Visual pre-calibration-based zoom monitoring method for closest distance between two targets in spherical space

Abstract

The invention discloses a visual pre-calibration-based zoom monitoring method for the nearest distance between two targets in a spherical space, and belongs to the field of zoom visual measurement. The problem that when a plurality of fixed-focus cameras are adopted to monitor the closest distance between the two targets to determine the collision risk, the closest distance between the two targets cannot be calculated correctly due to the fact that the observation view angle is shielded is solved. The invention provides a layout scheme for acquiring monitoring images in a multi-view manner by using two sets of zoom cameras, in the layout scheme, the two sets of cameras are mounted on a high-latitude circle of a northern hemisphere of a spherical sky, the optical axes of the cameras point to the center of the sphere, and three monitoring ranges with different magnifications are adopted through zooming, so that zoom monitoring of the center of a target chamber is realized; and the problem that the closest distance between the two targets cannot be calculated correctly due to the fact that the observation visual angle is shielded when multiple fixed-focus cameras are used for monitoring the closest distance between the two targets is solved. The method is mainly applied to target chamber center collision risk monitoring.

Description

Technical Field

[0001] This invention belongs to the field of zoom vision measurement. Background Technology

[0002] In a high-power laser target ablation device, multiple pieces of equipment (such as robotic arms and parallel platforms) within the spherical enclosed target chamber all enter, exit, and operate at the center of the chamber. Figure 1 As shown. During the experiment, the equipment needs to be adjusted in multiple degrees of freedom. There is a risk of collision between the equipment during the movement process, which requires real-time monitoring of the movement process. The closer to the center of the target chamber, the higher the risk of collision, and the higher the accuracy requirement for collision monitoring.

[0003] In existing technologies, multiple fixed-focus cameras are used to monitor the closest distance between two targets to determine the risk of collision. However, this method cannot meet the requirement of higher accuracy as the target chamber center approaches. This is because multiple devices operating simultaneously near the target chamber center may obstruct the observation view. Furthermore, the obstruction between devices becomes more severe as the target chamber center approaches, resulting in more challenging monitoring conditions. Consequently, the closest distance between two targets monitored by the fixed-focus cameras cannot be correctly calculated. Therefore, these problems urgently need to be addressed. Summary of the Invention

[0004] The purpose of this invention is to solve the problem that existing methods using multiple fixed-focus cameras to monitor the closest distance between two targets to determine collision risk suffer from obstructed viewing angles, resulting in the inability to correctly calculate the closest distance between the two targets. This invention provides a zoom-based method for monitoring the closest distance between two targets in spherical space based on visual pre-calibration.

[0005] A method for close-range zoom detection of two targets in a spherical space based on visual pre-calibration includes the following steps:

[0006] Step 1: Zoom Camera Setup and Calibration: Set up two zoom cameras on the same latitude circle in the upper hemisphere of the spherical space of the spherical target chamber and set them up in a mirror image. The two zoom cameras are used as a group of zoom cameras and are calibrated at three different magnifications. The external parameters between the two zoom cameras are pre-calibrated, and the external parameters after calibration at each magnification are used as a preset memory point parameter. The three magnifications correspond to three fields of view with decreasing monitoring radii: large, medium, and small.

[0007] Step 2: Two zoom cameras are used to monitor target A and target B respectively. If target A or target B is observed by multiple fields of view at the same time, the current zoom camera selects the preset memory point parameter corresponding to the field of view with the smallest monitoring radius and performs image acquisition on the target corresponding to the zoom camera.

[0008] Step 3: Based on the image of target A captured by the first zoom camera, calculate the pose of target A relative to the first zoom camera. Based on the image of target B captured by the second zoom camera, calculate the pose of target B relative to the second zoom camera. Using the extrinsic parameters between the two zoom cameras and the pose of target B relative to the second zoom camera, calculate the pose of target B relative to the first zoom camera.

[0009] Step 4: Obtain the CAD point clouds of targets A and B; based on the pose relationship between targets A and B relative to the first zoom camera, transform the CAD point clouds of targets A and B into the same coordinate system; use the nearest neighbor search algorithm to calculate the distance between the nearest points in the CAD point clouds of targets A and B, and use this distance as the nearest distance between the two targets.

