A method for optical axis calibration of an optoelectronic tracker

Abstract

The application discloses a kind of optical axis calibration methods suitable for photoelectric tracker, this method is aimed at the precision of existing calibration technology Low, poor adaptability Problem, realize high-precision calibration by multi-source data fusion and adaptive algorithm optimization, calibration preparation stage sets up calibration target containing known three-dimensional coordinate characteristic mark point and completes equipment initialization;Multi-source data acquisition stage synchronously obtains calibration target image and characteristic point three-dimensional coordinate data and pre-processes;Characteristic extraction and matching stage is based on the improved SIFT algorithm Extraction characteristic point, establish image and three-dimensional coordinate mapping relationship in combination with FLANN matching algorithm;Optical axis offset calculation stage solves optical axis offset by perspective projection model and least square method;Optical axis adjustment and verification stage drives servo module to compensate offset and is confirmed accuracy by three rechecks.The calibration precision of the application reaches 0.1mrad, can work stably in low illumination, complex environment such as field, degree of automation is high and cost is controllable, suitable for the optical axis calibration scene of various photoelectric trackers.

Description

Technical Field

[0001] This invention relates to the field of photoelectric tracking technology, specifically to a method for calibrating the optical axis of a photoelectric tracker. Background Technology

[0002] Optical trackers are high-precision measurement and tracking devices widely used in aerospace, precision manufacturing, and other fields. Their core performance depends on the stability and accuracy of the optical axis. If the optical axis deviates, it will directly lead to decreased tracking accuracy, increased measurement errors, and even affect the normal operation of the equipment. Therefore, regularly calibrating the optical axis of the optical tracker is a necessary step to ensure its reliable operation.

[0003] Currently, the optical axis calibration methods for photoelectric trackers are mainly divided into two categories: manual calibration and automatic calibration.

[0004] Manual calibration primarily relies on operators visually observing and manually adjusting to achieve optical axis correction. While this method is simple to operate, it has significant limitations: calibration accuracy is greatly affected by subjective factors such as operator experience and fatigue, resulting in poor repeatability; the calibration process is time-consuming and inefficient; and it is difficult to meet the requirements of high-precision and high-reliability applications.

[0005] While automatic calibration methods have improved calibration efficiency to some extent, they still have the following problems:

[0006] 1. Reliance on high-precision external reference equipment: Some automatic calibration methods require high-precision external equipment such as laser trackers and coordinate measuring machines as references, resulting in high system costs. In addition, the calibration process has strict requirements on the environment (such as temperature, vibration, cleanliness, etc.), and has poor applicability in environments such as the field, mobile platforms or complex industrial sites.

[0007] 2. Insufficient stability of feature extraction: Automatic calibration methods based on image feature matching perform well in scenes with good lighting conditions and obvious features. However, in cases of low illumination, strong light interference, blurred target features, or cluttered backgrounds, feature extraction becomes difficult, the matching success rate is low, and the calibration accuracy decreases or even calibration fails.

[0008] 3. Weak environmental adaptability: Existing methods are often sensitive to conditions such as lighting, distance, and target attitude, and it is difficult to maintain stable calibration performance in dynamic or harsh environments.

[0009] 4. Complex system integration: Some calibration systems have complex structures, require the collaboration of multiple devices, and have cumbersome calibration processes, which are not conducive to rapid on-site deployment and maintenance.

[0010] In summary, existing optical axis calibration methods for photoelectric trackers generally suffer from limited calibration accuracy, poor environmental adaptability, high system cost, and complex operation procedures, making it difficult to meet the increasingly demanding application requirements for high precision, high reliability, and strong adaptability. Therefore, there is an urgent need for an optical axis calibration method that can achieve high precision, automation, and low cost in complex environments. Summary of the Invention

[0011] The purpose of this invention is to overcome the problems of low accuracy, poor adaptability, high cost, or complex operation in existing optical axis calibration methods for photoelectric trackers, and to provide an optical axis calibration method suitable for photoelectric trackers. This method can achieve high-precision calibration under different environmental conditions, has low equipment dependence, is easy to operate, and effectively improves calibration efficiency and reliability.

