Disturbance satellite target tracking and positioning method fusing binocular information

Abstract

The embodiment of the invention provides a disturbance satellite target tracking and positioning method fused with binocular data. The method comprises the following steps: initializing a target tracker based on angular point information, ID information and the like of a disturbance sub-star target collected by a camera; a group of KCF filters are adopted to construct target trackers for disturbing the child satellites, and the target trackers mainly comprise four angular point trackers and a center tracker; tracking the image in the image frame in real time based on the tracker to calculate the target responsivity, and updating the current tracker position and pyramid level at the highest response position; the target credibility is constructed based on the sub-star target tracking condition to represent the tracking stability of the target; finally, a binocular position resolving method or a target-based pose resolving method is selected and used based on the target credibility, fusion of binocular positioning and monocular target resolving is completed, and visual positioning for the disturbance sub-star target is achieved. According to the technical scheme provided by the embodiment of the invention, a reference can be provided for target tracking and positioning of the disturbance sub-star fused with binocular data.

Description

[Technical field]

[0001] The invention relates to the technical field of space robots, and in particular to a disturbed satellite target tracking and positioning method integrating binocular data. [Background technology]

[0002] With the increasing demand for space exploration, the significance and role of autonomous on-orbit services through space robots have become increasingly prominent, and have been widely used to perform various tasks of large spacecraft, becoming one of the most important ways of on-orbit services. On-orbit target capture is an important part of on-orbit operation tasks, and accurate and efficient target capture requires high-precision visual servo capture technology, among which the accuracy of visual positioning directly affects the success rate of subsequent servo capture. Therefore, it is of great practical significance to study visual positioning detection methods that take into account both stability and accuracy. At present, targets are widely used for spacecraft positioning. Target-based spacecraft positioning can obtain more accurate spacecraft posture information based on the arm-mounted camera, which is convenient for subsequent servo capture operations.

[0003] However, the existing target-based satellite positioning is still in the single-frame instant solution. It has a good positioning effect for hovering or slow-moving satellite targets, but the positioning success rate is low and the target attitude often fluctuates when the satellite is rapidly disturbed or the target image moves, which greatly affects the positioning effect of dynamic spacecraft and easily leads to insufficient solution stability and accuracy during the positioning process. In addition, this method of single-use monocular image for target positioning cannot guarantee the robustness of positioning, and cannot fully utilize the binocular camera currently carried at the end of the aerospace manipulator. Therefore, it is of great research significance to propose a disturbed satellite target tracking and positioning method that integrates binocular information. [Summary of the invention]

[0004] In view of this, an example of the present invention provides a method for tracking and positioning a disturbed satellite target by fusing binocular information, comprising:

[0005] According to the selected ID information and target parameters of target size, firstly, the target feature is extracted based on the image collected by the camera to obtain the four corner point information and target number information of the target;

[0006] Initializing the target tracker according to the corner point information and the target number information, generating a target tracker group, that is, using a group of trackers to form the target tracker of this article;

[0007] Using the initialized target tracker, tracking the change of the target position in the subsequent real-time image stream, and calculating the target response map; judging whether the target tracker is lost based on the response map, and if not, updating the target tracker position in real time; and if the tracker is lost, reinitializing the target tracker based on the first two steps;

[0008] The trustworthiness of the tracking is calculated in real time based on the position update of the tracker. The trustworthiness directly represents the reliability of the tracking of the group of trackers. The final pose solution strategy is switched according to the trustworthiness. If the trustworthiness is high, the pose solution method based on the target is selected. If the trustworthiness is low, the position solution method based on binocular stereo matching is selected.

[0009] Based on the trustworthiness index and the calculation results of the posture solution method, the posture information of the satellite target is output to complete the target positioning.