[0010] Preferably, the method for calibrating the zoom camera group at three different magnifications is implemented using Zhang's calibration method.

[0011] Preferably, the method for obtaining three magnifications corresponding to three fields of view—large, medium, and small—with decreasing monitoring radii, respectively, is as follows:

[0012] The center of the spherical space of the spherical target chamber is taken as the monitoring center. The optical axes of the two zoom cameras are pointed to the monitoring center. Concentric large, medium and small circular monitoring fields of view are defined within the cross section where the monitoring center is located. The monitoring radius of the large, medium and small circular monitoring fields of view decreases in sequence. The large, medium and small circular monitoring fields of view are the large, medium and small fields of view corresponding to the three magnifications.

[0013] Preferably, the extrinsic parameters between the two zoom cameras include the rotation vector between the coordinate systems of the two zoom cameras. Translation vector .

[0014] Preferably, the method for obtaining the CAD point clouds of targets A and B is as follows:

[0015] The CAD model of each target is sampled using the Poisson disk sampling method to obtain the corresponding CAD point cloud.

[0016] Preferably, the nearest neighbor search algorithm is implemented using the KD-Tree nearest neighbor search algorithm.

[0017] Preferably, the two zoom phases are positioned at 120° east and west longitudes in the spherical space, respectively.

[0018] A visually pre-calibrated spherical two-target closest-range zoom monitoring device includes a storage device, a processor, and a computer program stored in the storage device and executable on the processor. The processor executes the computer program to implement the visually pre-calibrated spherical two-target closest-range zoom monitoring method. A computer-readable storage device stores a computer program that, when executed, implements the visually pre-calibrated spherical two-target closest-range zoom monitoring method. A computer program product includes a computer program that, when executed by a processor, implements the visually pre-calibrated spherical two-target closest-range zoom monitoring method.

[0019] The beneficial effects of this invention are:

[0020] This invention provides two sets of zoom cameras to achieve zoom monitoring of the target chamber center. This avoids the problem of obstructed viewing angles when using multiple fixed-focus cameras to monitor the closest distance between two targets, which prevents the accurate calculation of the closest distance between the two targets. On the one hand, a three-level pre-calibration scheme is proposed, which pre-calibrates the target in different object-side fields of view to avoid the time-consuming on-site calibration during the monitoring process and ensures the real-time performance of the monitoring. On the other hand, when the target is observed by multiple fields of view at the same time, the current zoom camera selects the preset memory point parameters corresponding to the field of view with the smallest monitoring radius. The principle is to use the smallest possible field of view to improve monitoring accuracy, which is suitable for places with higher accuracy requirements for collision monitoring.

[0021] This invention provides a reusable monitoring technology, particularly suitable for space-based platforms, capable of measuring the position and size of ultra-high-speed, small-sized flying targets, such as for spacecraft in-orbit detection of the size and impact location of incoming space debris. Attached Figure Description

[0022] Figure 1 This is a schematic diagram illustrating the principle of multiple devices (such as robotic arms, parallel platforms, etc.) entering and exiting a spherical target chamber in the background technology.

[0023] Figure 2 This is a diagram showing the relative positional relationship between the camera and the target in a single-camera target six-free target pose monitoring method based on hierarchical structure and similarity in the existing technology; where R0 and T0 are the rotation vector and translation vector between the target coordinate system and the camera coordinate system, respectively.

[0024] Figure 3 This is a diagram showing the relative positions of two zoom cameras in the visual pre-calibration-based zoom monitoring method for the closest distance between two targets in spherical space, as described in this invention.

[0025] Figure 4This is a schematic diagram illustrating the principle of pre-calibrating the external parameters between two zoom cameras using Zhang's calibration method at three different magnifications.

[0026] Figure 5 This invention is based on the pose relationship diagram between the camera and two targets;

[0027] Figure 6 It is the CAD model of target B;

[0028] Figure 7 This is the CAD point cloud diagram of target B;

[0029] Figure 8 This is a schematic diagram illustrating the principle of transforming point clouds to the same coordinate system;

[0030] Figure 9 This is a schematic diagram illustrating the process of finding the nearest distance between point cloud A and point cloud B based on KD-Tree nearest neighbor search. Detailed Implementation

[0031] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0033] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the scope of the invention.