[0012] To achieve the above objectives, the present invention adopts the following technical solution:

[0013] A method for calibrating the optical axis of an optical tracker, the optical tracker comprising an optical imaging module, a laser ranging module, a servo drive module, and a data processing module, the method comprising the following steps:

[0014] Step 1: Calibration preparation stage: Set up a calibration target within the observation field of view of the photoelectric tracker. The calibration target has multiple feature markers with known three-dimensional coordinates; adjust the attitude of the photoelectric tracker and complete the equipment initialization, and preset the calibration parameters;

[0015] Step 2: Multi-source data acquisition stage: Control the optical imaging module to acquire target images, and simultaneously control the laser ranging module to acquire distance data of feature markers. Combine the photoelectric tracker attitude angle data to convert it into three-dimensional coordinate data in the device coordinate system; preprocess the image data and three-dimensional coordinate data.

[0016] Step 3: Feature extraction and matching stage: Based on the improved SIFT algorithm, extract the image coordinates and feature descriptors of feature markers in the image, and use the FLANN matching algorithm to filter feature point pairs with high matching degree to establish a mapping relationship set between image coordinates and three-dimensional coordinates;

[0017] Step 4: Optical axis offset calculation stage: Construct an imaging geometric model based on the perspective projection model, substitute the mapping relationship into the model and solve it using the least squares method to obtain the theoretical direction vector of the current optical axis; calculate the angle between the theoretical direction vector and the design reference direction vector to obtain the optical axis offset;

[0018] Step 5: Optical axis adjustment stage: Adjust the attitude of the optical components of the photoelectric tracker by controlling the servo drive module according to the optical axis offset;

[0019] Step 6: Optical axis verification stage: Repeat the calibration process of steps 2-5 until the optical axis offset is less than the preset calibration accuracy threshold.

[0020] Furthermore, in step 1, the calibration target is a checkerboard target or a dot array target, with no less than 9 feature markers that are evenly distributed; the three-dimensional coordinates of the feature markers are calibrated by a laser tracker or a coordinate measuring machine, with a measurement accuracy of no less than ±0.01mm.

[0021] Furthermore, in step 2, the preprocessing includes: using a bilateral filtering algorithm to denoise the image, using an adaptive histogram equalization algorithm to enhance the image grayscale contrast; using a Kalman filtering algorithm to filter the three-dimensional coordinate data, and using the 3σ criterion to remove outliers.

[0022] Furthermore, in step 3, the improved SIFT algorithm enhances the stability of feature extraction in low-light environments by using adaptive Gaussian filtering parameters. The formula for calculating the filter kernel size k is:

[0023]

[0024] in, Based on the size of the filter kernel, To adjust the coefficient, The average gray level of the current image. The standard grayscale mean. This represents the current image grayscale variance.

[0025] Furthermore, in step 4, the expression for the imaging geometric model is:

[0026]

[0027]

[0028] in The image coordinates of the feature marker points, To correct the distorted three-dimensional coordinates of the feature marker points in the equipment coordinate system, The focal length of the optical imaging module. These are the coordinates of the principal point of the optical imaging module.

[0029] Furthermore, in step 4, the optical axis offset includes a horizontal offset. and vertical offset The vector dot product formula is used for calculation:

[0030]

[0031]

[0032] in, Theoretical direction vector , Design reference direction vector .

[0033] Furthermore, the preset calibration parameters in step 1 include: image acquisition frame rate of 10-25fps, laser ranging frequency of 20-50Hz, feature matching threshold of 0.6-0.8, and calibration accuracy threshold of 0.05-0.1mrad.

[0034] Furthermore, in step 3, after the FLANN matching algorithm filters matching point pairs, the RANSAC algorithm is further used to remove erroneous matching points, thereby improving the matching accuracy.

