A hole positioning method for automobile parts processing

Abstract

The present invention relates to the field of image processing, and more specifically to a method for hole positioning in automotive parts processing. The method involves acquiring an image of the area to be holed on a workpiece and calibrating a ROI region. The ORB algorithm is then used to detect key points and main directions on the ROI image. The core step is to calculate the local illumination perturbation for each key point neighborhood image block, dynamically adjust the normalization range accordingly, and perform intensity remapping on the image block, generating an ICE-rBRIEF feature descriptor on this basis. The generated descriptor is then matched with a reference model. A local non-uniform two-dimensional deformation field model is established using high-confidence matching points. Finally, the theoretical hole position is accurately compensated to guide the operation of the hole-forming mechanism. The present invention analyzes the local illumination perturbation characteristics, adaptively adjusts the intensity normalization range of the image block, generates the ICE-rBRIEF feature descriptor, and combines it with deformation compensation to achieve high-precision hole positioning.

Description

Technical Field

[0001] The present invention relates to the field of image processing, and in particular to a hole positioning method for automobile parts processing. Background Art

[0002] The automotive parts addressed by this invention, such as the battery tray for new energy vehicles, are key structural components that support the power battery pack. These parts are typically manufactured from lightweight, high-strength metals such as aluminum alloys and are characterized by large size, thin walls, and relatively complex structures. During their production, numerous holes must be precisely machined in specific planar areas to secure battery modules, connect to the vehicle body, or install other accessories. The accuracy of these holes directly impacts the vehicle's assembly quality, safety performance, and range.

[0003] While traditional hole-drilling methods based on specialized fixtures can guarantee accuracy to a certain extent, they are limited for large parts like battery trays, which are prone to uneven deformation due to processes like welding and stamping. Therefore, the use of non-contact, highly adaptable vision-guided technology for hole-drilling positioning has become an important development direction.

[0004] Among vision-based localization methods, the ORB algorithm is commonly used for feature point extraction and description due to its excellent balance between speed and robustness. This algorithm first identifies keypoints using the FAST detector and then generates a binary string for each keypoint using the rBRIEF descriptor. However, when the ORB algorithm is applied to automotive parts with metal surfaces, such as battery trays, the performance of its rBRIEF descriptor is severely challenged. Metal surfaces are prone to specular reflections, and lighting conditions in industrial sites are often not perfectly uniform. These factors cause the grayscale values ​​of point pairs, which rBRIEF relies on for comparison, to become highly unstable within the image blocks surrounding keypoints. Even subtle changes in lighting or viewing angle can cause the highlight state of the same physical feature point to change dramatically, leading to significant differences in its rBRIEF descriptor. This descriptor instability can directly reduce the success rate of subsequent feature matching and increase false matches. Consequently, deformation compensation models built based on these unreliable matching results are inaccurate and ultimately unable to meet the requirements for high-precision hole positioning, limiting the potential application of vision technology in precision machining. Summary of the Invention

[0005] In order to solve the problem that the instability of the above-mentioned descriptors may lead to subsequent feature matching errors, the present invention proposes a hole positioning method for automobile parts processing, comprising: obtaining an image of an automobile part to be processed and calibrating a ROI area; detecting key points and calculating their main directions in the ROI area using the ORB algorithm; determining a neighborhood image block after rotation alignment with each key point as the center; counting the number of pixels in each neighborhood image block whose grayscale value is greater than a set upper threshold and less than a set lower threshold, and the ratio of the number of pixels to the total number of pixels in the neighborhood image block constitutes a local illumination disturbance degree; and determining an adaptive effective intensity normalization method for the neighborhood image block based on the local illumination disturbance degree. The method comprises the following steps: applying the adaptive effective normalization lower limit and upper limit to the neighborhood image block to perform intensity remapping to obtain a remapped image block; generating a feature descriptor of local intensity normalization rBRIEF on the remapped image block; matching the feature descriptor with a feature descriptor in a preset reference model to obtain a high-confidence matching feature point pair; using the high-confidence matching feature point pair to establish a deformation field model for characterizing the local non-uniform two-dimensional deformation of the ROI area relative to the preset reference model; accurately compensating the theoretical hole opening position according to the deformation field model to obtain the actual hole opening position, and guiding the hole opening actuator to complete the hole opening operation.

