CN-119887845-B - Disturbance satellite target tracking and positioning method integrating binocular information

CN119887845BCN 119887845 BCN119887845 BCN 119887845BCN-119887845-B

Abstract

The embodiment of the invention provides a disturbance satellite target tracking and positioning method fusing binocular data, which comprises initializing a target tracker based on angular point information, ID information and the like of a disturbance star target acquired by a camera, constructing a target tracker of a disturbance star by adopting a group of KCF filters, mainly comprising four angular point trackers and a central tracker, calculating target responsiveness in an image frame based on a tracking image of the tracker in real time, updating the current tracker position and pyramid level at the highest response, constructing target trustworthiness based on the tracking condition of the star target to characterize the tracking stability of the target, and finally finishing the fusion of binocular positioning and monocular target solving based on the target trustiness by selecting a binocular position solving method or a pose solving method based on the target to realize the visual positioning of the disturbance star target. According to the technical scheme provided by the embodiment of the invention, reference can be provided for tracking and positioning of the scrambling star target fused with binocular data.

Inventors

CHEN GANG
WANG HAITONG
LI TONG
WANG YIFAN
REN YI
WANG YIFAN
Gao Xianyuan

Assignees

北京邮电大学

Dates

Publication Date: 20260508
Application Date: 20250114

Claims (8)

1. A disturbance satellite target tracking and positioning method integrating binocular information is characterized by comprising the following steps: S1, selecting a target ID, performing target feature extraction based on an image acquired by a camera, acquiring four corner information of a target, and matching the target ID information; s2, initializing a target tracker according to the corner information and the target ID information to generate a target tracker group, namely adopting a group of trackers to form a target tracker; S3, tracking the change of the target position in a real-time image stream by using an initialized target tracker, and calculating a target response diagram; s4, judging whether the target tracker is lost in tracking or not based on the response graph; S5, if the target tracker is lost, executing the steps S1 and S2 to reinitialize the target tracker, if the target tracker is not lost, updating the position of the target tracker in real time, and calculating the credibility of the target tracker based on the position updating condition of the target tracker; S6, selecting a final pose resolving strategy according to the credibility, selecting a target-based pose resolving method when the credibility is higher than a set threshold value, selecting a binocular stereo matching-based position resolving method when the credibility is lower than the threshold value, and resolving relative position or pose information of the target; And S7, outputting pose information of the satellite target based on the credibility and the resolving result, and completing target positioning of the disturbance satellite.
2. The method according to claim 1, wherein the step S1 comprises: s1.1, selecting ID information of a target and taking the size and the size of the target as known target parameters; S1.2, acquiring an image by adopting a camera and recording the image as an image Performing matching judgment image based on predefined target ID information and identified target ID information Which image blocks are target objects, and marking the extracted target objects as ; S1.3 targeting the detected target object Extracting characteristic information corresponding to the target, including coordinates of the target in the center of the image And the pixel coordinates of the four corner points Wherein c represents an image The pixel coordinates of the four corners of m are marked.
3. The method according to claim 2, wherein the step S2 comprises: S2.1 setting function Identification image In order to get Is of the center and the size Is a block of an image; S2.2 based on the center coordinates described in S1.3 Pixel coordinates of four corner points Five feature image blocks are generated And The image block is the designed KCF tracker with the side length of Is a tracking image of (1); s2.3 generating a target global tracker by using a KCF filtering algorithm based on the image block described in S2.1 And four target corner trackers ; S2.4 is based on the target global tracker and four target corner trackers described in S2.3: Wherein the method comprises the steps of KCF tracker for representing mark center and four corners, and recording pyramid level corresponding to the initialized tracker at this time as Representing a pyramid level for target m correlation at time t, the pyramid level And the scale change characterization parameters are used for characterizing the approaching and the separating of the images.
4. A method according to claim 3, wherein said step S3 comprises: S3.1, performing response calculation in real time in an image stream by using a target tracker group and representing the response calculation in a form of a response graph, and in order to accelerate the response calculation, using a multi-scale parallel calculation tracker group, performing pyramid level calculation in step S2.4 Two adjacent pyramid dimensions 、 The above response conditions include: Pyramid-level-based using the set of target trackers Performing frequency domain correlation operation on the real-time image to obtain a response chart ; Pyramid-level-based using the set of target trackers Performing frequency domain correlation operation on the real-time image to obtain a response chart ; Pyramid-level-based using the set of target trackers Performing frequency domain correlation operation on the real-time image to obtain a response chart ; Wherein the method comprises the steps of Representing element-wise multiplication, representing complex conjugation, Representing all sides in an image frame at the current pyramid level as A set of patch blocks; S3.2, weighting and calculating a response graph of the current image according to weights based on the response graph calculated in each pyramid level in the S3.1, wherein the response graph comprises the following steps: Wherein the method comprises the steps of Is the scale of The weight of the upper layer is calculated, Representing dimensions The response diagram at the time t is finally obtained A response image for the current detection; S3.3 Each pyramid level described based on S3.1 And 、 Comparing the response conditions of the images, and updating the pyramid level of the maximum response of the response graph to the new pyramid level recorded by the current tracking image 。
5. The method according to claim 4, wherein the step S4 includes: S4.1 design function For computing in a tracker group Maximum response value corresponding to position p in the corresponding response diagram, where Representing a filter centered around a pixel p in an image I Is a response to (a); s4.2 based on the response map Obtaining a tracker corresponding to the central coordinate in the tracker group by adopting a maximum response value calculation function Maximum response value of (2) Define the tracker loss threshold as If the maximum response value is below the loss threshold, then the marker tracking is deemed lost.
6. The method according to claim 5, wherein the step S5 includes: S5.1 based on the S4.2 judgment criterion, if the response is maximum Below the loss threshold If the target tracker group is lost, the target tracker group is required to be initialized again, and the steps S1 and S2 are re-executed for re-tracking; s5.2 based on the S4.2 judgment criterion, if the response is maximum Above the loss threshold And if the target tracking is successful, updating the position of the new target in the image to the position p of the maximum response corresponding to the tracker group in the image, calculating the reliability of the corresponding tracking based on the position p, and reflecting the reliability of the tracking by the reliability, wherein the reliability index mainly comprises two parts of contents.
7. The method according to claim 6, wherein the step S6 includes: setting up a confidence threshold based on the confidence level of S6 When the credibility is larger than the threshold value, the target image definition is considered to be higher, and the target pose information is output by using a target-based pose solving method, wherein the specific formula is expressed as follows: In the middle of Representing the pose corresponding to the mark m at the time t, and calculating the credibility of the target when Below a threshold value When the method is used, a binocular position resolving method with stronger resolving robustness is adopted, otherwise, a pose resolving strategy based on a target with higher precision is adopted; when the reliability of the current target tracker is low, namely the image representing the faster movement of the target is blurred, a clear and error-free target image cannot be obtained, a central pixel point of the target is obtained through the target tracker, and the actual position information of the target is obtained by combining a binocular stereo matching mode based on the position of the pixel point in the binocular camera; And the pose resolving strategy is that when the reliability of the current target tracker is higher, the target is uniformly moved, the target image is clear, the target image with higher definition is obtained, and based on the image, the accurate target pose information is calculated by using the OpenCV-based target pose resolving method.
8. The method according to claim 7, wherein the step S7 includes: Based on pose information Corresponding trust level And (3) finishing positioning of the target of the disturbance sub-star, outputting the calculated target pose information, and circularly executing the steps S5-S7 to perform continuous target positioning operation so as to realize continuous tracking and pose calculation of the target of the disturbance sub-star.