[0034] Source of idea:

[0035] When using a fixed-focus camera to monitor the collision risk between targets, it cannot meet the requirement of increasing accuracy as the target approaches the center of the target chamber. Existing technologies, such as single-camera methods based on hierarchical structure and similarity for six-free target pose monitoring, etc. Figure 2 As shown. Inspired by the above method, this invention provides a method for close-range zoom monitoring of two targets in a spherical space based on visual pre-calibration, which can achieve collision risk assessment. To meet the requirement that the collision risk increases as the target chamber center approaches, and the accuracy requirement for collision monitoring increases, two sets of cameras are used as zoom cameras to achieve zoom monitoring of the target chamber center. A three-level pre-calibration scheme is proposed, which performs pre-calibration in different object-side field-of-view ranges to meet the requirement that the collision risk increases as the target chamber center approaches, and the accuracy requirement for collision monitoring increases. The specific implementation method provided is as follows:

[0036] Specific Implementation Method 1: Combination Figure 3 As shown, this invention provides a method for close-range zoom monitoring of two targets in a spherical space based on visual pre-calibration. The method includes the following steps:

[0037] Step 1: Zoom Camera Setup and Calibration: Set up two zoom cameras on the same latitude circle in the upper hemisphere of the spherical space of the spherical target chamber and set them up in a mirror image. The two zoom cameras are used as a group of zoom cameras and are calibrated at three different magnifications. The external parameters between the two zoom cameras are pre-calibrated, and the external parameters after calibration at each magnification are used as a preset memory point parameter. The three magnifications correspond to three fields of view with decreasing monitoring radii: large, medium, and small.

[0038] Step 2: Two zoom cameras are used to monitor target A and target B respectively. If target A or target B is observed by multiple fields of view at the same time, the current zoom camera selects the preset memory point parameter corresponding to the field of view with the smallest monitoring radius and performs image acquisition on the target corresponding to the zoom camera.

[0039] Step 3: Based on the image of target A captured by the first zoom camera, calculate the pose of target A relative to the first zoom camera. Based on the image of target B captured by the second zoom camera, calculate the pose of target B relative to the second zoom camera. Using the extrinsic parameters between the two zoom cameras and the pose of target B relative to the second zoom camera, calculate the pose of target B relative to the first zoom camera.

[0040] Step 4: Obtain the CAD point clouds of targets A and B; based on the pose relationship between targets A and B relative to the first zoom camera, transform the CAD point clouds of targets A and B into the same coordinate system; use the nearest neighbor search algorithm to calculate the distance between the nearest points in the CAD point clouds of targets A and B, and use this distance as the nearest distance between the two targets.

[0041] To address the issue of obstructed observation views closer to the center of the target chamber within a spherical space, a layout scheme is proposed that utilize two sets of zoom cameras to acquire monitoring images from multiple perspectives. In this scheme, the two cameras are mounted on a high-latitude circle in the northern hemisphere of the sphere, specifically distributed at 120° east and west longitudes, with the camera optical axes pointing towards the center of the sphere. Figure 3 As shown, by using zoom, three different magnifications are employed for the monitoring range. The monitoring range decreases towards the center of the target chamber, while the monitoring accuracy increases.

[0042] In practical applications, calculating the target's pose relative to the camera based on images captured by the camera is achieved using existing technologies, such as the existing single-camera method for six-free target pose monitoring based on hierarchical structure and similarity. Calculating the pose of target B relative to the first zoom camera using the extrinsic parameters between two zoom cameras and the pose of target B relative to the second zoom camera is also achieved using existing technologies, as the extrinsic parameters between the two zoom cameras already include the rotation vector between the coordinate systems of the two zoom cameras. Translation vector This provides the data foundation for the solution.

[0043] Preferably, the method for obtaining three magnifications corresponding to three fields of view—large, medium, and small—with decreasing monitoring radii, respectively, is as follows:

[0044] The center of the spherical space of the spherical target chamber is taken as the monitoring center. The optical axes of the two zoom cameras are pointed to the monitoring center. Concentric large, medium and small circular monitoring fields of view are defined within the cross section where the monitoring center is located. The monitoring radius of the large, medium and small circular monitoring fields of view decreases in sequence. The large, medium and small circular monitoring fields of view are the large, medium and small fields of view corresponding to the three magnifications.