[0006] Compared with the existing technology that uses the standard rBRIEF descriptor, the present invention generates an ICE-rBRIEF feature descriptor that is more robust to metal surface highlights and uneven lighting by performing adaptive intensity normalization preprocessing on the key point neighborhood image blocks based on the local illumination perturbation characteristics. This significantly improves the stability and accuracy of feature matching under harsh lighting conditions, thereby improving the accuracy of the subsequent deformation compensation model and the final hole positioning. It is particularly suitable for automotive metal parts with complex surface reflective characteristics.

[0007] Furthermore, the calculation method of the local illumination disturbance degree is specifically as follows:

[0008] ;

[0009] in Indicates key points The local illumination disturbance of the corresponding neighborhood image block; Indicates the number of pixels in the image block whose grayscale value is less than the set lower threshold; Indicates the number of pixels in the image block whose grayscale value is greater than the set upper threshold; Indicates the total number of pixels in the neighborhood image block.

[0010] Furthermore, the calculation method of the adaptive effective normalization lower limit and upper limit is specifically as follows:

[0011] ;

[0012] ;

[0013] in represents the effective normalized lower bound of the adaptation; represents the effective normalized upper bound of the adaptation; 、 、 Image blocks The 10th, 50th, and 90th percentile grayscale values; Represents key points The local illumination perturbation of the corresponding neighborhood image block.

[0014] Compared with the normalization method using a fixed range or simple maximum and minimum values ​​in the prior art, the present invention dynamically adjusts the effective range of normalization through the local illumination disturbance degree, so that it shrinks towards the median when the illumination disturbance is severe. This can more effectively suppress the influence of extreme pixel values ​​on the subsequent rBRIEF descriptor generation, and enhance the robustness of the descriptor to illumination changes.

[0015] Furthermore, the calculation method of the remapped image block is specifically as follows:

[0016] ;

[0017] in Represents the image block after remapping Center coordinates The grayscale value of the pixel at ; Represents the original rotation-aligned image patch Center coordinates The grayscale value of the pixel at ; represents the effective normalized lower bound of the adaptation; represents the effective normalized upper bound of the adaptation; represents a very small positive number; Function represents a truncation function used to convert Restricted to within the interval; Represents the rounding function.

[0018] Compared with the existing technology that directly generates descriptors on the original image blocks, the present invention effectively compresses the extreme grayscale values ​​affected by highlights or shadows by linearly mapping the pixel grayscale to the dynamic range determined by the upper and lower limits of adaptive effective normalization, so that the remapped image blocks have more balanced contrast, providing a basis for generating a more stable local intensity normalized rBRIEF feature descriptor.

[0019] Furthermore, the neighborhood image block is an image area of ​​a set size determined with the key point as the center.

[0020] Furthermore, it also includes preprocessing the ROI area, specifically: grayscale processing the ROI area; applying adaptive histogram equalization processing to the grayscale ROI area; applying a Gaussian smoothing filter or a median filter to the ROI area after the adaptive histogram equalization processing for denoising.

[0021] Compared with existing technologies that use simple preprocessing such as standard Gaussian smoothing or median filtering, the present invention can more effectively handle the complex lighting conditions and image noise in industrial sites through targeted background illumination correction and edge-preserving denoising algorithms, providing higher-quality image input for the stability and accuracy of subsequent ORB key point detection.

[0022] Furthermore, after the ORB algorithm detects the key points, it also includes sorting and screening according to the Harris corner response values ​​of the key points, and calculating the main direction of the key points based on the intensity centroid method.

[0023] Furthermore, the feature descriptor is matched with the feature descriptor in the preset reference model to obtain high-confidence matching feature point pairs, which also includes: using a nearest neighbor search based on Hamming distance, and combining the nearest neighbor distance ratio test and cross-validation strategy to perform preliminary matching point pair screening.

[0024] Furthermore, the deformation field model is a two-dimensional thin plate spline interpolation model.

[0025] Furthermore, guiding the hole-opening actuator to complete the hole-opening operation also includes: converting the actual hole-opening position into three-dimensional physical coordinates in the working coordinate system of the hole-opening actuator through pre-calibrated internal and external parameters of the camera and the hand-eye calibration relationship between the camera and the hole-opening actuator, and controlling the hole-opening actuator to operate according to the three-dimensional physical coordinates.

