A deburring device for a tooling fixture and an automatic burr positioning method

Abstract

The present invention relates to the technical field of deburring, and in particular to a deburring device for a tooling fixture and a method for automatically locating burrs. The method comprises: obtaining a plurality of monitoring images of the tooling fixture to be deburred at each viewing angle; performing a Gaussian pyramid scale transformation on each monitoring image, and performing corner point detection, and dividing the corner points in each monitoring image into two categories according to the presentation of the corner points in the scale map; combining the relative distance and edge change between each corner point and other corner points in the corresponding scale map of each monitoring image to obtain the texture representation value of each corner point in the scale map; combining the texture representation value of each corner point in the corresponding scale map of all monitoring images at the same viewing angle to obtain the texture feature value of each corner point in the monitoring image, and then dividing and screening the burr area for each category. The present invention improves the accuracy of locating the burr area and the burr removal effect.

Description

Technical Field

[0001] The present invention relates to the technical field of deburring, and in particular to a deburring device for a tooling fixture and an automatic burr positioning method. Background Art

[0002] Fixtures and jigs are devices used during the manufacturing process to secure, support, position, or restrict the position and state of a product during processing. They play a vital role in ensuring product quality, improving production efficiency, and ensuring operational safety. During the production process, jigs and jigs utilize various machining equipment, such as lathes, milling machines, grinders, and EDMs, to perform cutting, drilling, and grinding operations on materials to create the various components of the jigs and jigs. This process can generate burrs, which can affect the product's processing accuracy and quality during use.

[0003] Binocular vision generally uses two cameras to simultaneously capture two digital images of an object from different angles, extract the pixel coordinates of the points in the two images, find the corresponding points in the two images, calculate the offset between the corresponding points, and restore the depth information of the object based on the principle of triangulation. When deburring tooling, it is necessary to locate the burr area in the acquired part. During positioning, the binocular camera can be used to capture the part information from different angles, and the spatial depth of the product can be determined by the matching of the binocular cameras. However, the image changes caused by the burrs during binocular camera positioning may lead to incorrect feature point matching, which will directly affect the calculation of the disparity map and the accuracy of the depth information. At the same time, the three-dimensional burrs may be obscured by the structure of the product itself or other burrs, which will prevent the burrs from being fully captured by the binocular camera, resulting in inaccurate positioning, which in turn affects the subsequent burr removal effect. Summary of the Invention

[0004] In order to solve the problem that the existing method of deburring the tooling fixture is not accurate in locating the burr area, which affects the burr removal effect, the purpose of the present invention is to provide a tooling fixture deburring device and a burr automatic positioning method. The technical solutions adopted are as follows:

[0005] In a first aspect, the present invention provides a method for automatically locating burrs, the method comprising the following steps:

[0006] Acquire several monitoring images of the tooling to be deburred at each viewing angle;

[0007] Each surveillance image is scaled using a Gaussian pyramid and corner detected. The corners in each surveillance image are divided into two categories based on how they appear in the scale map. The texture representation value of each corner in the scale map is obtained by combining the relative distances and edge changes between each corner in the corresponding scale map.

[0008] Combining the texture representation value of each corner point in the corresponding scale map of all monitoring images under the same viewing angle, the texture feature value of each corner point in the monitoring image is obtained; and the texture feature value is used to filter the burr area for each category.

[0009] Preferably, the method of dividing the corner points in each surveillance image into two categories according to the presentation of the corner points in each surveillance image in the scalogram includes:

[0010] For any corner point in the candidate monitoring image, the total number of times the corner point is judged as a corner point in all scale images corresponding to the candidate monitoring image is recorded as the scale feature value of the corner point;

[0011] Based on the scale feature values of all corner points in the candidate monitoring image, the maximum inter-class variance method is used to classify all corner points in the candidate monitoring image into two categories; the candidate monitoring image is any monitoring image.

[0012] Preferably, the step of combining the relative distance and edge change between each corner point and other corner points in the corresponding scalogram of each surveillance image to obtain the texture representation value of each corner point in the scalogram includes:

[0013] The corner point directly connected to the candidate corner point through the edge line is recorded as the reference corner point of the candidate corner point; the discrete degree of the slope of all pixels on the edge line between the candidate corner point and each of its reference corner points is calculated and recorded as the first eigenvalue between the candidate corner point and each of its reference corner points;

[0014] Obtaining a texture representation value of the candidate corner point based on the Euclidean distance between the candidate corner point and all reference corner points and the first eigenvalue, wherein the Euclidean distance is negatively correlated with the texture representation value, and the first eigenvalue is positively correlated with the texture representation value;

[0015] The candidate corner point is any corner point in any scale map.

[0016] Preferably, the calculation of the discrete degree of the slopes of all pixel points on the edge line between the candidate corner point and each of its reference corner points includes: taking the variance of the slopes of all pixel points on the edge line connecting the candidate corner point and each of its reference corner points as the corresponding discrete degree.

