Active near-infrared anti-occlusion planar target with spatially complementary redundant encoding, pose determination method and surgical navigation system

Abstract

The application discloses a space complementary redundant coding active near-infrared anti-occlusion plane target, a pose determination method and a surgical navigation system. The plane target comprises a film plane target and a backlight side near-infrared light emitting structure. A 3*3 region in the center of a checkerboard grid is provided with nine identity coding units, and the periphery is provided with eight redundant coding units. Except for the center position, the remaining identity coding units are one-to-one corresponding to the redundant coding units and are centrally symmetrically distributed. Each coding unit is provided with a binary value represented by a circular pattern with or without a circular pattern. When single-sided occlusion occurs, the remaining circular patterns and corner points are detected to determine the direction and reconstruct the complete grid. The occluded coding bits are recovered by using the corresponding redundant coding units. The pose is calculated according to the remaining corner points. The identity recognition continuity and the pose estimation robustness are improved without increasing the hardware conditions. The system is linked with a mechanical arm through a calibration plate. Multiple groups of images are collected at a time to complete camera calibration and hand-eye calibration, and a unified coordinate system is quickly established.

Description

Technical Field

[0001] This invention relates to the field of surgical navigation technology, and in particular to an active near-infrared anti-occlusion planar target with spatial complementary redundancy coding, a pose determination method, and a surgical navigation system. Background Technology

[0002] Near-infrared optical positioning systems are an important component of orthopedic surgical robots, neurosurgical navigation equipment, and other precision surgical navigation systems. A typical near-infrared optical positioning system includes a near-infrared camera, an optical target, and a pose calculation device. The near-infrared camera acquires images of the optical target, and the pose calculation device calculates the target's position and orientation relative to the near-infrared camera based on feature points on the target and their geometric relationships.

[0003] Existing surgical navigation systems typically employ spatial targets composed of three or four reflective spheres, luminescent spheres, or near-infrared LEDs. The system identifies the target and calculates its pose based on the geometric relationships between the spherical feature points. However, due to limitations in the surgical navigation area, target size, and instrument placement, the number of spherical feature points that can be placed on a single target is limited. When one or more spherical feature points are obscured by medical personnel's hands, surgical instruments, gauze, patient tissue, or blood contamination, the number of feature points available for identification and pose estimation is significantly reduced, easily leading to target identification loss or pose tracking interruption.

[0004] Increasing the number of targets or cameras can reduce the impact of occlusion to some extent, but it will increase the size of the surgical navigation system, the amount of calibration work, the installation space and the computational complexity, which is not conducive to its application in space-constrained surgical environments.

[0005] Compared to spherical targets, planar targets can accommodate more checkerboard corner points, line segments, and coding units within a limited area, potentially improving occlusion robustness. However, existing visible light planar markers are typically designed for augmented reality or simultaneous localization and mapping scenarios, with their design goals usually being identity recognition and pose estimation under complete marking conditions. This makes it difficult to simultaneously meet the requirements of surgical navigation for small size, near-infrared imaging, local occlusion recovery, and high-precision pose estimation.

[0006] Some existing planar markers use multiple independent coded blocks to achieve occlusion tolerance. When one coded block is occluded, the system searches for other complete coded blocks for identification. This method requires at least one complete coded block to be retained and does not establish a one-to-one recovery relationship for individual identity bits. When the occluded area spans multiple coded blocks, it may still lead to recognition failure.

[0007] Furthermore, the identification and orientation encoding of ordinary planar targets typically rely on complete outer boundaries or complete corner features. When the target edges are occluded, the system struggles to determine the row and column positions of the remaining pattern within the complete target, thus failing to correctly read the identification code. Even if some corner points can still be detected, the stability of ordinary planar target pose estimation methods may decrease due to the reduced number of visible corner points, uneven corner point distribution, and image point detection errors.

[0008] In view of this, this case involves in-depth research into the aforementioned issues, which led to the formation of this case. Summary of the Invention

[0009] The purpose of this invention is to provide an active near-infrared anti-occlusion planar target with spatial complementary redundancy coding. By backing up the centrally symmetrical identity code and the redundancy code one-to-one, the occluded code bits can be recovered from the visible side when one side is occluded. This solves the problem of recognition interruption caused by local occlusion during surgery without adding hardware.

[0010] To achieve the above objectives, the present invention adopts the following technical solution: An active near-infrared anti-masking planar target with spatial complementary redundancy coding includes a film planar target and a near-infrared emitting structure disposed on the backlight side of the film planar target. The film planar target has a checkerboard grid and is provided with multiple identity coding units, multiple redundancy coding units and multiple direction coding units. Each of the identity coding units is used to represent a coding bit in the target identity coding. At least some of the identity coding units are respectively established in a one-to-one correspondence with one of the redundant coding units. The corresponding identity coding units and the redundant coding units represent the same binary coding value and are set at spatially separated positions on the film plane target. The corresponding identity coding units and the redundant coding units are located on opposite sides of the center of the film plane target, so that when the film plane target is partially occluded on one side, the corresponding identity coding bits are recovered based on a set of corresponding identity coding units and one unoccluded coding unit in the redundant coding units.

[0011] In a preferred embodiment of the present invention, the checkerboard grid comprises 25 square units in 5 rows × 5 columns, forming 36 grid intersections in 6 rows × 6 columns. Nine identity coding units are disposed in the 3 rows × 3 columns area in the center of the checkerboard grid, and eight redundant coding units correspond one-to-one with the eight identity coding units other than the central identity coding unit. When the row and column coordinates of the identity coding unit are (r, c), the row and column coordinates of the corresponding redundant coding unit are (6-r, 6-c), so that the corresponding identity coding units and the redundant coding units are centrally symmetrically distributed with respect to the center of the checkerboard grid.

[0012] In a preferred embodiment of the present invention, the direction encoding unit includes a left-top corner direction encoding unit, a left-bottom corner direction encoding unit, a right-top corner direction encoding unit, and a right-bottom corner direction encoding unit respectively disposed at the four corners of the chessboard grid. Each of the left-top corner direction encoding unit, the left-bottom corner direction encoding unit, the right-top corner direction encoding unit, and the right-bottom corner direction encoding unit includes two encoding units. Each of the left-top corner direction encoding unit, the left-bottom corner direction encoding unit, the right-top corner direction encoding unit, and the right-bottom corner direction encoding unit is a unique two-bit binary code. The direction encoding unit, the identity encoding unit, and the redundancy encoding unit all represent binary encoding values ​​by setting a circular pattern with the opposite brightness attribute to the square unit they belong to. Setting the circular pattern represents the binary value 1, and not setting the circular pattern represents the binary value 0.

[0013] In a preferred embodiment of the present invention, the near-infrared light-emitting structure includes a diffuser plate, a light guide plate, a reflective film, near-infrared LED beads, and an aluminum support heat sink. The near-infrared LED beads are disposed on at least one side of the light guide plate for coupling near-infrared light from the side of the light guide plate into the interior of the light guide plate. The diffuser plate is disposed between the film plane target and the light guide plate. The reflective film is disposed on the side of the light guide plate opposite to the film plane target. The aluminum support heat sink is disposed on the side of the reflective film opposite to the light guide plate, and the aluminum support heat sink is an aluminum plate or an aluminum strip. The center wavelength of the near-infrared LED beads is 850nm.