[0045] In practical applications, the Zhang's calibration method is used to calibrate zoom camera groups at three different magnifications. The extrinsic parameters (R1, T1), (R2, T2), and (R3, T3) between the two cameras are pre-calibrated under different fields of view, such as... Figure 4 As shown, the camera's pre-calibration position is set as a preset memory point, and the preset memory point parameters are recorded. The camera can be reset to the corresponding memory point using these parameters. Since the coordinate systems and pitch angles of the two cameras remain unchanged, the parameters are the same for all three fields of view: R1 = R2 = R3, T1 = T2 = T3. R1 to R3 are the rotation vectors between the coordinate systems of the two zoom cameras in the large, medium, and small fields of view, respectively, and T1 to T3 are the translation vectors between the coordinate systems of the two zoom cameras in the large, medium, and small fields of view, respectively.

[0046] To detect the closest distance between targets A and B, the poses of targets A and B are monitored using two cameras, O1 and O3, as follows: Figure 5 As shown, the pose relationship between two targets A and B is solved by pre-calibrating the extrinsic parameter relationship (R, T) between two cameras O1 and O3. R and T are the rotation vector and translation vector between the coordinate systems of the two zoom cameras, respectively. The pose of targets A and B includes position (x, y, z) and attitude (α, β, γ are the angles around the x, y, and z axes, respectively).

[0047] The corresponding CAD point clouds are obtained by using the Poisson disk sampling method on the CAD models of targets A and B.

[0048] Furthermore, the method for obtaining the CAD point clouds of targets A and B is as follows:

[0049] The CAD model of each target is sampled using the Poisson disk sampling method to obtain the corresponding CAD point cloud. For example... Figure 6 and Figure 7 As shown.

[0050] In practical applications, based on the pose relationship between targets A and B relative to the first zoom camera, the CAD point clouds of targets A and B are transformed into the same coordinate system, such as... Figure 8 As shown.

[0051] In practical applications, the nearest neighbor search algorithm used to calculate the distance between the nearest points in the CAD point clouds of target A and target B is specifically implemented using the KD-Tree nearest neighbor search algorithm. This KD-Tree nearest neighbor search algorithm is an existing technology. When using this algorithm for calculation, since the point cloud has three dimensions (X, Y, Z), the algorithm first calculates the dimension K with the largest variance, and then divides the point cloud into left and right subtrees based on the median of dimension K. For the point clouds in the left and right subtrees, the median of dimension K is repeatedly calculated for partitioning. The point cloud data generates a binary tree, i.e., a three-dimensional KD-Tree. The KD-Trees for the CAD point clouds of target A and target B are constructed using the above method.

[0052] For the CAD point clouds of target A and target B, first select the point 'a' closest to the centroid in the CAD point cloud of target A, and perform a KD-Tree nearest neighbor search algorithm between point a and the CAD point cloud of target B to obtain the nearest point 'b' to point a in the CAD point cloud of target B, and the distance S1 between point a and point b. Then, use point b to perform a KD-Tree nearest neighbor search on the CAD point cloud of target A to obtain the nearest point 'c' in the CAD point cloud of target A, and the distance S2 between point b and point c. Then, use point c to perform a KD-Tree nearest neighbor search on the CAD point cloud of target B to obtain the nearest point 'd' in the CAD point cloud of target B, and the distance S3 between point c and point d. Repeat the above process until the updated distance Si no longer decreases. Si is the closest distance between the CAD point clouds of target A and target B, and i is a variable.

[0053] Specific Embodiment Two: The visual pre-calibration-based close-range zoom monitoring device for two targets in spherical space, as described in this embodiment, includes a storage device, a processor, and a computer program stored in the storage device and executable on the processor. The processor executes the computer program to implement the visual pre-calibration-based close-range zoom monitoring method for two targets in spherical space as described in Specific Embodiment One. Specific Embodiment Three: A computer-readable storage device storing a computer program, which, when executed, implements the visual pre-calibration-based close-range zoom monitoring method for two targets in spherical space as described in Specific Embodiment One. Specific Embodiment Four: A computer program product including a computer program, which, when executed by a processor, implements the visual pre-calibration-based close-range zoom monitoring method for two targets in spherical space as described in Specific Embodiment One.

