point track that falls into any track gate; Update the track list based on the updated track.

[0114] The embodiment of the present invention provides a method, system, electronic device and storage medium for processing multi-target tracks of infrared panoramic images based on LMB, which establishes the state of weak targets and sensor measurements in complex scenes as a random finite set model, and uses the theory of random finite set statistics to process the tracking task of multiple targets; random finite set statistics provides a theoretical optimal solution and an engineering implementation method for the multi-target track processing task in infrared panoramic images; the multi-target track processing method for infrared panoramic images based on label multi-Bernoulli filtering does not introduce data association, avoids track combination explosion, has low computational complexity, is engineering-friendly, can naturally complete the start, maintenance and termination of the track, and can realize strong maneuvering target tracking by combining with interactive multi-models; adding an extended target model can realize the detection and tracking of extended targets such as missile tail flames and blurred ships, which can effectively solve the problems faced by infrared detection multi-target track processing; two-frame navigation, anti-interference and motion noise suppression are realized simultaneously, which greatly improves the detection capability of infrared detection equipment; the anti-interference success rate of weak target tracks is improved by more than 50% compared with the original technology; the navigation of weak targets with only 50% detection probability is established. The navigation success rate is increased by more than 45%. To address the problem of asynchronous measurement data, the 360° image area scanned by the infrared detection equipment is evenly divided into 16 sectors, and the track information is sent by sector. At the same time, processing by sector brings another problem. During the Mth circle, the target is in the Nth sector. During the M+1th circle, the target may fly to the N-1 or N+1 sector. This makes data processing difficult. Therefore, when designing the algorithm, multiple adjacent sector information is considered for track initiation and track update. When inputting the measurement point track data, SVM is first used for pre-segmentation. Class processing; at the same time, before continuing the track, the standard gating technology is used to divide the LMB form multi-target prediction density function into different groups; each group of measurement points is adjacent to the track space position; group processing significantly reduces the computational complexity of the track continuation process; consider introducing the multi-model idea into the labeled multi-Bernoulli filter, and use a combination of multiple motion models to handle the multi-target track processing problem of infrared detection equipment; using the "label technology", the Gaussian component inherits the corresponding historical track label in each iteration, and the established stable track is sent out according to the batch number.

[0115] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0116] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may 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.

[0117] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes 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 a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0118] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0119] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0120] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0121] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A multi-target track processing method for infrared panoramic images based on LMB, characterized in that: The multi-target track processing method comprises: Step 1, dividing the infrared panoramic image into sectors, reading measurement points according to each sector, and classifying the points using SVM; Step 2, establish a track list, including new point tracks and surviving tracks after label multi-Bernoulli algorithm prediction filtering; if the track list is empty, execute step 4; Step 3, grouping and processing the established track list based on the standard gating technology; executing step 4 for the points that do not fall into any track gate, and updating the tracks across adjacent sectors based on the label multi-Bernoulli algorithm for the points that fall into any track gate; and updating the surviving tracks in the track list in the next cycle based on the updated tracks; Step 4: Create new points for the input points that have not been successfully associated with the track. If there are new points, they are used to establish new points in the track list in the next cycle.

2. The multi-target track processing method according to claim 1, characterized in that: The process of starting a new track across adjacent sectors based on the label multi-Bernoulli algorithm in step 3 includes: When the current sector number of the trace is N+1 and the current scanning cycle number is M+1, the trace of the Nth sector in the Mth scanning cycle and the traces of the N-1th, Nth and N+1th sectors in the M+1th scanning cycle are collected; The targets are screened and the label multi-Bernoulli algorithm filtering and updating are performed. When the point track is determined to be within the track starting gate, the target is batched and the track is output.

3. The multi-target track processing method according to claim 1, characterized in that: In step 3, the point track is compared with the predicted track of the track based on the standard gating technology and then grouped and processed, including: The predicted coordinates of the track in the predicted track are taken out in sequence, step 4 is executed for the track points that do not fall into any track gate, and the track across adjacent sectors is updated based on the label multi-Bernoulli algorithm for the track points that fall into any track gate.

4. The multi-target track processing method according to claim 1, characterized in that: The process of updating the track across adjacent sectors based on the label multi-Bernoulli algorithm for the point track falling into any track gate in step 3 includes: When the current sector number of the trace is N+1 and the current scanning cycle number is M+1, the traces of the sectors N-1, N and N+1 of the M+1th scanning cycle are collected; When the point track data of the first sector of any M+1th scanning cycle is collected, the track of the Mth scanning cycle is predicted by label multi-Bernoulli algorithm filtering; The point tracks of the M+1th scanning cycle are grouped according to the track information of the Mth scanning cycle to generate various track groups, and the tracks of each track group are subjected to label multi-Bernoulli algorithm filtering and updating processing.

5. The multi-target track processing method according to claim 4, characterized in that: The performing label multi-Bernoulli algorithm filtering and updating processing on the tracks of each track grouping includes: performing label multi-Bernoulli algorithm filtering and updating on the tracks under different association conditions, and calculating the track weights under different association conditions.

6. The multi-target track processing method according to claim 5, characterized in that: After the label multi-Bernoulli algorithm filtering and updating processing is performed on the tracks of each track group, the method further includes: normalizing the track weights of each track group, clipping and merging the updated tracks, and then outputting the updated tracks.

7. The multi-target track processing method according to claim 6, characterized in that: The clipping of the updated track includes: clipping the track whose weight value is less than a set threshold.

8. An infrared panoramic image multi-target track processing system based on LMB, characterized in that: include: Point track classification module, track list establishment module, point track generation module, track start module and track continuation module; The point trace classification module is used to divide the infrared panoramic image into sectors, read the measurement point traces according to each sector, and classify the point traces using SVM; The track list establishment module is used to establish a track list, including new point tracks and surviving tracks after label multi-Bernoulli algorithm prediction filtering; The point track regeneration module performs point track regeneration on the points that cannot be associated with the track chain list, so as to enter the track chain list in the next cycle; The track start module refers to the successful association of the input track with the new track of the previous cycle and giving a new track batch number; The track continuation module is used to compare the point track with the predicted track of the track based on the standard gating technology and then perform grouping processing; the point track that does not fall into any track wave gate enters the point track regeneration module, and the point track that falls into any track wave gate is updated across adjacent sectors based on the label multi-Bernoulli algorithm; the track linked list is updated based on the updated track.

9. An electronic device, characterized in that: The method comprises a memory and a processor, wherein the processor is used to implement the steps of the LMB-based infrared panoramic image multi-target track processing method as described in any one of claims 1 to 7 when executing a computer management program stored in the memory.

10. A computer-readable storage medium, characterized in that: A computer management program is stored thereon, and when the computer management program is executed by the processor, the steps of the LMB-based infrared panoramic image multi-target track processing method as described in any one of claims 1 to 7 are implemented.

Images