[0014] By adopting the aforementioned design scheme, the beneficial effects of the present invention are: By establishing a one-to-one correspondence between each identity coding bit (excluding the central coding bit) and a redundant coding unit, and placing the corresponding coding units on both sides of the target center to form a spatially complementary layout, the occluded identity coding bit can be directly recovered from the corresponding redundant coding unit on its symmetrical side when unilateral occlusion occurs. This achieves precise fault tolerance at the coding bit level and significantly improves the continuity and robustness of target identity recognition under partial occlusion conditions. At the same time, this structure can maintain complete identity decoding when the unilateral occlusion area does not exceed 50% without increasing the number of cameras or additional targets, reducing system complexity and hardware costs, and providing a reliable coding hardware foundation for anti-occlusion localization in surgical navigation.

[0015] The purpose of this invention is to provide a method for determining the pose of an active near-infrared anti-occlusion planar target using spatial complementary redundancy coding. By using orientation recognition, mesh reconstruction, redundancy coding recovery, and residual corner point pose calculation, this method solves the problems of target identity loss and tracking interruption during intraoperative local occlusion, and improves tracking continuity and pose estimation robustness without increasing hardware.

[0016] To achieve the above objectives, the present invention adopts the following technical solution: A method for determining the pose of an active near-infrared anti-occlusion planar target using the above-mentioned spatial complementary redundancy coding includes the following steps: S1, acquire near-infrared images of the planar target using a near-infrared camera, and determine the region of interest of the planar target; S2, Detect unobstructed circular patterns and checkerboard corner points within the region of interest; S3. Based on the circular pattern, the corner points of the checkerboard grid, and the relationship between the light and dark pixels in the neighborhood of the corner points of the checkerboard grid, determine the direction of the planar target, and reconstruct the complete grid according to the preset row and column relationship of the checkerboard grid. S4. Based on the complete grid, determine the logical positions of each identity coding unit and each redundant coding unit, and determine whether each coding unit is in an identifiable or unidentifiable state. S5, for the corresponding identity coding units and redundant coding units, the corresponding identity coding bits are recovered according to the coding units that are in the identifiable state, and the identity coding bits are combined to obtain the identity coding of the planar target. S6. Calculate the spatial pose of the planar target relative to the near-infrared camera based on the image coordinates of the unobstructed checkerboard corner points, the known coordinates of the corresponding corner points in the planar target coordinate system, and the intrinsic parameters of the near-infrared camera.

[0017] In a preferred embodiment of the present invention, step S5 restores the identity code bits according to the following rules: When the corresponding identity coding unit and redundant coding unit are both in an identifiable state and have the same coding value, the coding value is confirmed to be the corresponding identity coding bit. When one of them is in an identifiable state and the other is in an unidentifiable state, the encoding value of the encoding unit in the identifiable state is determined as the corresponding identity encoding bit; When both are in an unrecognizable state, or both are in a recognizable state but their encoding values ​​are inconsistent, the current image frame is marked as an encoding recovery anomaly.

[0018] In a preferred embodiment of the present invention, step S3 includes: fitting the row direction and column direction of the chessboard grid according to the collinear relationship between the corner points of the visible chessboard grid and the spacing relationship between adjacent grid points; The orientation of the planar target is determined based on at least one identifiable orientation encoding unit, the relative positions between visible circular patterns, and the alternating light and dark relationships of the checkerboard pattern. Extend grid lines along the row and column directions to map the visible checkerboard corner points to a preset complete grid, thereby determining the logical row and column positions of each identity encoding unit and redundant encoding unit within the occluded area.

[0019] In a preferred embodiment of the present invention, when the unilateral continuous occlusion area does not exceed 50% of the effective coding pattern area of ​​the film planar target, the identity code and spatial pose of the planar target are output when the following conditions are met simultaneously: The central identity coding unit is in a recognizable state; At least one of the corresponding identity coding units and redundant coding units in each set is in a recognizable state; The orientation and complete grid of the planar target can be determined based on the remaining direction encoding and the corner points of the checkerboard grid. There are at least four non-collinear visible checkerboard corner points that can be correlated with known coordinates in the planar target coordinate system.

[0020] By adopting the aforementioned design scheme, the beneficial effects of the present invention are: By detecting the remaining circular patterns and checkerboard corner points and combining the brightness and darkness relationships of the corner point neighborhoods, the target orientation can be accurately determined and the complete mesh can be reconstructed when the target edge or corner is occluded, overcoming the limitation of ordinary planar targets relying on complete boundary orientation. Furthermore, by utilizing the one-to-one correspondence between the centrally symmetric distributed identity coding units and redundant coding units, when any coding bit is unrecognizable due to occlusion, it can be directly supplemented and restored using the corresponding visible coding value, achieving precise fault tolerance at the coding bit level and effectively avoiding the problem of target identity loss due to local occlusion. Finally, based on the remaining visible checkerboard corner points and their correspondence with the target coordinate system, the spatial pose is continuously calculated. Under the condition that the occlusion on one side does not exceed 50% of the effective area, it can still output stable and reliable position and attitude information. Thus, without increasing the number of cameras or targets, the tracking continuity and pose estimation robustness of the surgical navigation system in complex intraoperative occlusion environments are significantly improved.

[0021] The purpose of this invention is to provide an active near-infrared anti-occlusion surgical navigation system with spatial complementary redundancy coding. By linking a calibration plate with a robotic arm, multiple sets of images are acquired at once to simultaneously complete camera calibration and hand-eye calibration, quickly unifying various coordinate systems. This solves the problems of cumbersome calibration processes and excessively long preoperative preparation times in existing systems, achieving efficient and accurate system-level navigation and positioning.

[0022] To achieve the above objectives, the present invention adopts the following technical solution: An active near-infrared anti-occlusion surgical navigation system with spatial complementary redundancy coding includes a near-infrared camera assembly for capturing real-time position images of a calibration board or target. The calibration plate assembly, consisting of a calibration plate film and a near-infrared outer light source, is used in the calibration process of near-infrared cameras. The target, employing the aforementioned spatial complementary redundancy coding active near-infrared anti-masking planar target, is used to track the relative position of surgical instruments or patients. The robotic arm assembly changes the current position and orientation of a calibration plate or planar target by driving the movement of multi-axis joints; Computer components used to process images captured and transmitted back by the camera and run related algorithm programs; The calibration plate assembly is fixed to the end of the robotic arm, driving the robotic arm to change the spatial position and orientation of the calibration plate. At least 15 sets of calibration plate images at different positions are acquired by the near-infrared camera assembly. The computer calculates the internal and external parameters of the camera based on the calibration plate photos acquired by the camera, and unifies the coordinate systems of the camera, target and robotic arm.