[0026] Compared with existing technologies that only output image coordinates or rely on manual conversion, the present invention clarifies the complete conversion path from image coordinates to the physical coordinates of the actuator. Through precise camera calibration and hand-eye calibration, it ensures that the visual positioning results can be accurately transmitted and act on the opening operation in the physical space. It is a key link in achieving high-precision automated opening.

[0027] The technical effects of the present invention are:

[0028] This paper proposes a method for generating an illumination-robust and adaptive local intensity normalized rBRIEF feature descriptor and applies it to hole location in large automotive metal parts. Specifically, the method first calculates the local illumination perturbation by analyzing the grayscale distribution of the rotationally aligned neighborhood image patch of each ORB keypoint. This metric effectively quantifies the degree to which highlights or deep shadows, common on metal surfaces, affect this local image patch. The calculated local illumination perturbation is then used to dynamically adjust the effective upper and lower bounds used for subsequent intensity normalization. This adjustment enables the normalization process to adaptively narrow its reference dynamic range based on the severity of local illumination, thereby suppressing the influence of extreme pixel values. Furthermore, the adaptively determined effective normalization range is used to perform intensity remapping on the original neighborhood image patch, generating a new image patch that mitigates the effects of uneven illumination, such as highlights and shadows. The final rBRIEF binary descriptor is generated by performing point-by-point grayscale comparisons on this specifically optimized, more illumination-robust image patch. This makes the descriptor more robust to sudden changes in lighting caused by the reflective properties of metal surfaces, significantly improves the stability and accuracy of feature matching, and lays a solid foundation for subsequent high-precision deformation compensation and hole positioning. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood by reading the following detailed description with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present invention are shown in an illustrative and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0030] Figure 1 is a flow chart schematically illustrating a method for hole positioning for automobile parts processing according to an embodiment of the present invention;

[0031] Figure 2 The figure schematically shows a grayscale diagram of a battery tray for a new energy vehicle according to an embodiment of the present invention. DETAILED DESCRIPTION

[0032] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work shall fall within the scope of protection of the present invention.

[0033] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0034] An embodiment of a hole positioning method for automobile parts processing:

[0035] like Figure 1 As shown, a hole positioning method for automobile parts processing of the present invention includes:

[0036] S1. Target area image acquisition and ROI area calibration.

[0037] Prior art often uses specialized fixtures and CNC machine tools to locate holes in automotive parts. However, for large, thin-walled parts like the lower housing of new energy vehicle battery trays, which undergo welding or complex extrusion molding, traditional rigid positioning methods struggle to maintain high precision at all critical holes due to unpredictable and uncontrollable welding deformation and molding tolerances. Furthermore, these methods are costly and lack flexibility.

[0038] In this embodiment, the battery tray of a new energy vehicle is taken as an example. Figure 2 As shown, there are multiple flat areas for installing battery modules or connecting to the vehicle body, and the hole position accuracy of these areas is required to be high. A battery tray lower shell that has just been welded or stamped and still needs to be drilled is fixed to the visual inspection station of the automated processing unit through a universal fixture. The visual system should include at least one industrial digital camera, which can be installed via a C-arm or fixed bracket so that its lens optical axis is roughly perpendicular to a specific flat functional area on the workpiece where the hole is expected to be opened. Start the camera to capture a two-dimensional digital image of the flat area.

[0039] In order to focus computing resources and accurately limit the scope of subsequent analysis, the operator can use the human-computer interaction interface to interactively (for example, drag the mouse to select) mark the precise region of interest (ROI) on the real-time displayed workpiece image, referring to the outline of the pre-loaded CAD two-dimensional engineering drawing of the corresponding area of ​​the workpiece. Alternatively, the system can automatically identify and frame the ROI through preliminary rough positioning and template matching. Exemplary explanation: After the camera captures a battery tray, the operator selects a 200mm×300mm mounting plane in the middle as the ROI. If the image resolution is 0.1mm / pixel, one possible case of the ROI parameter is: ,in Indicates the center coordinates of the ROI area; Indicates the width and height of the ROI area.

[0040] S2. ROI image preprocessing and ORB feature point detection.

[0041] The acquired ROI raw image data, especially those from industrial sites, is often accompanied by interference such as uneven lighting and random noise. To improve the quality of subsequent feature point detection, image preprocessing is required and the FAST component of the ORB algorithm is used to initially detect key points.