[0010] In the above method, the target feature extraction process mainly extracts the target information in the image, and each target information contains data content such as the corresponding four corner point coordinates and target ID information;

[0011] The content mainly represents the location information of the target in the image, which can be expressed as:

[0012] For the target objects detected in image I, the formula can be expressed as:

[0013]

[0014] Where c represents the pixel coordinates of the four corners of the marker m in image I, C(m)∈R 2 represents the center coordinate of the marker, and A(m) is the observed target area.

[0015] In the above method, a target tracker group is generated based on the target information. Each group of trackers includes four target corner trackers and one target global tracker. Each initially defined tracker includes five target image blocks, namely: P(c1,τ s ),P(c2,τ s ),P(c3,τ s ) and P(c4,τ s ) where the function P(o,τ s ), the identification image I is Centered at The image block is the tracking image with a side length of τ of the KCF tracker designed in this paper.

[0016] Based on the image block, a target global tracker is generated using the KCF filtering algorithm and four target corner trackers for:

[0017]

[0018] in Represents the KCF tracker for the marker center and four corners; also records the current marker image pyramid level here It represents the pyramid level related to the target m at time t, that is, the parameter representation used to characterize the image scale change (far away or close).

[0019] In the above method, the initialized target tracker is used to track the change of the target position in the subsequent real-time image stream and update the target tracker position in real time; this mainly includes completing the response of the target tracker in the subsequent image (t>1) based on the image response. The response image can be obtained by performing frequency domain correlation operations on the current image using the filter group defined in the target tracker defined in the previous article. Where ⊙ represents element-by-element multiplication, * represents complex conjugation, and X represents the set of all patches with a side length of τ in the current image frame. In the process of response calculation, the scale factor, i.e., the pyramid level, needs to be considered. Therefore, a multi-scale parallel calculation and weight fusion method is proposed to update the response degree. That is, the response map is obtained from two adjacent pyramid scales. Calculate the response graph. The specific calculation is as follows:

[0020] Using the target tracker group, based on the pyramid level Perform frequency domain correlation operations on real-time images to obtain response graphs

[0021] Using the target tracker group, based on the pyramid level Perform frequency domain correlation operations on real-time images to obtain response graphs

[0022] Using the target tracker group, based on the pyramid level Perform frequency domain correlation operations on real-time images to obtain response graphs

[0023] And extract the response based on the weight to obtain a more accurate response, that is, the response map of the current target can be defined as

[0024]

[0025] where w l is the weight on scale l, Represents the response diagram at time t on scale l, and the final result is is the response image of this detection. And according to the three pyramid scales calculated this time and The pyramid level with the maximum response in the response graph is selected as the new pyramid level recorded by the current tracking image.

[0026] Get the corresponding graph Then the maximum response position is calculated and recorded as p, and the new target tracker position can be updated as In this formula, PSR represents the filter centered on pixel p in image I. Based on the response graph The maximum response value calculation function is used to obtain the tracker corresponding to the center coordinate in the tracker group. Maximum response value of the tracker used to track the center of the marker The tracker loss threshold is defined as χ. If the maximum response value is lower than the loss threshold, the marker tracking is considered lost.

[0027] In the above method, the trustworthiness index mainly represents the clarity of the target image under the motion condition, and its calculation mainly includes:

[0028] The trustworthiness index is mainly composed of two parts. The first part is to preliminarily determine the trustworthiness by extracting the target polygonal bit sequence b(p) based on the four corner point coordinates obtained by the KCF target tracker and the normalized Hamming distance between the tracking target bit sequence b(m) detected during the initialization detection. The F1 function is constructed based on b(p) and b(m) to preliminarily determine the trustworthiness:

[0029]

[0030] Where H represents the normalized Hamming distance calculation function. The trustworthiness index of the first part is calculated as above.

[0031] The second is the loss of the four corner point filters created in the target tracker group. According to the loss of the four corner point trackers created in the target tracker group, that is, the response of the four corner point trackers, the reliability of the detection can also be determined. Based on this, the F2 function is constructed to further determine the reliability:

[0032]

[0033] in

[0034]

[0035] Where PSR represents the maximum response of the tracker, and χ represents the threshold of target tracker loss, that is, if the target tracker response is less than the threshold, the target tracker is considered to be lost.