[0054] While the invention has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the invention. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that different dependent claims and features described herein can be combined in ways different from those described in the original claims. It is also understood that features described in conjunction with individual embodiments can be used in other described embodiments.

Claims

1. A method for zoom monitoring of the closest distance between two targets in a spherical space based on visual pre-calibration, characterized in that, The method includes the following steps: Step 1: Zoom Camera Setup and Calibration: Set up two zoom cameras on the same latitude circle in the upper hemisphere of the spherical space of the spherical target chamber and set them up in a mirror image. The two zoom cameras are used as a group of zoom cameras and are calibrated at three different magnifications. The external parameters between the two zoom cameras are pre-calibrated, and the external parameters after calibration at each magnification are used as a preset memory point parameter. The three magnifications correspond to three fields of view with decreasing monitoring radii: large, medium, and small. Step 2: Two zoom cameras are used to monitor target A and target B respectively. If target A or target B is observed by multiple fields of view at the same time, the current zoom camera selects the preset memory point parameter corresponding to the field of view with the smallest monitoring radius and performs image acquisition on the target corresponding to the zoom camera. Step 3: Based on the image of target A captured by the first zoom camera, calculate the pose of target A relative to the first zoom camera. Based on the image of target B captured by the second zoom camera, calculate the pose of target B relative to the second zoom camera. Using the extrinsic parameters between the two zoom cameras and the pose of target B relative to the second zoom camera, calculate the pose of target B relative to the first zoom camera. Step 4: Obtain the CAD point clouds of targets A and B; based on the pose relationship between targets A and B relative to the first zoom camera, transform the CAD point clouds of targets A and B into the same coordinate system; use the nearest neighbor search algorithm to calculate the distance between the nearest points in the CAD point clouds of targets A and B, and use this distance as the nearest distance between the two targets.

2. The method for close-range zoom monitoring of two targets in spherical space based on visual pre-calibration according to claim 1, characterized in that, The Zhang's calibration method was used to calibrate the zoom camera group at three different magnifications.

3. The method for close-range zoom monitoring of two targets in spherical space based on visual pre-calibration according to claim 1, characterized in that, The method for obtaining three magnifications corresponding to three fields of view—large, medium, and small—with decreasing monitoring radii, respectively, is as follows: The center of the spherical space of the spherical target chamber is taken as the monitoring center. The optical axes of the two zoom cameras are pointed to the monitoring center. Concentric large, medium and small circular monitoring fields of view are defined within the cross section where the monitoring center is located. The monitoring radius of the large, medium and small circular monitoring fields of view decreases in sequence. The large, medium and small circular monitoring fields of view are the large, medium and small fields of view corresponding to the three magnifications.

4. The method for close-range zoom monitoring of two targets in spherical space based on visual pre-calibration according to claim 1, characterized in that, The extrinsic parameters between the two zoom cameras include the rotation vector between the coordinate systems of the two zoom cameras. Translation vector .

5. The method for close-range zoom monitoring of two targets in spherical space based on visual pre-calibration according to claim 1, characterized in that, The method for obtaining the CAD point clouds of targets A and B is as follows: The CAD model of each target is sampled using the Poisson disk sampling method to obtain the corresponding CAD point cloud.

6. The method for close-range zoom monitoring of two targets in spherical space based on visual pre-calibration according to claim 1, characterized in that, The nearest neighbor search algorithm is implemented using the KD-Tree nearest neighbor search algorithm.

7. The method for close-range zoom monitoring of two targets in spherical space based on visual pre-calibration according to claim 1, characterized in that, The two zoom phases are respectively located at 120° east and west longitude in the spherical space.

8. A zoom monitoring device for the closest distance between two targets in a spherical space based on visual pre-calibration, comprising a storage device, a processor, and a computer program stored in the storage device and executable on the processor, characterized in that, The processor executes a computer program to implement the method for close-range zoom monitoring of two targets in spherical space based on visual pre-calibration as described in any one of claims 1 to 7.

9. A computer-readable storage device storing a computer program, characterized in that, When the computer program is executed, it implements the method for close-range zoom monitoring of two targets in spherical space based on visual pre-calibration as described in any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the method for close-range zoom monitoring of two targets in spherical space based on visual pre-calibration as described in any one of claims 1 to 7.

Images