[0042] First, the input color image is converted to a single-channel grayscale image. Next, to mitigate the effects of uneven illumination, adaptive histogram equalization can be applied to the grayscale image. Finally, to smooth random noise and preserve edge details, a Gaussian smoothing filter or median filter with a 5x5 kernel, for example, can be applied. The aforementioned preprocessing algorithms are well-known and will not be detailed here.

[0043] On the pre-processed ROI grayscale image, the standard key point detection process of the ORB algorithm is applied. The process first uses the FAST (Features from Accelerated Segment Test) corner detection algorithm to quickly identify candidate corner points. Subsequently, to ensure the quality and stability of the corner points, the Harris corner response values ​​of these candidate corner points are usually calculated and sorted and filtered according to the response values. In one embodiment, the top candidate corner points with the highest response values ​​can be selected. In another embodiment, a response value threshold can be set to select only corner points with a response value greater than the threshold. The ORB algorithm can also calculate the main direction of each selected key point based on the intensity centroid method, and the main direction is used for the rotation invariance of the subsequent feature descriptor.

[0044] S3. Extract key point neighborhood image blocks and rotate and align them; analyze local illumination characteristics of neighborhood image blocks and calculate normalization parameters; remap neighborhood image block intensities based on adaptive effective normalization range.

[0045] After S2 obtains the ORB keypoints and their principal directions, it is necessary to generate a feature descriptor for each keypoint that is robust to lighting variations, particularly specular reflections and shadows on metal surfaces. The standard rBRIEF descriptor performs point-to-point grayscale comparisons on image patches that have been rotated and aligned around the keypoints. This approach becomes extremely unstable when faced with dramatic lighting changes common on metal surfaces, severely impacting subsequent matching accuracy.

[0046] In this embodiment, an illumination-robust and adaptive local intensity normalization rBRIEF (ICE-rBRIEF) descriptor is proposed. The core idea is to perform adaptive intensity normalization preprocessing based on local illumination characteristics analysis on the image blocks in the neighborhood of each key point before applying the binary string generation rule of the rBRIEF standard.

[0047] S3.1: Extract keypoint neighborhood image patches and rotate and align them.

[0048] First, for each ORB key point output by S2, a fixed-size neighborhood image block is extracted from the preprocessed ROI grayscale image with its coordinates as the center. In this embodiment, the size of the image block can be set to 31x31 pixels; then, the main direction information of the key point is used to rotate the image block so that its main direction is aligned to a unified reference direction (for example, the positive direction of the x-axis). This step is a standard component of the ORB algorithm and is used to ensure that the subsequently generated descriptors are rotationally invariant. This step extracts and rotates the neighborhood image block for each key point. .

[0049] S3.2: Analysis of local illumination characteristics of neighborhood image blocks and calculation of normalization parameters.

[0050] The highlights or deep shadows on the metal surface will cause the grayscale distribution of pixels in the neighborhood image block to be extreme, making the rBRIEF descriptor based on the fixed comparison method unstable. Therefore, it is necessary to quantify the degree of this illumination disturbance. In this embodiment, for the rotated and aligned neighborhood image block output in step S3.1, (Assume it contains pixels), calculate their local illumination disturbance .

[0051] First calculate the image block The 10th percentile of all pixel grayscale values and the 90th percentile In this embodiment, As the lower threshold, As the upper threshold. These two values ​​can relatively robustly represent the main grayscale range of the image block excluding extreme highlights and dark spots. In addition to the 10th percentile and 90th percentile, the implementer can also set it to the 15th percentile and 85th percentile according to actual conditions or further experiments; then count the grayscale values ​​in the image block that are strictly less than Number of pixels , and the grayscale value is strictly greater than Number of pixels , these pixels can be regarded as pixels that are more seriously affected by the light disturbance; finally, the proportion of these extreme pixels is calculated to obtain the local light disturbance degree ,have:

[0052] ;

[0053] in Indicates key points The local illumination disturbance of the corresponding neighborhood image block; Indicates the number of pixels in the image block whose grayscale value is less than the lower threshold; Indicates the number of pixels in the image block whose grayscale value is greater than the upper threshold; Represents the total number of pixels in the image block, which is .