[0036] Combining the above F1 and F2 functions, the trustworthiness measure of this paper is constructed as follows:

[0037]

[0038] Based on the trustworthiness, the ambiguity and trustworthiness of the target tracked by the tracker can be conveniently evaluated.

[0039] In the above method, the pose solution method is switched based on the trustworthiness, and the position or pose information of the target is finally output. The main calculation process is as follows:

[0040] A trustworthiness threshold ε is set as a judgment indicator. When the trustworthiness is less than the threshold, it is considered that the target image is highly blurred and the position information is calculated using a binocular position calculation method. When the trustworthiness is greater than the threshold, it is considered that the target image is highly clear and the target-based position calculation method is used to output the target position information. The specific formula for judgment can be expressed as follows:

[0041]

[0042] In the formula Indicates the pose corresponding to the marker m at time t. When the calculated target trustworthiness w m When the value is lower than the threshold ε, a more robust binocular position solution method is used. Otherwise, a more accurate target-based pose solution strategy is used.

[0043] Binocular position solution method: When the current target tracker has low reliability, that is, the target moves quickly and the image is blurred, a clear target image cannot be obtained. However, the target center pixel point can be obtained through the target tracker. Based on the pixel position in the binocular camera, the actual position information of the target can be obtained through binocular stereo matching. This method has excellent robustness.

[0044] Posture solution strategy: When the current target tracker has a high degree of trustworthiness, it means that the target moves evenly and the target image is clear. A target image with good clarity can be obtained. Based on this image, the opencv-based target pose solution method can be used to calculate the accurate target pose information.

[0045] Based on the posture information and the corresponding trustworthiness w m The positioning of the perturbed sub-satellite target is completed, the calculated target position and posture information is output, and the tracking and positioning steps are cyclically executed to perform continuous target positioning operations to achieve continuous tracking and position and posture solution of the perturbed sub-satellite target.

[0046] It can be seen from the above technical solutions that the embodiments of the present invention have the following beneficial effects:

[0047] The present invention proposes to track the randomly moving sub-satellite target through a target tracker group, to track the moving sub-satellite target information by constructing a group of KCF filters, to locate the pixel position of the target in real time through the tracking situation of the target tracker group, and to design a tracking trustworthiness mechanism for real-time judgment of the clarity of the target, and then switch to a more robust binocular position solution method or a more accurate target pose settlement method to obtain the target position / pose. The present invention constructs a real-time visual pose solution for the perturbed sub-satellite airborne target, taking into account both accuracy and robustness, and can achieve stable visual positioning of the robot.

Brief Description of the Drawings

[0048] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creativity and labor.

[0049] Figure 1 is an operation flow chart provided by an embodiment of the present invention;

[0050] Figure 2 This is a flow chart of the perturbation satellite position and posture solution algorithm provided by an embodiment of the present invention.

[0051] Figure 3 This is a diagram showing the initialization effect of the sub-satellite target tracker provided by an embodiment of the present invention.

[0052] Figure 4 The real-time position tracking curve of the sub-satellite disturbance target provided by the embodiment of the present invention is [Specific implementation method]

[0053] In order to better understand the technical solution of the present invention, the embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0054] It should be clear that the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] The embodiment of the present invention provides a method for tracking and positioning a disturbed satellite target by fusing binocular data. The method includes the following steps. For the operation process, please refer to Figure 1 :

[0056] Step 101, select the target ID and size parameters, extract the target features based on the image captured by the camera, and obtain the four corner point information of the target and the target number information.

[0057] Specifically, the target detection function is used to detect all target information in the camera image. After that, all target objects detected in the image I can be expressed by the formula:

[0058]

[0059] Where c represents the pixel coordinates of the four corners of the marker m in image I, C(m)∈R 2 represents the center coordinate of the marker, and A(m) is the observed target area.