[0023] In a preferred embodiment of the present invention, the near-infrared camera assembly comprises an industrial camera, an 850nm narrowband filter, and a lens; in the calibration plate assembly, the calibration plate film is printed with a checkerboard pattern, wherein the black part of the pattern is printed with a specific black ink and has the characteristic of being opaque; the near-infrared outer light source includes a diffuser plate, a light guide plate, a reflective film, an aluminum support heat sink, and several near-infrared LED beads; in the target assembly, the film planar target is used in conjunction with the near-infrared outer light source as the target tracked by the near-infrared camera.

[0024] By adopting the aforementioned design scheme, the beneficial effects of the present invention are: By fixing the calibration plate assembly to the end effector of the robotic arm and moving it in tandem, at least 15 sets of calibration plate images in different poses can be acquired simultaneously to complete both camera intrinsic parameter calibration and hand-eye calibration. This significantly simplifies the calibration process compared to the traditional method of performing camera and hand-eye calibration separately, and significantly shortens preoperative preparation time. Using the robotic arm base coordinate system as a unified world coordinate system, matrix operations are used to quickly unify the coordinate systems of the camera, calibration plate, and target, avoiding the accuracy loss and computational burden caused by multi-coordinate system transformation. At the same time, an anti-occlusion planar target is integrated as the tracking object, which can maintain stable tracking under local occlusion conditions during surgery without adding additional cameras or targets. This effectively reduces the system hardware cost and provides a highly efficient, accurate, low-cost, and interference-resistant system-level solution for surgical navigation. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the overall structure of the planar target of the present invention; Figure 2 This is a schematic diagram of the identification and coding area of ​​the planar target of the present invention; Figure 3 This is a schematic diagram of the redundant code region of the planar target of the present invention; Figure 4 This is a schematic diagram of the four directions of the planar target of the present invention; Figure 5 This is a schematic diagram of the normal detection process for the planar target of the present invention; Figure 6 These are schematic diagrams illustrating unilateral occlusion of the planar target of the present invention with different shapes and ranges; Figure 7 This is a schematic diagram of the identity code recovery process under the planar target occlusion state of the present invention. Detailed Implementation

[0026] 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.

[0027] like Figure 1 As shown, this embodiment provides an active near-infrared anti-obstruction planar target. The planar target includes a film planar target 1, a diffuser plate 2, a light guide plate 3, a reflective film 4, near-infrared LED beads 5, and an aluminum support heat sink 6; optionally, a hydrophobic transparent film is provided on the outer surface of the film planar target 1.

[0028] The active near-infrared anti-masking planar target has dimensions of 70mm × 60mm. The effective coding area of ​​film planar target 1 is 50mm × 50mm, consisting of 25 square units with a side length of 10mm, arranged in 5 rows × 5 columns. Adjacent square units are respectively set as high-brightness and low-brightness units in the near-infrared image, forming an alternating light and dark checkerboard pattern. The boundaries of the 5×5 square units form 36 grid intersection points in a 6×6 grid.

[0029] In one embodiment, the film target 1 is made of a film with a preset light-transmitting pattern. Positions on the film corresponding to high-brightness areas allow near-infrared light to pass through, while positions corresponding to low-brightness areas block or reduce the passage of near-infrared light.

[0030] The near-infrared emitting structure includes eight near-infrared LED beads 5 with a center wavelength of 850nm, a diffuser plate 2, a light guide plate 3, a reflective film 4, and an aluminum support heat sink 6. The eight near-infrared LED beads 5 are arranged at intervals along the side of the light guide plate 3 and connected in series. The rated forward voltage of each near-infrared LED bead 5 can be 1.5V, and the input voltage of the near-infrared emitting structure can be adjusted according to the camera exposure parameters. The light emitted by the near-infrared LED beads 5 enters the light guide plate 3 from the side. The reflective film 4 is disposed on the back of the light guide plate 3, reflecting the light facing the back back to the light guide plate 3. The diffuser plate 2 is disposed on the light-emitting surface of the light guide plate 3, homogenizing the light emitted from the light guide plate 3. A film target 1 is disposed on the outer side of the diffuser plate 2. The aluminum support heat sink 6 is disposed on the back of the reflective film 4, and can be made of aluminum plate or aluminum strip, used to fix the near-infrared emitting structure and conduct the heat generated by the near-infrared LED beads 5 during operation.

[0031] With the aforementioned side-incident light structure, near-infrared light can uniformly illuminate the film plane target 1 as a surface light source. When the target rotates at a large angle relative to the camera, the active light emission method can reduce the brightness attenuation caused by changes in reflection direction compared to the passive reflection method.

[0032] like Figures 2 to 4 As shown, the 5×5 square cell array is numbered from top to bottom as row 1 to row 5, and from left to right as column 1 to column 5.

[0033] Nine identity coding units 11 are located in the central 3×3 area formed by rows 2 to 4 and columns 2 to 4. The nine identity coding units 11 are numbered sequentially from bottom to top and from left to right within each row: the first identity coding unit is located at coordinates (4, 2); the second identity coding unit is located at coordinates (4, 3); the third identity coding unit is located at coordinates (4, 4); the fourth identity coding unit is located at coordinates (3, 2); the fifth identity coding unit is located at coordinates (3, 3); the sixth identity coding unit is located at coordinates (3, 4); the seventh identity coding unit is located at coordinates (2, 2); the eighth identity coding unit is located at coordinates (2, 3); and the ninth identity coding unit is located at coordinates (2, 4). These nine identity coding units 11 form a 9-bit binary identity code. When read in a preset order, it can represent a target identity number within the range of 0 to 511. In a specific example, the nine-bit code is 100000000. When the first identity coding unit 11 is the least significant bit and the ninth identity coding unit 11 is the most significant bit, the corresponding decimal identity number is 256.

[0034] Eight redundant coding units 12 are used to repeatedly represent the 1st, 2nd, 3rd, 4th, 6th, 7th, 8th, and 9th identity coding units 11, respectively. The correspondence is as follows:

[0035] Each set of identity coding units 11 and redundant coding units 12 are centrally symmetrically distributed with respect to the center of the chessboard (3,3). The fifth identity coding unit 11 at the center position does not have a corresponding redundant coding unit 12. Therefore, when it is necessary to fully recover the 9-bit identity code, the fifth identity coding unit 11 should remain identifiable.

[0036] This centrally symmetric distribution ensures that the identity encoding unit 11 located on one side of the film plane target 1 and the redundant encoding unit 12 located on the opposite side represent the same information. For example, when the left side of the film plane target 1 is occluded, the corresponding encoding unit on the right side can supplement the occluded encoding bits; when the right side of the film plane target 1 is occluded, the corresponding encoding unit on the left side performs restoration.

[0037] The orientation encoding unit 13 comprises eight encoding units, with each pair of units forming a corner orientation encoding unit. The positions of the four corner orientation encoding units are as follows: the upper left corner orientation encoding unit 131 is located at coordinates (1, 2) and (2, 1); the lower left corner orientation encoding unit 132 is located at coordinates (4, 1) and (5, 2); the upper right corner orientation encoding unit 133 is located at coordinates (1, 4) and (2, 5); and the lower right corner orientation encoding unit 134 is located at coordinates (4, 5) and (5, 4). Following a preset two-bit reading order, the lower left corner orientation encoding unit 132, the upper left corner orientation encoding unit 131, the upper right corner orientation encoding unit 133, and the lower right corner orientation encoding unit 134 represent 11, 10, 01, and 00, respectively. The four corner orientation encodings are distinct, enabling the image processing device to determine the orientation of the film plane target 1 after rotation by 0°, 90°, 180°, or 270°.