[0035] Furthermore, in step 4, lens distortion compensation is introduced when constructing the imaging geometry model. The lens distortion includes radial distortion and tangential distortion, and the distortion coefficients are obtained in advance through camera calibration.

[0036] Furthermore, in step 6, the optical axis verification stage adopts the principle of three-time re-inspection, that is, when the optical axis offset obtained by repeating steps 2-5 three times in a row is less than the preset calibration accuracy threshold, the calibration is determined to be completed.

[0037] Beneficial effects:

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] 1. High calibration accuracy: This invention constructs a multi-source data collaborative calibration system by fusing optical imaging data and laser ranging data, effectively suppressing measurement errors from a single data source; the improved SIFT algorithm (adaptive Gaussian filter parameter) can still stably extract feature points in low-light environments, and the least squares method is used to accurately solve the imaging geometric model, so that the optical axis calibration accuracy can stably reach below 0.1 mrad, meeting the requirements of high-precision applications.

[0040] 2. Strong environmental adaptability: The improved SIFT algorithm dynamically adjusts the filter kernel size based on the mean and variance of image grayscale, enabling stable extraction of feature points in low-light and strong interference environments; the calibration process does not rely on complex external reference equipment, but can be completed with only a simple calibration target, making it suitable for various operating scenarios such as indoor and outdoor environments.

[0041] 3. High degree of automation: From data acquisition and feature matching to optical axis adjustment, the entire calibration process is fully automated and requires no manual intervention, which effectively improves calibration efficiency and reduces the calibration time per session from more than 30 minutes under the existing technology to less than 5 minutes.

[0042] 4. Controllable Cost: The calibration target uses a simple checkerboard or dot array target. The coordinates of the feature points can be calibrated once using conventional high-precision equipment (such as laser trackers) and reused, avoiding continuous dependence on expensive external reference equipment and reducing equipment investment and maintenance costs. This invention is compatible with visible light and infrared imaging modules and can be directly integrated into existing photoelectric tracking systems, demonstrating good economic efficiency and promotional value.

[0043] 5. High system reliability: Through the "three-time re-inspection" mechanism and overcompensation adjustment strategy, the calibration results are ensured to be stable and reliable. If multiple adjustments still fail to meet the standards, the system will automatically trigger a fault alarm and prompt hardware checks to avoid miscalibration due to equipment abnormalities, thereby improving the robustness and security of the system. Attached Figure Description

[0044] Figure 1 This is a flowchart of the optical axis calibration method for the photoelectric tracker in the embodiment.

[0045] Figure 2 This is a schematic diagram of the system layout of the photoelectric tracker and the calibration target in the embodiment. In the figure: 1-photoelectric tracker, 11-optical imaging module, 12-laser ranging module, 13-servo drive module, 14-data processing module, 2-calibration target, 3-observation field of view, 4-equipment coordinate system (X-axis horizontal to the right, Y-axis vertical upward, Z-axis pointing to the calibration target).

[0046] Figure 3 This is a schematic diagram of the target structure in the embodiment. In the figure: 5-aluminum alloy substrate (size 1m×1m), 6-circular feature mark (diameter 5mm, sprayed with matte black), 7-high gloss white background, 8-three-dimensional coordinate annotation of feature mark (example: P1(X1,Y1,Z1)). Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it.

[0048] The optical axis calibration method for photoelectric trackers proposed in this embodiment is applicable to photoelectric tracker systems including an optical imaging module, a laser ranging module, a servo drive module, and a data processing module. The optical imaging module can be a visible light camera or an infrared thermal imager. The optical axis calibration principle is the same for both imaging modules; the difference lies in the fact that a heat source calibration target is required for infrared thermal imagers. The technical solution of this invention will be described in detail below using a visible light camera as an example, in conjunction with the accompanying drawings.