[0060] That is, after the above steps, the target corner point position information in the initial frame can be extracted and recorded in the list.

[0061] Step 102, based on the extracted target information, initialize the target tracker and generate a target filter group, that is, use a group of filters to form the target tracker of this article.

[0062] Please refer to Figure 2 , which is a schematic diagram of the target tracker generated in this paper.

[0063] That is, after obtaining the target coordinates, first define is the area of ​​the tracking block designed in this paper. Based on this, the function P(o,τ s ), the identification image I is Centered at This function is used to obtain the tracking image block of the tracker with the position information of the target obtained as the center. P(c1,τ s ),P(c2,τ s ),P(c3,τ s ), based on the tracking image block, the tracker group of target m at time t can be defined as:

[0064]

[0065] in Represents the KCF tracker for the center and four corners of the marker; at the same time, the current identification image pyramid level is initialized here It represents the pyramid level related to the target m at time t, that is, the parameter representation used to characterize the image scale change (far away or close).

[0066] Step 103, based on the initialized tracker, tracks the changes of the target position in the image stream in real time. The real-time tracking of the target position mainly includes: based on the initialized target tracker, tracking the changes of the target position in the subsequent real-time image stream, and updating the target tracker position in real time; this mainly includes completing the response of the target tracker in the subsequent image (t>1) based on the image response. The response image can be obtained by performing frequency domain correlation operations on the current image using the filter group defined in the target tracker defined above. Where ⊙ represents element-by-element multiplication, * represents complex conjugation, and X represents the set of all patches with a side length of τ in the current image frame. In the process of response calculation, the scale factor, i.e., the pyramid level, needs to be considered. Therefore, a multi-scale parallel calculation and weight fusion method is proposed to update the response degree. That is, the response map is obtained from two adjacent pyramid scales. Calculate the response graph. The specific calculation is as follows:

[0067] Using the target tracker group, based on the pyramid level Perform frequency domain correlation operations on real-time images to obtain response graphs

[0068] Using the target tracker group, based on the pyramid level Perform frequency domain correlation operations on real-time images to obtain response graphs

[0069] Using the target tracker group, based on the pyramid level Perform frequency domain correlation operations on real-time images to obtain response graphs

[0070] And extract the response based on the weight to obtain a more accurate response, that is, the response map of the current target can be defined as

[0071]

[0072] where w l is the weight on scale l, Represents the response diagram at time t on scale l, and the final result is is the response image of this detection. And according to the three pyramid scales calculated this time and The pyramid level with the maximum response in the response graph is selected as the new pyramid level recorded by the current tracking image.

[0073] In the response diagram Then the maximum response position is calculated, which is the new target filter position p, and the new target tracker position will be updated as In this formula, PSR represents the filter centered on pixel p in image I. The corresponding target filter here is It will be updated to the position with the highest response in the current response map. When all target filters are updated, the pyramid response level of this frame will be updated to the pyramid level with the highest response corresponding to the filter with the highest response in the current frame according to the filter response. The maximum response value calculation function is used to obtain the tracker corresponding to the center coordinate in the tracker group. Maximum response value of the tracker used to track the center of the marker The tracker loss threshold is defined as χ. If the maximum response value is lower than the loss threshold, it is considered that the marker tracking is lost, and it is necessary to re-initialize the tracking, that is, repeat the operations performed in step 101 and step 102 to re-track and locate.

[0074] Step 104, based on the position update of the target tracker, the trustworthiness of the tracker is calculated in real time. The trustworthiness directly represents the tracking reliability of the current target tracker.

[0075] The specific target tracking trustworthiness index is mainly composed of two parts. The first part is to extract the target polygonal bit sequence b(p) based on the four corner point coordinates obtained by the KCF target tracker and the normalized Hamming distance between the tracking target bit sequence b(m) detected during the initialization detection to preliminarily determine the trustworthiness. The F1 function is constructed based on b(p) and b(m) to preliminarily determine the trustworthiness:

[0076]

[0077] Where H represents the normalized Hamming distance calculation function. The trustworthiness index of the first part is calculated as above.