Description

Disturbance satellite target tracking and positioning method integrating binocular information Technical Field The invention relates to the technical field of space robots, in particular to a disturbance satellite target tracking and positioning method integrating binocular data. Background Along with the increasing demand of space exploration, the significance and the effect of autonomous on-orbit service by a space robot are increasingly remarkable, and the space robot is widely applied to the execution of various tasks of a large spacecraft, and becomes one of the most important ways of on-orbit service. The on-orbit target capturing is an important link of an on-orbit operation task, and a high-precision visual servo capturing technology is required for accurate and efficient target capturing, wherein the precision of visual positioning directly influences the success rate of subsequent servo capturing, so that the visual positioning detection method with both stability and accuracy is studied, and has extremely strong practical significance. The targets are widely used for spacecraft positioning at present, and the spacecraft positioning based on the targets can obtain more accurate spacecraft pose information based on the arm-mounted camera, so that the subsequent servo capturing operation is convenient. However, the existing satellite positioning based on the target is still in single-frame instant calculation at present, the satellite positioning effect on the satellite target with lower hovering or moving speed is better, but the positioning success rate is low and the condition of target gesture fluctuation often occurs under the condition of quick disturbance of the subsatellite or target image movement, the positioning effect of the dynamic spacecraft is greatly influenced, and the stability and accuracy of the calculation in the positioning process are easily insufficient. In addition, the single target positioning by means of the monocular image cannot guarantee positioning robustness, and the binocular camera mounted at the tail end of the existing spaceflight mechanical arm cannot be fully utilized. Therefore, the disturbance satellite target tracking and positioning method fusing binocular information has important research significance. Disclosure of Invention In view of this, the embodiment of the present invention provides a method for tracking and positioning a disturbance satellite target by fusing binocular information, including: According to the selected ID information and target parameters of the size and the dimension of the target, firstly, extracting target characteristics based on images acquired by a camera to acquire four corner information and target number information of the target; initializing a target tracker according to the corner information and the target number information to generate a target tracker group, namely adopting a group of trackers to form a target tracker; the method comprises the steps of initializing a target tracker, calculating a target response graph by utilizing the initialized target tracker to track the change of the target position in a subsequent real-time image stream, judging whether the target tracker is lost or not based on the response graph, updating the target tracker position in real time if the target tracker is not lost, and reinitializing the target tracker based on the first two steps if the target tracker is lost; The reliability of tracking of the group of trackers is directly represented by the reliability, a final pose resolving strategy is switched according to the reliability, a pose resolving method based on a target is selected if the reliability is higher, and a position resolving method based on binocular stereo matching is selected if the reliability is lower; And outputting pose information of the satellite target based on the confidence index and a resolving result of the pose resolving method, and completing target positioning. In the method, the target feature extraction process mainly extracts target information in the image, and each target information comprises data contents such as corresponding four corner coordinates, target ID information and the like; The content mainly characterizes the position information of the target in the image, and can be expressed as follows: For images The target object detected in the above can be expressed as follows: wherein c represents an image The pixel coordinates of the four corners of the middle mark m,Representing the central coordinates of the marks, with simultaneous registrationIs the observed target region. In the method, target tracker groups are generated based on target information, each group of trackers comprises four target corner trackers and a target global tracker, and each initialization-defined tracker comprises five target image blocks, namely: And Wherein the function is setIdentification imageIn order to getIs of the center and the sizeThe image block is