[0049] In this embodiment, the optical imaging module of the photoelectric tracker uses a CMOS camera with a resolution of 1920×1080 and a focal length of f=25mm, and the principal point coordinates are... The laser ranging module has a measurement range of 20-100m and a measurement accuracy of ±0.1mm; the servo drive module has a control accuracy of 0.01mrad; the calibration target is a 10×10 checkerboard target with a checkerboard side length of 50mm. The three-dimensional coordinates of each feature marker point (checkerboard intersection) are pre-calibrated by the laser tracker (measurement accuracy ±0.01mm) and stored in the data processing module.

[0050] like Figure 1 As shown in the figure, the optical axis calibration method for photoelectric trackers proposed in this embodiment includes the following steps:

[0051] Step 1: Calibration Preparation Stage

[0052] Step 1.1: Fix the calibration target 50m away from the photoelectric tracker, ensuring it is within the tracker's field of view and unobstructed. The calibration target is made of an aluminum alloy substrate, 1m x 1m in size, with a high-contrast pattern formed by a glossy white background and matte black feature markers. The feature markers are circular, 5mm in diameter, with a 50mm center-to-center spacing between adjacent markers, arranged in a positive grid. The three-dimensional coordinates of each feature marker are pre-calibrated using a laser tracker in a standard laboratory environment, with a measurement accuracy of ±0.01mm. The calibration results are stored in the feature database of the data processing module. During installation, the calibration target must be leveled to ensure its plane is perpendicular to the theoretical direction of the photoelectric tracker's optical axis, with a perpendicularity error ≤0.05°.

[0053] Step 1.2: Start the photoelectric tracker. Adjust the azimuth and elevation angles of the device through the control interface. Using the real-time preview screen of the optical imaging module, ensure the target image occupies 60%-80% of the field of view, guaranteeing complete imaging of all feature markers. Start the optical imaging module to calibrate white balance and exposure parameters, the laser ranging module to correct baseline errors, and the data processing module to complete memory initialization and algorithm loading. After device initialization, all module status indicator lights should be constantly lit, with no fault alarm signals.

[0054] Step 1.3: Preset calibration parameters through the human-computer interaction interface of the data processing module. The image acquisition frame rate is dynamically set according to the distance to the calibration target. It is set to 10fps when the distance is ≤50m and increased to 25fps when the distance is >50m. The laser ranging frequency is set to twice the image acquisition frame rate to ensure that there are two sets of distance data for each imaging moment. The feature matching threshold is set to 0.7 (distance ratio method based on FLANN algorithm). The calibration accuracy threshold is set according to the application scenario of the photoelectric tracker. It is set to 0.05mrad for precision measurement scenario and 0.1mrad for ordinary tracking scenario.

[0055] Step 2: Multi-source data acquisition stage:

[0056] Step 2.1: Control the image acquisition card of the optical imaging module to continuously acquire images of the calibration target at a preset frame rate of 10fps. During the acquisition process, synchronization with the servo drive module is achieved through the trigger signal of the image acquisition card. 20 frames of calibration target images are continuously acquired. After format conversion (from RAW to BMP format), the image data is transferred to the buffer of the data processing module via DMA to avoid consuming CPU resources. Each frame of image is accompanied by an acquisition timestamp with an accuracy of ≤1μs for subsequent data synchronization.

[0057] Step 2.2: Based on the trigger signal of the image acquisition card, the laser ranging module is synchronously controlled to perform laser ranging on each feature marker point in a "point scan" mode. The ranging sequence is executed from left to right and from top to bottom according to the grid coordinates of the feature marker points. Ranging is performed on all 25 feature marker points of the checkerboard target, with each marker point measured 3 times. The average value is taken as the final distance data. The encoder of the servo drive module collects the azimuth angle (accuracy 0.001°) and pitch angle (accuracy 0.001°) data of the photoelectric tracker in real time. After associating the distance data with the same timestamp, the data is substituted into the coordinate transformation formula:

[0058] X=R×sinA×cosE, Y=R×sinE, Z=R×cosA×cosE

[0059] Where R represents distance data, A represents azimuth angle, and E represents elevation angle. The converted three-dimensional coordinate data (X, Y, Z) of the device coordinate system is transmitted to the data processing module and stored corresponding to the image data.