[0078] The second is the loss of the four corner point filters created in the target tracker group. According to the loss of the four corner point trackers created in the target tracker group, that is, the response of the four corner point trackers, the reliability of the detection can also be determined. Based on this, the F2 function is constructed to further determine the reliability:

[0079]

[0080] in

[0081]

[0082] Where PSR represents the maximum response of the tracker, and χ represents the threshold of target tracker loss, that is, if the target tracker response is less than the threshold, the target tracker is considered to be lost.

[0083] Combining the above F1 and F2 functions, the trustworthiness measure of this paper is constructed as follows:

[0084]

[0085] Based on the trustworthiness, the ambiguity and trustworthiness of the target tracked by the tracker can be conveniently evaluated.

[0086] Step 105, switch the final pose solution strategy according to the confidence level. If the confidence level is high, which means the image is clear and easy to accurately locate, a more accurate target-based pose solution method can be used. On the contrary, if the confidence level is lower than the set detection threshold, a binocular stereo positioning-based position solution method is used. Please refer to the relationship between the switching strategy based on confidence and the tracking and positioning algorithm process. Figure 3 .

[0087] Specifically, the target-based positioning algorithm is mainly based on the PnP algorithm to obtain the relative pose of the camera coordinate system relative to the marker coordinate system. This target-based positioning often requires sufficient image visibility so that the camera can correctly detect them. Therefore, it is necessary to ensure that the relative movement speed of the satellite and the camera is low so that a clear target image can be captured correctly. If the reliability of the solution in the previous step is high and the outline of the square target can be accurately located, the PnP pose solution method for the target in opencv can be called to obtain the relative position and pose.

[0088] The binocular pose solution method can use two-dimensional binocular images for stereo matching to restore the position of points in the three-dimensional world. This method has extremely high robustness, that is, it can obtain relatively accurate position information when the relative speed between the satellite and the camera is high, but it cannot obtain the satellite's attitude information. When the reliability of the previous solution is not high but the target is lost, the satellite positioning can be performed based on the target center pixel point in the left and right eye images.

[0089] Specifically, through the position of the target pixel in the binocular image, we can construct the mapping relationship between the two-dimensional binocular image and the image coordinates in the display world:

[0090]

[0091] Among them, M 11 、M 12 、M 13 and M 21 、M 22 、M 23Respectively represent the parameters in the left and right camera imaging models, M 11 、M 12 、M 21 、M 22 Indicates the internal parameters of the lower camera, M 13 、M 23 Represents the external parameters of the camera. The parameters in the above formula can be obtained by calibrating the binocular camera, and the position coordinate information in the real world can be solved by expanding the equation group.

[0092] In summary, since the pose solution based on the target in this paper can solve more accurate pose information when the target is moving at a low speed, and the position solution method based on binocular matching can provide a more robust position solution method when the target is tracked, in order to determine when to use pose solution and when to use position solution, the judgment threshold ε is defined. When the current frame confidence w of target tracking recognition is m When it is less than the threshold, the acquired target is considered to be blurred to a large extent, and the binocular position solution strategy is adopted. Otherwise, the target-based posture solution strategy is adopted. The specific judgment is shown in the following formula:

[0093]

[0094] In the formula Indicates the pose corresponding to the marker m at time t. When the calculated target confidence w m When the value is lower than the threshold ε, the binocular position solution method with stronger solution robustness is adopted. Otherwise, the target-based pose solution strategy with higher accuracy is adopted. Finally, the position and pose information of the disturbed satellite can be obtained. The pose solution effect of the disturbed satellite is as follows: Figure 4 shown.