[0038] When a circular pattern with the opposite brightness attribute to that of a square cell is set within a coding unit, the coding unit represents binary 1; when no circular pattern is set within the coding unit, it represents binary 0. When the square cell is a bright cell in the near-infrared image, the circular pattern is set to a low-brightness circle; when the square cell is a low-brightness cell, the circular pattern is set to a high-brightness circle. This opposite brightness attribute is beneficial for identifying circular patterns through grayscale thresholding and circular detection algorithms.

[0039] like Figure 5 As shown, target detection under normal conditions includes the following steps.

[0040] Step 1, Image Acquisition. A near-infrared camera acquires a near-infrared image of the active near-infrared anti-masking planar target. The near-infrared camera may include an industrial camera, a lens, and a narrowband filter with a center wavelength of 850 nm.

[0041] Step 2, Region of Interest (ROI) Search. Considering that small targets occupy few pixels in the camera's field of view, the original image is first downsampled to reduce the number of pixels to be searched. Adaptive thresholding is then performed on the downsampled image to obtain highlight and low-light regions. Subsequently, eight-connected component analysis is performed, and candidate target regions are selected based on the area, aspect ratio, and checkerboard alternation features of the connected regions. The positions of the candidate target regions are mapped back to the original resolution image to obtain the ROI for film plane target 1.

[0042] Step 3, Corner Detection. Detect checkerboard corners within the region of interest and perform sub-pixel optimization of the corner coordinates. When the target is fully visible, sub-pixel image coordinates of 36 checkerboard corners (6×6 grid) can be obtained.

[0043] Step 4, Circular Pattern Detection. The Hough circle transform is used to detect circular patterns in the film plane target 1, obtaining the center coordinates and radius of each circular pattern. Based on the positional relationship of the circular pattern center within the corresponding square unit, the circular pattern is mapped to the orientation coding unit 13, the identity coding unit 11, or the redundancy coding unit 12.

[0044] Step 5, Direction Determination. Based on the circular pattern distribution in the upper left corner direction coding unit 131, lower left corner direction coding unit 132, upper right corner direction coding unit 133, and lower right corner direction coding unit 134, read four sets of two-digit direction codes. Determine the upper left, lower left, upper right, and lower right directions of the target according to the direction coding table, and assign uniform row and column numbers to the intersection of the 5×5 square unit and the 6×6 grid point.

[0045] Step 6, Identity Decoding. Read the binary codes of the nine identity coding units 11 within the central 3×3 area and convert them into decimal identity numbers. In the fully visible state, eight redundant coding units 12 are also used to verify the eight peripheral identity coding bits. When the corresponding identity coding bits and redundant coding bits are inconsistent, the current image frame is marked as having a coding anomaly.

[0046] Step 7, Pose Estimation. Based on the image coordinates of the 36 checkerboard corner points and their known planar coordinates in the target coordinate system, the position and pose of the planar target relative to the near-infrared camera are calculated using either a planar projection transformation method or a perspective n-point method.

[0047] like Figure 6 and Figure 7 As shown, the film-plane target 1 may be partially obscured by medical personnel, surgical instruments, or other objects. The obscured area can be triangular, trapezoidal, rectangular, or an irregular continuous region.

[0048] Under different occlusion conditions, the corner points of the visible checkerboard grid can form different distributions such as 5×6, 4×6, 4×4, 3×4, 4×3, 3×3, 5×4, and 6×3. The restoration steps after partial occlusion are as follows.

[0049] Step 1, Image Acquisition. A near-infrared camera acquires a near-infrared image of the active near-infrared anti-masking planar target. The near-infrared camera may include an industrial camera, a lens, and a narrowband filter with a center wavelength of 850 nm.

[0050] Step 2, Region of Interest (ROI) Search. Considering that small targets occupy few pixels in the camera's field of view, the original image is first downsampled to reduce the number of pixels to be searched. Adaptive thresholding is then performed on the downsampled image to obtain highlight and low-light regions. Subsequently, eight-connected component analysis is performed, and candidate target regions are selected based on the area, aspect ratio, and checkerboard alternation features of the connected regions. The positions of the candidate target regions are mapped back to the original resolution image to obtain the ROI for film plane target 1.

[0051] Step 3, Corner Detection. Detect checkerboard corners within the region of interest and perform sub-pixel optimization of the corner coordinates. When the target is fully visible, sub-pixel image coordinates of 36 checkerboard corners (6×6 grid) can be obtained.

[0052] Step 4: Detect remaining circular patterns. Detect unoccluded circular patterns within the region of interest. Unoccluded circular patterns include one or more directional coding circles, identity coding circles, and redundancy coding circles.

[0053] Step 5: Detect remaining corner points. Detect visible checkerboard corner points and perform sub-pixel optimization.

[0054] Step 6: Determine the grid orientation. Based on the collinearity and spacing between visible corner points, fit the orientations of two sets of intersecting grid lines. Adjacent corner points in the same row or column have a fixed spacing in the planar target coordinate system. The image processing device determines the candidate row and column orientations based on the projection relationships of the visible corner points.

[0055] Step 7: Determine the target orientation. Read the circular patterns in the visible direction coding group. When at least one direction coding group is fully visible, compare its two-bit code with the preset direction coding table to determine the actual corner corresponding to that direction coding group. When a direction coding group is partially occluded, the target orientation can be determined by combining the relative positions between visible circular patterns, the row and column structure formed by visible corner points, the alternation relationship between bright and dark pixels near the corner points, and the position of the known direction coding group in the 5×5 grid, eliminating direction candidates that do not conform to the coding and checkerboard brightness and darkness arrangement rules.

[0056] Step 8: Reconstruct the complete mesh. Based on the determined target orientation, the row and column orientations of visible corner points, and the fixed spacing between adjacent grid points, map the visible corner points to a complete 6×6 mesh. For occluded grid points, determine their predicted image positions based on the geometric relationships of the complete mesh, thereby establishing a complete position index for 5×5 square cells. Reconstructing the mesh is not equivalent to creating a fictitious image texture for the occluded area; rather, it determines the logical row and column positions of the occluded area within the complete target, for use in subsequent bit mapping and pose corner point matching.

[0057] After the complete grid is established, the visibility of the nine central identity coding units 11 and the eight peripheral redundant coding units 12 is determined one by one.

[0058] The coding unit can be judged as unrecognizable based on the following factors: the predicted area of ​​the coding unit is covered by an occluder; the effective visible area of ​​the coding unit is lower than a preset ratio; the confidence of the circular detection within the coding unit is lower than a preset threshold; or the local grayscale distribution of the coding unit does not meet the preset characteristics of the high-brightness or low-brightness unit.

[0059] For the i-th identity code bit, its identity code detection result is defined as follows: The corresponding redundancy code detection result is Where i∈{1,2,3,4,6,7,8,9}; and define the visibility flag of the encoding unit. , ∈{0,1}, where 1 indicates identifiable and 0 indicates unidentifiable.