[0060] Step 2.3: The data processing module preprocesses the received image data and 3D coordinate data. Image denoising uses a bilateral filtering algorithm with a filter window size of 5×5, a spatial Gaussian standard deviation σ_d=5, and a gray-level similarity Gaussian standard deviation σ_r=20, removing high-frequency noise while preserving feature edge information. Gray-level enhancement uses an adaptive histogram equalization algorithm, dividing the image gray levels into 16 regions, with a contrast limiting factor of 2.0 for each region to avoid local over-enhancement. Coordinate data filtering uses a Kalman filtering algorithm, with both process noise and observation noise set to Gaussian white noise. Outlier removal uses the 3σ criterion, calculating the mean and standard deviation of the coordinate data, marking data exceeding the 3σ criterion range as outliers and removing them, while recording the outlier locations for avoidance during subsequent feature matching. In this embodiment, two outlier data are removed.

[0061] Step 3: Feature extraction and matching stage:

[0062] Step 3.1: Extract features from the preprocessed calibration target image using the improved SIFT algorithm. The algorithm flow includes:

[0063] 1) Scale Space Construction: To simulate the multi-scale features of an image, the scale space of the image is first constructed. The scale space is generated by performing a series of Gaussian filters at different scales on the image, i.e.

[0064]

[0065] in This is a Gaussian kernel function. The improvement of this invention lies in the kernel size of the Gaussian filter. It is not a fixed value, but rather adaptively determined based on the image's lighting conditions. Specifically, the mean gray value of the current image (or image region) is calculated before filtering. With variance Substitute into the adaptive formula:

[0066]

[0067] in, Based on the size of the filter kernel, To adjust the coefficient, The standard grayscale mean (usually 128) is used. When the image's grayscale mean is low (i.e., in low-light environments) and the variance is large, the filter kernel size is adjusted. The corresponding increase is used to suppress noise and enhance feature stability; conversely, under conditions of sufficient lighting and clear images, The value is set to a smaller value to retain more detailed information. Through the above adaptive adjustment, high-quality scale space can be constructed under different environmental conditions.

[0068] Subsequently, a Gaussian difference pyramid was constructed based on the Gaussian filtering results at different scales. The pyramid has a total of 6 layers, with each group containing 5 layers. The standard deviation of the Gaussian kernel increases from 1.6, which is used for subsequent keypoint detection.

[0069] 2) Key point detection: By comparing the gray values ​​of adjacent pixels in the Gaussian difference image, local extreme points are identified as candidate key points, and then low contrast and edge response points are eliminated by fitting a three-dimensional quadratic function.

[0070] 3) Direction assignment: Calculate the gradient direction histogram in a 16×16 neighborhood with the key point as the center, and take the direction corresponding to the peak value in the histogram as the main direction of the key point to ensure the rotation invariance of the feature descriptor.

[0071] 4) Feature descriptor generation: Divide the neighborhood of key points into 4×4 sub-regions, calculate the gradient histogram in 8 directions for each sub-region, and generate a 128-dimensional feature descriptor.

[0072] In this embodiment, based on the image grayscale mean μ=85, the standard grayscale mean... =128, grayscale variance σ=32, take , Substituting into the formula: k=3+0.1×(128-85) / 32≈4.34, we take k=4 and perform Gaussian filtering based on this filter kernel to ensure that stable feature points can still be extracted in low-light (illuminance ≤50 lux) environments.

[0073] Step 3.2: Use the FLANN matching algorithm to construct a KD tree index to perform cross-frame matching of feature markers in each frame of the image.