[0017] Preferably, combining the texture representation value of each corner point in the corresponding scale map in all surveillance images at the same viewing angle to obtain the texture feature value of each corner point in the surveillance image includes:

[0018] The variance of the texture representation values of the corresponding positions of the corner points to be analyzed in all scale images is recorded as the correction coefficient;

[0019] Using the correction coefficient, the texture representation values of the positions of the corner points to be analyzed in all monitoring images under the same viewing angle in all scale images are corrected to obtain the texture feature values of the positions of the corner points to be analyzed in the monitoring images;

[0020] The corner point to be analyzed is any corner point in any monitoring image.

[0021] Preferably, the method of using the correction coefficient to correct the texture representation values of the positions of the corner points to be analyzed in all the surveillance images at the same viewing angle in all the scale images to obtain the texture feature values of the positions of the corner points to be analyzed in the surveillance images includes:

[0022] The texture feature value of the corner point to be analyzed in all surveillance images under the same viewing angle is determined as the cumulative sum of the products of the texture representation values of the positions corresponding to the corner point to be analyzed in all surveillance images in all scale images and the correction coefficient.

[0023] Preferably, the method of using the texture feature value to divide and filter the burr area into each category includes:

[0024] For any surveillance image:

[0025] Calculate the average of the scale characteristic values of all corner points in each category and record it as the average scale characteristic value of each category; record the category with the largest average scale characteristic value as the first category, and the category with the smallest average scale characteristic value as the second category;

[0026] Based on the texture feature values of all corner points in each category in each surveillance image, the maximum inter-class variance method is used to divide all corner points in each category in each surveillance image into two subcategories again;

[0027] Calculate the average value of the texture feature values of all corner points in each subclass and record it as the average texture feature value of each subclass; record the subclass with the largest average texture feature value in the first class as the first subclass, and record the subclass with the largest average texture feature value in the second class as the second subclass;

[0028] A burr area is determined according to the first subcategory and the second subcategory.

[0029] Preferably, determining the burr area according to the first subcategory and the second subcategory includes:

[0030] All corner points in the first subclass and edge lines between the corner points form a first burr area;

[0031] All corner points in the second subclass and the edge lines between the corner points form a second burr area;

[0032] The burr region includes a first burr region and a second burr region.

[0033] In a second aspect, the present invention provides a deburring device for a tooling fixture, the device comprising:

[0034] An image acquisition module is used to obtain several monitoring images of the tooling to be deburred at each viewing angle;

[0035] The texture representation value determination module is used to perform Gaussian pyramid scale transformation on each surveillance image and perform corner detection. The corner points in each surveillance image are divided into two categories based on their presentation in the scale map. The texture representation value of each corner point in the scale map is obtained by combining the relative distance and edge changes between each corner point and other corner points in the corresponding scale map of each surveillance image.

[0036] The burr area screening module is used to combine the texture representation value of each corner point in the corresponding scale map of all monitoring images under the same viewing angle to obtain the texture feature value of each corner point in the monitoring image; and use the texture feature value to screen the burr area for each category;

[0037] The deburring module is used to screen the obscured structural feature points on the tooling to be deburred based on the relative distance between each position of the tooling to be deburred and the burr area and the fact that each position is judged as a corner point in the monitoring image under the same viewing angle; reconstruct the three-dimensional model of the tooling to be deburred by combining the obscured structural feature points and all corner points in all monitoring images under different viewing angles; and plan the deburring route of the deburring equipment and perform deburring processing based on the three-dimensional model and the position of the burr area.

[0038] Preferably, the screening of blocked structural feature points on the tooling to be deburred according to the relative distance between each position of the tooling to be deburred and the burr area and the fact that each position is determined to be a corner point in the monitoring image at the same viewing angle includes:

[0039] For any position of the tooling to be deburred:

[0040] Calculate the minimum value of the distance between the position and all corner points in the first subclass; count the number of times the position is determined to be a corner point in each surveillance image under the same viewing angle; and obtain the probability that the position is an occluded structure point in each surveillance image based on the minimum value and the number of times the position is determined to be a corner point in each surveillance image under the same viewing angle;

[0041] If the probability is greater than the probability threshold, it is determined that the position in the corresponding monitoring image belongs to an obscured structure point.