[0060] The restored identity code bits are determined according to the following rules: When the identity coding unit 11 is identifiable and the redundant coding unit 12 is not identifiable, the coding value of the identity coding unit 11 is adopted; When the identity coding unit 11 is unrecognizable and the redundant coding unit 12 is recognizable, the coding value of the redundant coding unit 12 is adopted. When both are recognizable and have the same encoded value, confirm the encoded value. When neither can be recognized, the encoded bit is marked as invalid. When both are recognizable but have different encoded values, the current frame is marked as abnormal.

[0061] The fifth identity code bit in the center does not have a corresponding redundant coding unit 12. Its recovery result is directly taken as the detection value of the fifth identity code unit 11, and the fifth identity code unit 11 is required to be in a recognizable state.

[0062] When all nine identity code bits are valid, the target identity number is calculated according to the preset bit weights.

[0063] exist Figure 7 In one example shown, one side of the film plane target 1 is obscured by a rectangle, making the 3rd, 6th, and 9th identification coding units 11 unrecognizable. The image processing device reads the redundant coding units 12 that are centrally symmetrical to the 3rd, 6th, and 9th identification coding units 11, recovers the 3rd, 6th, and 9th identification coding bits, and obtains the complete binary code 100000000, corresponding to the target identification number 256.

[0064] The centrally symmetric redundant structure in this embodiment is mainly designed for unilateral continuous occlusion. Unilateral continuous occlusion refers to an occlusion area that extends mainly from the left, right, top, or bottom of the target into the target, or forms a triangular, trapezoidal, rectangular, or irregular continuous area mainly located on one side of the target.

[0065] When the effective area of ​​unilateral occlusion does not exceed 50% of the effective coding pattern area of ​​the film plane target 1, identity recovery can continue as long as the following conditions are met simultaneously: the 5th identity coding unit 11 is identifiable; at least one of each of the 1st, 2nd, 3rd, 4th, 6th, 7th, 8th and 9th identity coding units 11 is identifiable with the corresponding redundant coding unit 12; the target orientation can be determined based on the remaining directional coding units 13 and the checkerboard structure; and there are at least four non-collinear visible corner points used to establish planar projection relationships.

[0066] When occlusion occurs simultaneously on the left and right sides, or simultaneously on the top and bottom sides, a certain identity coding unit 11 and its centrally symmetric redundant coding unit 12 may be occluded at the same time. In this case, the corresponding coding bit cannot be completely recovered by the single-frame redundancy mechanism of this embodiment.

[0067] Therefore, "50% occlusion" should be understood as a unilateral occlusion condition that meets the above visibility conditions, rather than 50% occlusion of any shape or position.

[0068] This embodiment provides an active near-infrared anti-occlusion surgical navigation system with spatial complementary redundancy coding. The system includes: an image processing device, a near-infrared camera assembly for capturing images, a robotic arm assembly for assisting navigation and positioning, a calibration plate assembly for calibrating the intrinsic and extrinsic parameters of the near-infrared camera, a target assembly for tracking a target reference point, and a telescopic linkage assembly for changing the spatial position of the near-infrared camera.

[0069] The near-infrared camera assembly comprises an industrial camera, an 850nm narrowband filter, and a lens, all connected sequentially via a C-mount mount, effectively filtering out light from other wavelengths. As a result, images captured by the near-infrared camera assembly exhibit clear foreground and background separation; the foreground primarily consists of bright patterns such as calibration boards or targets reflecting near-infrared light, while the background is predominantly black. The industrial camera acquires images of the film-mounted target 1 in real time and transmits the images to the image processing unit.

[0070] The calibration plate assembly consists of an A4-sized film with a specific inkjet-printed checkerboard pattern and a 21×30cm rectangular near-infrared outer light source. The rectangular near-infrared outer light source comprises 180 near-infrared LEDs, a diffuser plate, a light guide plate, a reflective film, and an aluminum support and heat sink. The black areas of the film are printed with a specific black ink, which is opaque. The near-infrared outer light source allows near-infrared light to penetrate the texture of the film, obtaining an image used for near-infrared camera calibration.

[0071] The target assembly consists of a 6cm square inkjet-printed film with redundant encoded information and a 6×6cm square near-infrared outer light source. The square near-infrared outer light source comprises eight near-infrared LEDs, a diffuser plate, a light guide plate, a reflective film, and an aluminum support and heat sink. The film is printed with a black and white checkerboard pattern, with added circular patterns as redundant encoded information.

[0072] The robotic arm assembly changes the current position and orientation of the calibration plate assembly or target assembly by driving the movement of its multi-axis joints. A telescopic linkage assembly is fixed to the base of the robotic arm, and a near-infrared camera assembly is mounted on the telescopic linkage. It is worth noting that in this embodiment, the active near-infrared anti-obstruction planar target can be mounted on at least one of the following objects: surgical instruments, patient reference frames, implant positioning tools, robotic arm end effectors, or bone fixation frames.

[0073] The image processing device sequentially performs region of interest search, checkerboard corner detection, circular code detection, target orientation determination, occlusion mesh reconstruction, identity code recovery, planar target pose estimation, and coordinate system transformation.

[0074] The system specifically includes the following steps: S1, Image Acquisition: The calibration plate is fixed to the end effector of the robotic arm, which is then driven to change multiple spatial poses. A near-infrared camera acquires images of the calibration plate in different spatial poses and determines the region of interest (ROI) of the target structure. The near-infrared camera may include an industrial camera, a lens, and a narrowband filter with a center wavelength of 850nm. The specific implementation steps are as follows: S11: Complete system hardware assembly and alignment calibration, and stably fix the calibration board at the end effector position of the six-axis robotic arm to ensure that there is no relative displacement between the calibration board and the end effector of the robotic arm and that the installation is firm, so as to avoid shaking and posture deviation during the shooting process and ensure the accuracy of posture data acquisition; at the same time, complete the communication connection between the near-infrared camera, the robotic arm and the host computer, and adjust the camera imaging parameters and the robotic arm motion parameters to ensure normal linkage operation of the equipment.

[0075] S12. After the equipment is debugged, the host computer outputs precise control commands to drive the first to sixth axes of the robotic arm to rotate and fine-tune in a coordinated manner. This precisely changes the spatial position and posture of the end effector of the robotic arm, thereby driving the calibration plate fixed at the end effector to complete multi-dimensional, multi-angle, and multi-distance pose transformations in three-dimensional space. This allows the calibration plate to appear in different areas of the camera's field of view and at different depths of field, ensuring that the acquired image dataset has rich spatial diversity and covers the effective imaging field of view and working depth of field range of the camera.

[0076] S13: After each adjustment of the robotic arm to the correct position and the calibration board to a stable, undisturbed position, the computer synchronously sends a shooting command, driving the near-infrared camera to acquire and store a single image of the calibration board. Leveraging the anti-interference characteristics of its 850nm narrowband filter, and combined with the uniform illumination effect of the calibration board's active near-infrared outer light source, the near-infrared camera acquires high-quality calibration images with clear foreground textures, a clean and interference-free background, and extremely high contrast. The checkerboard corner points in the images are complete, without distortion or shadows, fully meeting the accuracy requirements of the calibration algorithm.