[0054] When the grayscale distribution within the image block is relatively concentrated and there are fewer extreme bright and dark pixels (for example, a diffuse reflection surface with uniform illumination), and If an image patch has a large number of very dark or very bright pixels (for example, there are deep shadows and strong highlights at the same time), then The value of will be high, indicating that the image block is seriously disturbed by the lighting, that is, in the strong highlight area or deep shadow area on the surface of the battery tray, will be significantly higher; in areas with more uniform lighting, It will be low.

[0055] According to the local illumination disturbance of the image block , it is necessary to determine an effective dynamic range for subsequent intensity normalization. If it is very high, it means that the original dynamic range of the image block may have been destroyed by extreme values, and the effect of normalization directly using the original minimum and maximum values ​​or fixed percentile points is not good. To dynamically adjust the upper and lower limits used for normalization, so that they shrink toward the central grayscale trend of the image block, thereby suppressing the influence of extreme values.

[0056] set up 、 (median), Image blocks The 10th, 50th, and 90th percentile grayscale values ​​are:

[0057] ;

[0058] ;

[0059] in represents the effective normalized lower bound of the adaptation; represents the effective normalized upper bound of the adaptation; Represents key points The local illumination disturbance of the corresponding neighborhood image block. The above formula is obtained by The original robust grayscale range Towards the median Perform linear interpolation shrinkage.

[0060] when , that is, when the light disturbance is extremely small, and In this case, the normalization will use the original 10-90 percentile range; when , that is, when there is light disturbance, will be greater than , will be smaller than , both to Closer, the normalized effective dynamic range becomes narrower. The larger it is, the greater the shrinkage.

[0061] if reaches its theoretical maximum (e.g., assuming all pixels are below 10% or above 90%, representing extreme cases), and Will tend to , which will result in extremely low contrast in the normalized image blocks (close to a constant image), making subsequent rBRIEF point pair comparisons almost random. In extreme cases, this may actually increase the neglect of "structure" and rely more on a small number of pixels that are not fully compressed. However, the more practical effect is that by shrinking the dynamic range, pixels that were originally saturated due to highlights and pixels close to 0 due to shadows can be pulled into a more concentrated grayscale range after normalization, avoiding them from generating fixed 0 or 1 in rBRIEF binary comparisons, thereby increasing the descriptor's robustness to lighting changes.

[0062] In the highlight area of ​​the battery tray, High, leading to The range is significantly narrowed, suppressing highlights; in areas with uniform lighting, Low, the range is close to , maintaining the original contrast.

[0063] S3.3: Neighborhood image patch intensity remapping based on adaptive effective normalization range.

[0064] The effective normalized lower limit determined using S3.2 and upper limit , the original neighborhood image block after rotation alignment obtained in S3.1 Perform intensity remapping to generate a new, more lighting-robust image patch The subsequent rBRIEF descriptor will be calculated on this remapped image block.

[0065] For the original image block For each pixel in, record its gray value , remap it to a new grayscale value , and form a new image block .

[0066] ;

[0067] in Represents the image block after remapping Center coordinates The grayscale value of the pixel at ; Represents the original rotation-aligned image patch Center coordinates The grayscale value of the pixel at ; represents the effective normalized lower bound of the adaptation; represents the effective normalized upper bound of the adaptation; represents a very small positive constant, which can be set to 1e-6 in this embodiment to prevent division by zero errors; Function represents a truncation function used to convert Restricted to within the interval; Represents the rounding function.

[0068] The above steps convert the original pixel grayscale value Linearly adjusts from the input range dynamically Mapping to standard output ranges , if the original gray value of a pixel Lower than , then its remapped value will be truncated to 0; if the original grayscale value of a pixel Higher than , then its remapped value will be truncated to 255, and pixels in between will be linearly stretched to scope.

[0069] By dynamically adjusting and (based on ), the remapping process can effectively adjust the grayscale distribution of image blocks severely affected by highlights or deep shadows to a more standardized and contrast-controlled range, so that the subsequent rBRIEF point pair grayscale is less susceptible to the original extreme lighting conditions.

[0070] For the image blocks on the battery tray that are illuminated by the bright light, Higher, resulting in and The range between them is narrowed and may move upward as a whole. The normalization of the above formula will effectively suppress the brightness of the highlight area and enhance the contrast of the originally darker area. For areas with uniform lighting, Lower, the normalized range is closer to the original 10-90 percentile range, and the transformation has less effect.