[0095] The technical solution of the embodiment of the present invention has the following beneficial effects:

[0096] In the technical solution of the embodiment of the present invention, a method for tracking and positioning a disturbed satellite target that integrates binocular data is designed. First, the target tracker is initialized based on the disturbed sub-satellite target collected by the camera, a set of filters are used to track the sub-satellite target, and then the tracking is performed in parallel in the image, and the target position is updated based on the responsiveness cycle; and the stability of the tracking is characterized by constructing a target tracking trust index to evaluate the trustworthiness of the tracking target image; finally, the posture solution method is selected based on the trustworthiness of the sub-satellite tracking and the target tracking information. If the trustworthiness is high, the target posture solution method is selected, and if the trustworthiness is low, the binocular position solution method is selected to achieve the perturbed sub-satellite target positioning that takes into account both accuracy and robustness.

[0097] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

[0098] The contents not described in detail in the specification of the present invention belong to the common knowledge of those skilled in the art.

Claims

1. A method for tracking and positioning a disturbed satellite target by integrating binocular information, characterized in that: The method comprises: S1 selects target parameters such as target ID and target size, first extracts target features based on the image captured by the camera, obtains the four corner point information of the target and matches the target ID information; S2 initializes the target tracker according to the corner point information and the target ID information, and generates a target tracker group, that is, a group of trackers is used to form a target tracker; S3 uses the initialized target tracker to track the changes in the target position in the real-time image stream and calculate the target response map; S4 determines whether the target tracker has lost tracking based on the response graph; S5: If the target tracker is lost, execute steps S1 and S2 to reinitialize the target tracker; if the target tracker is not lost, update the position of the target tracker in real time, and calculate the trustworthiness of the tracker based on the position update of the target tracker; S6 selects a final pose solving strategy according to the trustworthiness. If the trustworthiness is higher than a set threshold, a pose solving method based on a target is selected. If the trustworthiness is lower than the threshold, a position solving method based on binocular stereo matching is selected to solve the relative position or pose information of the target. S7 outputs the position and posture information of the satellite target based on the trustworthiness and solution results, and completes the target positioning of the perturbation sub-satellite.

2. The method according to claim 1, characterized in that The step S1 comprises: S1.1 Select the target ID information and target size as known target parameters; S1.2 uses a camera to capture an image, which is recorded as image I. Based on the predefined target ID information and the identified target ID information, it is determined which image blocks in image I are target objects, and the extracted target objects are recorded as A(m); S1.3 For the detected target object A(m), extract the feature information corresponding to the target, including the target’s coordinates C(m)∈R at the center of the image 2 , and the pixel coordinates of the four corner points Where c represents the pixel coordinates of the four corners of marker m in image I.

3. The method according to claim 1, characterized in that The step S2 comprises: S2.1 Setting function P(o,τ s ), the identification image I is Centered at The image block; S2.2 Use the center coordinates and the coordinate information of the four corner points described in step S1.3 to generate five feature image blocks and P(c4,τ s ), the image block is the designed KCF tracker with a side length of τ s Tracking image of S2.3 Based on the image block, a target global tracker is generated using the KCF filtering algorithm and four target corner trackers S2.4 defines a target tracker group based on the five trackers as: in Represents the KCF tracker that marks the center and four corners; and records the pyramid level corresponding to the initialized tracker at this time as represents the pyramid level associated with target m at time t, the pyramid level A parameter used to characterize the scale change of images approaching and moving away.

4. The method according to claim 1, characterized in that: The step S3 comprises: S3.1 uses the target tracker group to calculate the response in real time in the image stream and present it in the form of a response graph. In order to accelerate the response calculation, a multi-scale parallel calculation tracker group is used to calculate the pyramid level described in step S2.