[0077] S14, repeat the above steps of robotic arm pose adjustment and image acquisition, continuously changing the spatial position and pitch and yaw attitude of the calibration plate, accumulating multiple sets of complete and valid calibration plate images to form a complete calibration image dataset. Multiple sets of images can construct sufficient overdetermined equations, effectively reducing the error in subsequent algorithm solutions and improving the accuracy and robustness of camera calibration and hand-eye calibration. After completing all image acquisition work, the calibration plate does not need to be retained and can be directly removed from the end effector of the robotic arm, completing the preoperative image acquisition process.

[0078] S2, Camera Calibration: Corner point recognition and extraction are performed on the acquired images using Zhang's calibration algorithm, and then the intrinsic and extrinsic parameters of each camera are calculated. The specific steps are as follows: S21, Region of Interest Search: Considering that small target structures occupy few pixels in the camera's field of view, the original image is first downsampled to reduce the number of pixels to be searched. Adaptive thresholding is performed on the downsampled image to obtain highlight and low-light regions. Subsequently, eight-connected component analysis is performed, and candidate target structure regions are selected based on the area, aspect ratio, and checkerboard alternation features of the connected regions. The positions of the candidate target structure regions are mapped back to the original resolution image to obtain the region of interest for film plane target 1.

[0079] S22, Corner Detection: Detect checkerboard corners within the region of interest and perform sub-pixel optimization of the corner coordinates. For each calibration board image, checkerboard corner detection is performed to obtain the sub-pixel image coordinates of the preset checkerboard corners. For each calibration board image, adaptive thresholding is used to separate the calibration board region from the background. The `findChessboardCorners` function in OpenCV is used to initially locate the checkerboard corners, and then the `cornerSubPix` function is used to obtain the sub-pixel precision corner coordinates, denoted as... ,in Indicates the first Zhang Image Indicates the first Corner points; S23, Camera Intrinsic Parameter Calibration: Let the homogeneous coordinates of the corner points on the calibration plate in the world coordinate system be... The camera imaging model is as follows: ; in: Scale factor; This is the camera intrinsic parameter matrix; and For the first The extrinsic parameters corresponding to the image; By utilizing multiple sets of corner point correspondences, the solution is obtained by minimizing the reprojection error. , and The LM algorithm is used for optimization, and the radial distortion coefficient is solved simultaneously. , ; S24, Output: Obtain the intrinsic parameter matrix of the near-infrared camera. Distortion coefficient , and the extrinsic parameter matrix corresponding to each image. This enables the mutual conversion between the camera coordinate system and the calibration plate coordinate system.

[0080] S3, Hand-eye calibration: The relative position information of the calibration plate in each set of images is collected, and the relative coordinate information between the calibration plate and the end of the robotic arm is obtained by using a calibration algorithm.

[0081] Let the pose of the robotic arm's end effector in the base coordinate system be... The data is read directly from the robotic arm controller. The pose of the calibration board in the camera coordinate system is The extrinsic parameter matrix is ​​given in step S23; The fixed conversion relationship between the camera and the end effector of the robotic arm is as follows: ; The fixed conversion relationship between the calibration plate and the end effector of the robotic arm is as follows: ; For the Group and No. The group of images has the following relationship: ; After transformation, we get: ; The equation is in the form of: Solve using the Tsai or Park method. .

[0082] Multiple equations can be constructed using multiple sets of images, and the optimal solution can be obtained through least squares optimization.

[0083] As a result, the relative coordinate system between the calibration plate and the robotic arm can be quickly obtained through the hand-eye calibration algorithm. .

[0084] S4, Coordinate Unification: In the coordinate unification unit, assuming the base of the robotic arm is the world coordinate system {W}, the position and orientation of the calibration plate relative to the base of the robotic arm, as well as the position and orientation of each near-infrared camera relative to the base of the robotic arm, are obtained through arithmetic operations of matrix inversion.

[0085] Specifically, for any point in space coordinates in the camera coordinate system Its coordinates transformed to the world coordinate system are: .

[0086] in: The extrinsic parameter matrix in step S23 and the matrix in step S3 The result of this joint derivation is that all coordinate systems are transformed into a coordinate system relative to the base of the robotic arm.

[0087] S5, Target Structure Registration: Select film planar target 1 with specific redundant coding information as the target to be tracked, and save the feature information of the pattern of the redundant coding information to the computer. The specific steps are as follows: S51, Obtain a standard template image of the target structure: Take a complete image of the target structure using a near-infrared camera assembly, and obtain a clear checkerboard and circular coding pattern after binarization.

[0088] S52, Decoding the symmetrical mapping relationship between the target structure ID and the storage center: According to Figures 2 to 4The encoding rule shown first detects the 36 corner points of a 5×5 chessboard to determine the boundaries of 25 square units. Then, it iterates through each square unit to determine if a circular pattern exists in its central area. If a circle exists, the corresponding bit is encoded as 1; otherwise, it is 0. The four corner orientation encoding units are used to determine the orientation of the target structure. The nine square units in the central 3×3 area are identity encoding units, which are converted into decimal numbers as the unique ID of the target structure. The eight outer square units are redundant encoding units, recording the centrosymmetric mapping relationship between each outer identity encoding unit (r, c) and its corresponding redundant encoding unit (6-r, 6-c), and storing this information in a computer database.

[0089] S53, storage template features: record the side length of the target structure, the known coordinates of each corner point of the 5×5 grid in the target structure coordinate system (with the center of the target structure as the origin), the binary value of each coding unit and the corresponding relationship, for subsequent tracking use.

[0090] S6, Target Structure Tracking: The computer reads the images captured by the near-infrared camera component in real time and searches the images for the existence of the target structure to be tracked. If no target structure is found, the process is skipped; if a target structure is found, the following anti-occlusion recognition and pose determination steps are performed: S61, Image Preprocessing and Candidate Region Detection: Gaussian filtering is applied to the real-time acquired image for noise reduction. Adaptive threshold segmentation is used to extract highlight regions. Contour analysis is used to select candidate regions with rectangular shapes and areas matching the target structure template. Perspective correction is performed on each candidate region to obtain a positive target structure image. Unoccluded circular patterns and checkerboard corner points are detected within the region of interest, and sub-pixel optimization is performed on the corner point coordinates.

[0091] S62, Target Structure Orientation Recognition and Complete Mesh Reconstruction: Based on the collinear relationship and equal spacing constraint between the corner points of the visible checkerboard grid, two sets of intersecting row and column directions are fitted. Adjacent corner points in the same row or column have a fixed spacing (10mm in this embodiment) in the plane target 1 coordinate system. The image processing device determines the candidate row and column directions based on the projection relationship of the visible corner points. Combining the circular pattern distribution in the visible direction coding units, the pixel relationship of alternating brightness near the visible corner points, and the preset position of the known direction coding group in the 5×5 grid, candidate directions that do not conform to the coding and checkerboard arrangement rules are eliminated, and the positive direction of the target structure is determined. The grid lines are extended along the determined row and column directions to map the visible corner points to the preset complete 6×6 grid, thereby determining the logical row and column positions of each identity coding unit and redundant coding unit in the occluded area. The purpose of reconstructing the grid is not to fabricate the image texture of the occluded area, but to determine the row and column indices of the occluded coding bits in the complete target structure for subsequent coding bit correspondence and pose corner point matching.