[0074] The matching process is divided into two steps: preliminary matching and precise matching. Preliminary matching adopts the nearest neighbor matching strategy, calculates the Euclidean distance between the feature descriptor to be matched and the target feature descriptor, and retains the candidate matching pair with the smallest distance. Preliminary matching introduces the distance ratio method, calculates the ratio of the nearest neighbor distance to the second nearest neighbor distance in the candidate matching pair. When the ratio is less than the preset feature matching threshold of 0.7, it is determined to be a valid matching pair; otherwise, false matching points are eliminated.

[0075] To further improve matching accuracy, the RANSAC algorithm was used to filter effective matching pairs. The number of iterations was set to 1000, the inlier threshold was 2 pixels, and the inlier rate of the final retained matching pairs was ≥90%.

[0076] Step 3.3: Determine the precise image coordinates (u, v) of the matched feature points based on the circular structural features of the feature markers. These coordinates have the top-left corner of the image as the origin, with the horizontal axis (u) and the vertical axis (v). The Hough circle transform is used to extract the center coordinates, with a circle detection radius of 4-6 mm and an accumulator threshold of 100, achieving an extraction accuracy of 0.1 pixels. The precise image coordinates (u, v) are then correlated one-to-one with the corresponding feature marker device coordinates (X, Y, Z) stored in the data processing module, establishing a mapping relationship set between image coordinates and 3D coordinates. In this embodiment, 23 sets of valid mapping relationships are established, satisfying the requirement that the number of samples is not less than 80% of the total number of feature markers.

[0077] Step 4: Calculation of optical axis offset:

[0078] Step 4.1: Construct the imaging geometric model of the photoelectric tracker based on the perspective projection model. This model takes into account the lens distortion error of the optical imaging module, including radial distortion and tangential distortion.

[0079] The formula for radial distortion correction is:

[0080]

[0081]

[0082] The formula for tangential distortion correction is:

[0083]

[0084]

[0085] The coordinates after total distortion correction are:

[0086]

[0087]

[0088] in These are the coordinates of the image before distortion. The image coordinates after distortion correction are the actual projected positions of the feature markers on the image after considering both radial and tangential lens distortion. The distance from the feature point to the image center. The radial distortion coefficient is... The tangential distortion coefficient is obtained in advance through camera calibration.

[0089] The corrected imaging geometry model expression is:

[0090]

[0091]

[0092] in The image coordinates of the feature marker points, To correct the distorted three-dimensional coordinates of the feature marker points in the equipment coordinate system, The focal length of the optical imaging module (corrected for temperature, with a compensation coefficient of -0.001 mm / ℃). In this embodiment, the principal point coordinates of the optical imaging module are... , , .

[0093] Step 4.2: Convert the image coordinates from the 23 sets of mapping relationships established in Step 3. With the corrected three-dimensional coordinates Substituting into the imaging geometry model, we construct the error equation:

[0094]

[0095]

[0096] in The image coordinates observed in step 3.3 are used; the objective function is established by solving the model parameters using the least squares method with the goal of minimizing the sum of squared errors:

[0097]

[0098] Through the objective function , , After taking the partial derivatives of the optical axis direction parameters and setting them equal to 0, the system of equations is obtained and solved using the LU decomposition method. Finally, the theoretical direction vector of the current optical axis in the device coordinate system is obtained. The vector has been normalized and satisfies In this embodiment, 23 sets of mapping relationships are substituted into the model and solved using the least squares method to obtain the theoretical direction vector of the current optical axis n=(0.0021, -0.0018, 0.999996);

[0099] Step 4.3: Retrieve the design reference direction vector of the optical axis from the device parameter library of the photoelectric tracker. In this embodiment, we take This vector is determined and stored by precision optical calibration at the time the device leaves the factory.