4. and two adjacent pyramid scales The responses include: Using the target tracker group, based on the pyramid level Perform frequency domain correlation operations on real-time images to obtain response graphs Using the target tracker group, based on the pyramid level Perform frequency domain correlation operations on real-time images to obtain response graphs Using the target tracker group, based on the pyramid level Perform frequency domain correlation operations on real-time images to obtain response graphs Where ⊙ represents element-by-element multiplication, * represents complex conjugate, Indicates that all the edges in the image frame at the current pyramid level are of length τ s A collection of patches; S3.2 The response map of the current image is weightedly calculated based on the response maps calculated in each pyramid level according to the weights: where w l is the weight on scale l, Represents the response diagram at time t on scale l, and the final result is is the response image of this detection; S3.3 Based on the pyramid levels described in S3.1 and The response of the response graph is compared, and the pyramid level of the maximum response of the response graph is updated to the new pyramid level recorded by the current tracking image.

5. The method according to claim 1, characterized in that The step S4 comprises: S4.1 Design function Used to calculate the tracker group The maximum response value corresponding to position p in the corresponding response map, where PSR represents the filter centered on pixel p in image I Response; S4.2 Based on the response diagram The maximum response value calculation function is used to obtain the tracker corresponding to the center coordinate in the tracker group. The maximum response value The tracker loss threshold is defined as χ. If the maximum response value is lower than the loss threshold, the marker tracking is considered lost.

6. The method according to claim 1, characterized in that The step S5 comprises: S5.1 Based on the judgment criteria described in S4, if the maximum response If it is lower than the loss threshold χ, it is considered lost and the target tracker group needs to be reinitialized and steps S1 and S2 need to be re-executed to track again; S5.2 Based on the judgment criteria described in S4, if the maximum response If the value is higher than the loss threshold χ, the target tracking is considered successful, and the position of the new target in the image is updated to the position p of the maximum response of the tracker group in the image. The trustworthiness of the corresponding tracking is calculated based on the position p. The trustworthiness reflects the reliability of the tracking. The trustworthiness index mainly consists of two parts. The specific calculation process includes: Define b(p) as the polygonal bit sequence of the target extracted based on the four corner point coordinates obtained by the KCF target tracker, b(m) as the tracking target bit sequence detected during initialization detection, calculate the normalized Hamming distance based on b(p) and b(m), and construct the F1 function: Where H represents the normalized Hamming distance calculation function; According to the loss of the four corner trackers created in the target tracker group, the F2 function is constructed to further determine the credibility: in Where maxPSR() represents the maximum response of the tracker; Combining F1 and F2 functions to build trustworthiness: Based on the trustworthiness, the ambiguity and trustworthiness of the target tracked by the tracker can be conveniently evaluated.

7. The method according to claim 1, characterized in that The step S6 comprises: Based on the trustworthiness described in S5, a trustworthiness threshold ε is set as a judgment indicator. When the trustworthiness is less than the threshold, it is considered that the target image is highly blurred and the position information is calculated using a binocular position calculation method. When the trustworthiness is greater than the threshold, it is considered that the target image is highly clear and the target-based position calculation method is used to output the target position information. The specific formula judgment can be expressed as follows: Where P t m Indicates the pose corresponding to the marker m at time t. When the calculated target trustworthiness w m When the value is lower than the threshold ε, a binocular position solution method with stronger solution robustness is adopted. Otherwise, a target-based pose solution strategy with higher accuracy is adopted. Binocular position solution method: When the current target tracker has low reliability, that is, the target moves quickly and the image is blurred, and a clear target image cannot be obtained, the target center pixel point can be obtained through the target tracker. Based on the pixel position in the binocular camera, the actual position information of the target can be obtained through binocular stereo matching. This method has excellent robustness. Posture solution strategy: When the current target tracker has a high degree of trustworthiness, it means that the target moves evenly and the target image is clear. A target image with good clarity can be obtained. Based on this image, the opencv-based target pose solution method can be used to calculate the accurate target pose information.

8. The method according to claim 1, characterized in that The step S7 comprises: Based on the posture information P t m and the corresponding trustworthiness w m The positioning of the perturbed sub-satellite target is completed, the calculated target position and posture information is output, and steps S5 to S7 are executed cyclically to perform continuous target positioning operations to achieve continuous tracking and position and posture solution of the perturbed sub-satellite target.

Images