[0092] S63, Redundant Encoding and Decoding and Identity Recovery: After the complete grid is established, the visibility of the nine central identity encoding units and the eight peripheral redundant encoding units is determined one by one. An encoding unit can be judged as unrecognizable based on the following factors: the predicted area of ​​the encoding unit is covered by an occlusion; the effective visible area of ​​the encoding unit is lower than a preset proportion; the confidence level of the circular detection within the encoding unit is lower than a preset threshold; or the local grayscale distribution of the encoding unit does not meet the preset characteristics of high-brightness or low-brightness units. For the i-th identity encoding bit, its identity encoding detection result is defined as... The corresponding redundancy code detection result is Where i∈{1,2,3,4,6,7,8,9}; and define the visibility flag of the encoding unit. , ∈{0, 1}, where 1 indicates identifiable and 0 indicates unidentifiable. The recovered identity code bits are determined according to the following rules: When the identity coding unit is identifiable and the redundant coding unit is not identifiable, the coding value of the identity coding unit shall be used; When the identity coding unit is unidentifiable but the redundant coding unit is identifiable, the coding value of the redundant coding unit is used; When both are recognizable and have the same encoded value, confirm the encoded value. When both are unrecognizable, the encoded bit is marked as invalid, and the current frame is determined to have failed to recover. When both are recognizable but have different encoded values, the current frame is marked as an encoding anomaly and discarded.

[0093] The fifth identity coding bit in the center has no corresponding redundant coding unit. Its recovery result is directly taken from the detection value of the fifth identity coding unit, and the center bit is required to be in a recognizable state. If the center bit is occluded, the current frame is directly determined to be invalid. When all nine identity coding bits are valid, the complete target structure identity number is obtained according to the preset bit weight combination.

[0094] S64, pose calculation and coordinate unification: Extract the pixel coordinates of all unoccluded checkerboard corner points on the target structure that have established row and column mappings, and use the camera intrinsic parameter matrix. The distortion coefficients are then processed to remove distortion, and the pose of the target structure in the camera coordinate system is solved using the PnP algorithm. The three-dimensional coordinates of the target structure center are obtained. At this point, at least four non-collinear visible corner points are required. Finally, using the transformation relationships in the coordinate unification unit, the target structure coordinates are transformed to the world coordinate system: .Will The data is output in real time to the robotic arm or other downstream systems to complete the surgical navigation and positioning function. When the target structure is occluded by continuous occlusion on one side, the effective occlusion area does not exceed 50% of the effective coding pattern area of ​​the target structure, and the center identity coding bit is visible, the above-mentioned complete mesh reconstruction, redundancy recovery, and pose calculation process is executed; when the occlusion area exceeds 50% or the center position is occluded, the current frame tracking is determined to be unsuccessful and discarded, and the system waits for the next frame image.

[0095] During implementation, this device can clearly capture images of the calibration board and target structure using a monocular near-infrared camera assembly. By taking multiple images of the calibration board in different positions and orientations at once, the device can complete the calibration of the intrinsic and extrinsic parameters of the near-infrared camera assembly and the hand-eye calibration of the near-infrared camera assembly and the robotic arm assembly. This enables a rapid calibration process for the surgical navigation and positioning system, reducing the time spent on preoperative preparation. In the target structure tracking phase, through a centrally symmetric spatial complementary redundancy coding and mesh reconstruction algorithm, the device can accurately recover the target structure identity and continuously output six-degree-of-freedom pose even when unilateral local occlusion does not exceed 50% of the effective area and the center position is visible. This significantly improves the system's anti-interference capability in complex surgical environments and avoids tracking interruptions caused by local occlusion.

[0096] The above embodiment uses a 5×5 square unit, nine identity coding units 11 in the central 3×3 area and eight redundant coding units 12 on the periphery, which is a preferred embodiment of the present invention.

[0097] While maintaining the following core technical relationships, the coding pattern can undergo equivalent changes: identity coding unit 11 and redundant coding unit 12 represent the same target identity; at least some identity coding units 11 correspond one-to-one with spatially separated redundant coding units 12; the corresponding coding units are set in positions that are not easily covered by the same unilateral occlusion area; after occlusion, the coding position in the complete grid can be determined based on the remaining features; the missing identity coding bits are recovered based on the unoccluded corresponding coding units; and the target pose is calculated based on the remaining planar feature points.

[0098] The circular pattern in the encoding unit can also be replaced with an ellipse, a ring, or other simple geometric shape that can be reliably distinguished from the checkerboard background, but the ability to determine the binary encoded value through image processing should be maintained.

[0099] In the calibration plate assembly, the film size is not limited to A4 paper size, and the size of the near-infrared outer light source is not limited to 21×30cm; both can be adjusted according to actual calibration requirements. In the target assembly, the film size is not limited to 6cm×6cm, and the size of the near-infrared outer light source is not limited to 6×6cm; both can be adjusted according to actual tracking accuracy requirements. The number of near-infrared LED beads is not limited to 180 or 8; it can be increased or decreased according to illumination uniformity requirements.

[0100] This invention is not limited to orthopedic surgical robots, but can also be used in neurosurgical navigation, dental surgical navigation, interventional surgical positioning, industrial robot visual positioning, and other scenarios that require the identification and pose of planar targets under partial occlusion conditions.

[0101] During implementation, this device can clearly capture images of the calibration board and target structure using a monocular near-infrared camera simultaneously. By taking multiple images of the calibration board in different positions and orientations at once, the device can complete the calibration of the near-infrared camera's internal and external parameters, as well as the hand-eye calibration of the near-infrared camera and the robotic arm. This enables a rapid calibration process for the surgical navigation and positioning system, reducing the time spent on preoperative preparation. In the target structure tracking phase, through a centrally symmetric spatial complementary redundancy coding and mesh reconstruction algorithm, the device can accurately recover the target structure's identity and continuously output its six-degree-of-freedom pose, even when unilateral local occlusion does not exceed 50% of the effective area and the center remains visible. This significantly improves the system's anti-interference capability in complex surgical environments and avoids tracking interruptions caused by local occlusion.

[0102] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An active near-infrared anti-masking planar target with spatial complementary redundancy coding, characterized in that: The device includes a film planar target and a near-infrared emitting structure disposed on the backlight side of the film planar target. The film planar target has a checkerboard grid and is provided with multiple identity coding units, multiple redundant coding units and multiple directional coding units. Each of the identity coding units is used to represent a coding bit in the target identity coding. At least some of the identity coding units are respectively established in a one-to-one correspondence with one of the redundant coding units. The corresponding identity coding units and the redundant coding units represent the same binary coding value and are set at spatially separated positions on the film plane target. The corresponding identity coding units and the redundant coding units are located on opposite sides of the center of the film plane target, so that when the film plane target is partially occluded on one side, the corresponding identity coding bits are recovered based on a set of corresponding identity coding units and one unoccluded coding unit in the redundant coding units.