[0100] The theoretical direction vector is calculated using the dot product formula. relative to the design reference direction vector The angle between them, including the horizontal offset and vertical offset :

[0101]

[0102]

[0103] In the horizontal offset Reflects the offset of the optical axis in the horizontal plane (XZ plane), the vertical offset amount It reflects the offset of the optical axis in the vertical plane (YZ plane); in this embodiment:

[0104]

[0105]

[0106] A temperature compensation factor is also introduced during the calculation process. The direction vector is corrected based on the real-time temperature value collected by the ambient temperature sensor (accuracy ±0.1℃). The correction coefficient is obtained by fitting experimental data to ensure the accuracy of offset calculation under different temperature environments.

[0107] Step 5: Optical axis adjustment stage:

[0108] The data processing module will calculate the horizontal offset of the optical axis. and vertical offset The control signals are converted into pulse control signals according to the control protocol of the servo drive module. The pulse equivalent of the azimuth axis is 0.001° / pulse, and the pulse equivalent of the pitch axis is 0.001° / pulse. These are converted into control signals: horizontal adjustment -0.12mrad and vertical adjustment 0.10mrad. The control signals are transmitted to the servo drive module via the CAN bus, driving the stepper motors of the azimuth and pitch axes to adjust the attitude. The adjustment process adopts a segmented uniform speed control strategy. In the initial stage, the speed is 1° / s to quickly approach the target position. When the offset is less than 0.5mrad, the speed is switched to low-speed fine adjustment of 0.1° / s to avoid overshoot. After the adjustment is completed, the servo drive module feeds back the position signal, and the encoder collects the adjusted axis angle position in real time and transmits it to the data processing module.

[0109] Step 6: Optical Axis Verification Stage

[0110] Step 6.1: Repeat steps 2-5 to recheck the optical axis offset. During the recheck, keep the calibration target position and calibration parameters unchanged, shorten the data acquisition time to 50% of the initial acquisition time, and focus on acquiring data from the core area of ​​the feature marker points. After the recheck is completed, the data processing module calculates the adjusted optical axis offset to obtain the adjusted horizontal offset. Vertical offset All of them are less than the preset calibration accuracy threshold of 0.1 mrad.

[0111] Step 6.2: To ensure the reliability of the verification, the "three-time re-inspection" principle is adopted, meaning that the offset in three consecutive re-inspections must meet the accuracy requirements. In this embodiment, the results of the three consecutive re-inspections all meet the requirements. and The calibration is complete. The data processing module automatically generates a calibration report, which includes information such as calibration time, ambient temperature, data collection volume at each stage, initial and final offset values, and adjustment parameters. The report is stored in the local database and simultaneously uploaded to the cloud server.

[0112] If the re-inspection fails, return to step 5. Calculate a new adjustment based on the offset from the previous re-inspection. The adjustment value is 1.2 times the current offset (i.e., by introducing an overcompensation coefficient to accelerate convergence). Repeat this adjustment until the accuracy requirements are met after three consecutive re-inspections. If the accuracy is still not met after more than five adjustments, a fault alarm signal is triggered, prompting the operator to check the equipment hardware status.

[0113] In this embodiment, the calibration process took 4.2 minutes and the calibration accuracy reached 0.08 mrad. Stable feature extraction and matching were still achieved under low illumination conditions (ambient light intensity of 50 lux), and the calibration effect was better than the existing technology.

[0114] It should be noted that the above embodiments are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications or equivalent substitutions made by those skilled in the art to the technical solutions of the present invention without departing from the design concept of the present invention should fall within the scope of protection of the present invention.