[0071] The standard rBRIEF descriptor selects predefined (e.g. 256) point pairs on the Gaussian-smoothed, rotationally aligned image blocks. , compare its grayscale size: if ， The descriptor's The ICE-rBRIEF descriptor of the present invention is an image block with more robust illumination obtained after S3.3 intensity remapping. On the grayscale value of these 256 points and to generate the final 256-bit binary descriptor.

[0072] S4. Matching of ICE-rBRIEF feature descriptors.

[0073] After generating the more illumination-robust ICE-rBRIEF descriptor, the next step is to match it with the descriptor in the pre-stored reference model to establish the feature point correspondence between the current workpiece ROI and the reference model.

[0074] First, the ROI of the ideal workpiece to be located needs to be processed. The S2 and S3 methods (i.e., ORB keypoint detection and ICE-rBRIEF descriptor generation) are used to extract reference keypoints and their ICE-rBRIEF descriptors to build a reference descriptor library. Then, a fast matching algorithm based on Hamming distance is used. For each ICE-rBRIEF descriptor in the current ROI, the descriptor with the smallest Hamming distance to it is searched in the reference descriptor library as the best match. At the same time, to improve the reliability of the matching, cross-validation and / or the nearest neighbor distance ratio test (Lowe's Ratio Test, i.e., the Hamming distance of the best match is significantly smaller than the Hamming distance of the second best match, for example, less than 0.75 times) can be used.

[0075] This step can finally output a set of geometrically verified high confidence matching feature point pairs. Each pair contains the coordinates of a key point in the current ROI image. and the corresponding key point coordinates in the reference model (in this embodiment, the CAD coordinate system) .

[0076] For example, if you input 100 ICE-rBRIEF descriptors for the current ROI and 500 ICE-rBRIEF descriptors for the reference model, you might get 60 highly confident matching pairs after matching and filtering, such as [current point (100, 150) -> reference point (50.5, 75.2)], [current point (320, 210) -> reference point (160.1, 105.8)], etc.

[0077] S5. Modeling of the local non-uniform two-dimensional deformation field within the ROI.

[0078] First, a mathematical model that can fit non-rigid deformation is selected, such as the 2D Thin Plate Spline (TPS) model, which can find a smooth mapping function that minimizes the overall bending energy. , and thus can accurately pass through all given matching control point pairs; then by solving a set of linear equations related to the number of control points, the coefficients of the TPS model can be determined. After the model is established, for any point in the ROI in the reference coordinate system, its corresponding position in the current actual deformed workpiece image can be calculated through the TPS model.

[0079] This step can output a parameterized two-dimensional non-rigid deformation field model, which in this embodiment is the coefficient matrix of the TPS model. Exemplary explanation: Based on the 60 pairs of matching points output by S4, a TPS model is calculated, which can be expressed as a function:

[0080] ;

[0081] in Indicates the coordinates of key points in the reference model CAD coordinate system; Represents a smooth mapping function.

[0082] S6. Accurate compensation and guidance of the target opening position.

[0083] The precise local deformation field model established based on S5 can accurately map the theoretical hole position on the CAD design drawing to the current actual deformed workpiece ROI, thereby guiding the hole-opening actuator to perform high-precision operations.

[0084] First, the theoretical two-dimensional coordinate list of all target holes to be processed in the ROI in the workpiece CAD design coordinate system is recorded as ; Then for each theoretical hole position , using the deformation field model output by S5 (such as ) to perform forward mapping and calculate its precise compensated position in the actual ROI image coordinate system of the current workpiece ,Right now: ; Further calculate the actual opening position in the image coordinate system , using pre-calibrated camera internal and external parameters and hand-eye calibration (if the drilling tool is gripped by a robot), these parameters are converted into 3D physical coordinates within the drilling actuator (such as the industrial robot end-of-tool coordinate system or the CNC machine tool table coordinate system). For drilling perpendicular to the ROI plane, the Z-axis coordinate is typically determined by the workpiece surface height and tool length. Ultimately, the control system drives the drilling actuator (such as the drilling spindle, laser head, or punching module) to precisely move to the calculated 3D physical coordinates and execute the drilling operation.