2. The active near-infrared anti-masking planar target with spatial complementary redundancy coding according to claim 1, characterized in that: The checkerboard grid comprises 25 square units in 5 rows × 5 columns, forming 36 grid intersections in 6 rows × 6 columns. Nine identity coding units are located in the 3 rows × 3 columns area in the center of the checkerboard grid. Eight redundant coding units correspond one-to-one with the eight identity coding units other than the central identity coding unit. When the row and column coordinates of an identity coding unit are (r, c), the row and column coordinates of its corresponding redundant coding unit are (6-r, 6-c), so that the corresponding identity coding units and redundant coding units are centrally symmetrically distributed with respect to the center of the checkerboard grid.

3. The active near-infrared anti-masking planar target with spatial complementary redundancy coding according to claim 1, characterized in that: The direction encoding unit includes a left-side direction encoding unit, a left-side direction encoding unit, a right-side direction encoding unit, and a right-side direction encoding unit respectively disposed at the four corners of the chessboard grid. Each of the left-side direction encoding unit, the left-side direction encoding unit, the right-side direction encoding unit, and the right-side direction encoding unit includes two encoding units. Each of the left-side direction encoding unit, the left-side direction encoding unit, the right-side direction encoding unit, and the right-side direction encoding unit is a unique two-bit binary code. The direction encoding unit, the identity encoding unit, and the redundancy encoding unit all represent binary encoding values ​​by setting a circular pattern with the opposite brightness attribute to the square unit they belong to. Setting the circular pattern represents the binary value 1, and not setting the circular pattern represents the binary value 0.

4. The active near-infrared anti-masking planar target with spatial complementary redundancy coding according to claim 1, characterized in that: The near-infrared emitting structure includes a diffuser plate, a light guide plate, a reflective film, near-infrared LED beads, and an aluminum support heat sink. The near-infrared LED beads are disposed on at least one side of the light guide plate to couple near-infrared light from the side of the light guide plate into the interior of the light guide plate. The diffuser plate is disposed between the film plane target and the light guide plate. The reflective film is disposed on the side of the light guide plate opposite to the film plane target. The aluminum support heat sink is disposed on the side of the reflective film opposite to the light guide plate, and the aluminum support heat sink is an aluminum plate or aluminum strip. The center wavelength of the near-infrared LED beads is 850nm.

5. A method for determining the pose of an active near-infrared anti-occlusion planar target using spatial complementary redundancy coding as described in any one of claims 1 to 4, characterized in that: Includes the following steps: S1, acquire near-infrared images of the planar target using a near-infrared camera, and determine the region of interest of the planar target; S2, Detect unobstructed circular patterns and checkerboard corner points within the region of interest; S3. Based on the circular pattern, the corner points of the checkerboard grid, and the relationship between the light and dark pixels in the neighborhood of the corner points of the checkerboard grid, determine the direction of the planar target, and reconstruct the complete grid according to the preset row and column relationship of the checkerboard grid. S4. Based on the complete grid, determine the logical positions of each identity coding unit and each redundant coding unit, and determine whether each coding unit is in an identifiable or unidentifiable state. S5, for the corresponding identity coding units and redundant coding units, the corresponding identity coding bits are recovered according to the coding units that are in the identifiable state, and the identity coding bits are combined to obtain the identity coding of the planar target. S6. Calculate the spatial pose of the planar target relative to the near-infrared camera based on the image coordinates of the unobstructed checkerboard corner points, the known coordinates of the corresponding corner points in the planar target coordinate system, and the intrinsic parameters of the near-infrared camera.

6. The pose determination method for an active near-infrared anti-occlusion planar target using spatial complementary redundancy coding according to claim 5, characterized in that: Step S5 restores the identity code bits according to the following rules: When the corresponding identity coding unit and redundant coding unit are both in an identifiable state and have the same coding value, the coding value is confirmed to be the corresponding identity coding bit. When one of them is in an identifiable state and the other is in an unidentifiable state, the encoding value of the encoding unit in the identifiable state is determined as the corresponding identity encoding bit; When both are in an unrecognizable state, or both are in a recognizable state but their encoding values ​​are inconsistent, the current image frame is marked as an encoding recovery anomaly.

7. The method for determining the pose of an active near-infrared anti-occlusion planar target using spatial complementary redundancy coding according to claim 5, characterized in that: Step S3 includes: fitting the row and column directions of the chessboard grid based on the collinearity between the corner points of the visible chessboard and the spacing between adjacent grid points; The orientation of the planar target is determined based on at least one identifiable orientation encoding unit, the relative positions between visible circular patterns, and the alternating light and dark relationships of the checkerboard pattern. Extend grid lines along the row and column directions to map the visible checkerboard corner points to a preset complete grid, thereby determining the logical row and column positions of each identity encoding unit and redundant encoding unit within the occluded area.

8. The method for determining the pose of an active near-infrared anti-occlusion planar target using spatial complementary redundancy coding according to claim 5, characterized in that: When the area of ​​continuous occlusion on one side does not exceed 50% of the effective coding pattern area of ​​the film planar target, the identity code and spatial pose of the planar target are output when the following conditions are met simultaneously: The central identity coding unit is in a recognizable state; At least one of the corresponding identity coding units and redundant coding units in each set is in a recognizable state; The orientation and complete grid of the planar target can be determined based on the remaining direction encoding and the corner points of the checkerboard grid. There are at least four non-collinear visible checkerboard corner points that can be correlated with known coordinates in the planar target coordinate system.

9. An active near-infrared anti-occlusion surgical navigation system with spatial complementary redundancy coding, characterized in that: Includes a near-infrared camera assembly for capturing real-time positional images of a calibration board or target; The calibration plate assembly, consisting of a calibration plate film and a near-infrared outer light source, is used in the calibration process of near-infrared cameras. The target, employing the active near-infrared anti-masking planar target with spatial complementary redundancy coding as described in any one of claims 1 to 4, is used to track the relative position of surgical instruments or patients. The robotic arm assembly changes the current position and orientation of a calibration plate or planar target by driving the movement of multi-axis joints; Computer components used to process images captured and transmitted back by the camera and run related algorithm programs; The calibration plate assembly is fixed to the end of the robotic arm, driving the robotic arm to change the spatial position and orientation of the calibration plate. At least 15 sets of calibration plate images at different positions are acquired by the near-infrared camera assembly. The computer calculates the internal and external parameters of the camera based on the calibration plate photos acquired by the camera, and unifies the coordinate systems of the camera, target and robotic arm.

10. The active near-infrared anti-occlusion surgical navigation system with spatial complementary redundancy coding according to claim 9, characterized in that: The near-infrared camera assembly consists of an industrial camera, an 850nm narrowband filter, and a lens; in the calibration plate assembly, the calibration plate film is printed with a checkerboard pattern, wherein the black part of the pattern is printed with a specific black ink, which has the characteristic of being opaque; the near-infrared outer light source includes a diffuser plate, a light guide plate, a reflective film, an aluminum support heat sink, and several near-infrared LED beads; in the target assembly, the film planar target is used in conjunction with the near-infrared outer light source as the target tracked by the near-infrared camera.

Images