Claims

1. A method for calibrating the optical axis of a photoelectric tracker, the photoelectric tracker comprising an optical imaging module, a laser ranging module, a servo drive module, and a data processing module, characterized in that: The method includes the following steps: Step 1: Calibration preparation stage: Set up a calibration target within the observation field of view of the photoelectric tracker. The calibration target has multiple feature markers with known three-dimensional coordinates; adjust the attitude of the photoelectric tracker and complete the equipment initialization, and preset the calibration parameters; Step 2: Multi-source data acquisition stage: Control the optical imaging module to acquire target images, and simultaneously control the laser ranging module to acquire distance data of feature markers. Combine the photoelectric tracker attitude angle data to convert it into three-dimensional coordinate data in the device coordinate system; preprocess the image data and three-dimensional coordinate data. Step 3: Feature extraction and matching stage: Based on the improved SIFT algorithm, extract the image coordinates and feature descriptors of feature markers in the image, and use the FLANN matching algorithm to filter feature point pairs with high matching degree to establish a mapping relationship set between image coordinates and three-dimensional coordinates; Step 4: Optical axis offset calculation stage: Construct an imaging geometric model based on the perspective projection model, substitute the mapping relationship into the model and solve it using the least squares method to obtain the theoretical direction vector of the current optical axis; calculate the angle between the theoretical direction vector and the design reference direction vector to obtain the optical axis offset; Step 5: Optical axis adjustment stage: Adjust the attitude of the optical components of the photoelectric tracker by controlling the servo drive module according to the optical axis offset; Step 6: Optical axis verification stage: Repeat the calibration process of steps 2-5 until the optical axis offset is less than the preset calibration accuracy threshold.

2. The optical axis calibration method for a photoelectric tracker according to claim 1, characterized in that, In step 1, the calibration target is a checkerboard target or a dot array target, with no less than 9 feature markers that are evenly distributed; the three-dimensional coordinates of the feature markers are calibrated by a laser tracker or a coordinate measuring machine, with a measurement accuracy of no less than ±0.01mm.

3. The optical axis calibration method for a photoelectric tracker according to claim 1, characterized in that, In step 2, the preprocessing includes: using a bilateral filtering algorithm to denoise the image, using an adaptive histogram equalization algorithm to enhance the grayscale contrast of the image, using a Kalman filtering algorithm to filter the three-dimensional coordinate data, and using the 3σ criterion to remove outliers.

4. The optical axis calibration method for a photoelectric tracker according to claim 1, characterized in that, In step 3, the improved SIFT algorithm enhances the stability of feature extraction in low-light environments by using adaptive Gaussian filtering parameters. The formula for calculating the filter kernel size k is: in, Based on the size of the filter kernel, To adjust the coefficient, The average gray level of the current image. The standard grayscale mean. This represents the current image grayscale variance.

5. The optical axis calibration method for a photoelectric tracker according to claim 1, characterized in that, In step 4, the expression for the imaging geometric model is: in The image coordinates of the feature marker points, To correct the distorted three-dimensional coordinates of the feature marker points in the equipment coordinate system, The focal length of the optical imaging module. These are the coordinates of the principal point of the optical imaging module.

6. The optical axis calibration method for a photoelectric tracker according to claim 1, characterized in that, In step 4, the optical axis offset includes a horizontal offset. and vertical offset The vector dot product formula is used for calculation: in, Theoretical direction vector , Design reference direction vector .

7. The optical axis calibration method for a photoelectric tracker according to claim 1, characterized in that, The preset calibration parameters in step 1 include: image acquisition frame rate 10-25fps, laser ranging frequency 20-50Hz, feature matching threshold 0.6-0.8, and calibration accuracy threshold 0.05-0.1mrad.

8. The optical axis calibration method for a photoelectric tracker according to claim 1, characterized in that, In step 3, after the FLANN matching algorithm filters matching point pairs, the RANSAC algorithm is further used to remove erroneous matching points, thereby improving the matching accuracy.

9. The optical axis calibration method for a photoelectric tracker according to claim 1, characterized in that, In step 4, lens distortion compensation is introduced when constructing the imaging geometry model. The lens distortion includes radial distortion and tangential distortion, and the distortion coefficients are obtained in advance through camera calibration.

10. The optical axis calibration method for a photoelectric tracker according to claim 1, characterized in that, In step 6, the optical axis verification stage adopts the principle of three-time re-inspection, that is, when the optical axis offset obtained by repeating steps 2-5 three times in a row is less than the preset calibration accuracy threshold, the calibration is determined to be completed.

Images