[0085] Example: If there is a hole position (100.0, 200.0) mm in CAD. After the TPS model of S5 After the transformation, its position in the current image is obtained and then converted into the actual processing target point in the robot coordinate system, for example (X=350.5, Y=520.1, Z=-150.0) mm (Z is the processing depth), then the robot moves to this position to complete the drilling.

Claims

1. A hole positioning method for automobile parts processing, characterized in that: The method comprises: acquiring an image of an automobile part to be processed and calibrating a ROI region; detecting key points in the ROI region using an ORB algorithm and calculating the main direction thereof; determining a neighborhood image block after rotational alignment with each key point as the center; and counting the number of pixels in each neighborhood image block whose grayscale value is greater than a set upper threshold and less than a set lower threshold, wherein the ratio of the number of pixels to the total number of pixels in the neighborhood image block constitutes a local illumination disturbance degree. Based on the local illumination disturbance degree, determining an adaptive effective normalization lower limit and an upper limit for intensity normalization of the neighborhood image block; applying the adaptive effective normalization lower limit and upper limit to the neighborhood image block to perform intensity remapping to obtain a remapped image block; generating a local intensity normalized rBRIEF feature descriptor on the remapped image block; the local illumination disturbance degree is calculated specifically as follows: ； in Indicates key points The local illumination disturbance of the corresponding neighborhood image block; Indicates the number of pixels in the image block whose grayscale value is less than the set lower threshold; Indicates the number of pixels in the image block whose grayscale value is greater than the set upper threshold; Represents the total number of pixels in the neighborhood image block; The calculation method of the adaptive effective normalization lower limit and upper limit is specifically as follows: ； ； in represents the effective normalized lower bound of the adaptation; represents the effective normalized upper bound of the adaptation; 、 、 Image blocks The 10th, 50th, and 90th percentile grayscale values; Indicates key points The local illumination disturbance of the corresponding neighborhood image block; The feature descriptor is matched with the feature descriptor in the preset reference model to obtain a high-confidence matching feature point pair; the high-confidence matching feature point pair is used to establish a deformation field model for characterizing the local non-uniform two-dimensional deformation of the ROI area relative to the preset reference model; the theoretical hole opening position is accurately compensated according to the deformation field model to obtain the actual hole opening position, and the hole opening actuator is guided to complete the hole opening operation.

2. A hole positioning method for automobile parts processing according to claim 1, characterized in that: The calculation method of the remapped image block is specifically as follows: ； in Represents the image block after remapping Center coordinates The grayscale value of the pixel at ; Represents the original rotation-aligned image patch Center coordinates The grayscale value of the pixel at ; represents the effective normalized lower bound of the adaptation; represents the effective normalized upper bound of the adaptation; represents a very small positive number; Function represents a truncation function used to convert Restricted to within the interval; Represents the rounding function.

3. The method for hole positioning for automobile parts processing according to claim 1, characterized in that: The neighborhood image block is an image area of ​​a set size determined with the key point as the center.

4. A hole positioning method for automobile parts processing according to claim 1, characterized in that: The method also includes pre-processing the ROI area, specifically: Grayscale the ROI area; Apply adaptive histogram equalization to the grayscale ROI area; Apply a Gaussian smoothing filter or a median filter to the ROI area after adaptive histogram equalization for denoising.

5. The method for hole positioning for automobile parts processing according to claim 1, characterized in that: After the ORB algorithm detects the key points, it also includes sorting and screening according to the Harris corner response values ​​of the key points, and calculating the main direction of the key points based on the intensity centroid method.

6. A hole positioning method for automobile parts processing according to claim 1, characterized in that: Matching the feature descriptor with a feature descriptor in a preset reference model to obtain a high-confidence matching feature point pair, including: The nearest neighbor search based on Hamming distance is used, combined with the nearest neighbor distance ratio test and cross-validation strategy to perform preliminary matching point pair screening.

7. A hole positioning method for automobile parts processing according to claim 1, characterized in that: The deformation field model is a two-dimensional thin plate spline interpolation model.

8. The method for hole positioning for automobile parts processing according to claim 1, characterized in that: Guide the drilling actuator to complete the drilling operation, including: The actual hole opening position is converted into a three-dimensional physical coordinate in the working coordinate system of the hole opening actuator through the pre-calibrated internal and external parameters of the camera and the hand-eye calibration relationship between the camera and the hole opening actuator, and the hole opening actuator is controlled to operate according to the three-dimensional physical coordinate.

Images