[0042] The present invention has at least the following beneficial effects:

[0043] The present invention first performs a Gaussian pyramid scale transformation on each monitoring image of the tooling fixture to be deburred to obtain a corresponding scale map, performs corner point detection on each monitoring image and scale map, and then divides the corner points in the monitoring image into two categories based on the presentation of the corner points in the scale map. One category is stable points with obvious scale features, that is, structural corner points or large burr corner points of the tooling fixture itself to be deburred, and the other category is points with unstable scale features, that is, structural corner points or small burr corner points of the tooling fixture itself; based on the relative distance and edge change between each corner point and other corner points in the scale map, the texture distribution of the corner points in the scale map is evaluated to obtain a texture expression value; based on the texture expression value of each corner point in the corresponding scale map of all monitoring images under the same viewing angle, the texture feature value of each corner point in the monitoring image is obtained, and the texture feature value is used to screen the corner points in each category again to determine the burr area, thereby realizing subsequent deburring processing, making the positioning of the burr area more accurate, and improving the burr removal effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0045] Figure 1 A flow chart of a method for automatically locating burrs provided by an embodiment of the present invention;

[0046] Figure 2 This is a structural block diagram of a deburring device for a tooling fixture provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0047] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the following is a detailed description of a deburring device for a tooling fixture and an automatic burr positioning method proposed in accordance with the present invention in combination with the accompanying drawings and preferred embodiments.

[0048] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0049] The following describes in detail a tooling fixture deburring device and a specific solution for an automatic burr positioning method provided by the present invention with reference to the accompanying drawings.

[0050] An embodiment of a burr automatic positioning method:

[0051] The specific scenario targeted by this embodiment is: in the process of deburring the tooling fixture using deburring equipment, it is necessary to first locate the burr area on the tooling fixture to be deburred. This embodiment uses a binocular camera to collect image information of the tooling fixture to be deburred that needs to be polished, obtains corner points in images taken at different angles through the SIFT algorithm, determines the burr distribution information in space by analyzing the corner points, and then performs the burr polishing operation based on the acquired spatial information.

[0052] This embodiment proposes a method for automatically locating burrs. Figure 1 As shown, a burr automatic location method of this embodiment includes the following steps:

[0053] Step S1: Acquire several monitoring images of the tooling to be deburred at each viewing angle.

[0054] First, in this embodiment, when acquiring image data of the jig to be deburred, binocular cameras are mounted above and below the polished part. During polishing, the jig rotates. In specific applications, the implementer sets the image acquisition frequency based on the time it takes for the jig to complete one rotation. All cameras in the binocular camera have the same acquisition frequency and field of view. That is, at the same moment, all cameras capture images of the same area of the jig to be deburred, and each image captured by a camera represents an image from a single viewing angle. This embodiment uses images of a single area of the jig to be deburred as an example; the method provided in this embodiment can be used to process other areas. In this embodiment, 20 images of the jig to be deburred are captured from each viewing angle. In specific applications, the implementer can set the number of captured images based on the specific situation. It should be noted that during image acquisition, the field of view of each camera is fixed. Each of the collected surveillance images is pre-processed and background information removed, and the processed images are recorded as surveillance images. The image pre-processing and background information removal processes are both existing technologies and will not be described in detail here.

[0055] So far, this embodiment has collected multiple monitoring images of the tooling fixture to be deburred from different viewing angles.

[0056] In step S2, a Gaussian pyramid scale transformation is performed on each surveillance image, and corner point detection is performed. The corner points in each surveillance image are divided into two categories according to their presentation in the scale map. The texture representation value of each corner point in the scale map is obtained by combining the relative distance and edge change between each corner point and other corner points in the corresponding scale map of each surveillance image.

[0057] When binocular cameras are used to locate burrs, the fixture to be deburred has a complex three-dimensional geometric structure, which may cause one part of the product to block another part. The blocked part may not be captured by both cameras at the same time, resulting in incomplete three-dimensional information of this part. Therefore, the product is rotated during the inspection process to capture images from multiple angles to obtain complete three-dimensional information.

[0058] When performing spatial positioning based on captured surveillance images, feature points must be extracted from the images. These feature points are then used to determine correspondences in stereo matching. A stereo matching algorithm is used to find corresponding feature points in different surveillance images. This step is based on the principle of disparity, which measures the positional difference between two images of the same object. A disparity map is calculated using matched feature point pairs, and the depth information of each feature point is then calculated based on the camera's internal and external parameters. Based on this depth information and image features, image processing algorithms (such as edge detection and morphological operations) are used to identify and locate glitches.

[0059] When matching information in an image, it is impossible for all points to participate in the matching. Therefore, this embodiment uses the SIFT (Scale-Invariant Feature Transform) algorithm to extract scale-invariant feature points in the image, namely, corner points.

[0060] Specifically, each surveillance image is first subjected to a Gaussian pyramid scale transform to obtain multiple scale maps corresponding to each surveillance image. The parameters of the Gaussian pyramid scale transform are set by the implementer based on the specific situation. By setting different parameters, the surveillance image is converted into image information of different scales. Gaussian pyramid scale transform is a prior art and will not be described in detail here. Then, SIFT corner detection is performed on each surveillance image and each obtained scale map, respectively, to obtain multiple corner points in each surveillance image and multiple corner points in each scale map. SIFT corner detection is a prior art and will not be described in detail here.

[0061] When acquiring SIFT feature points, the acquired burrs may block or change the original feature points, causing these feature points to be lost in the image. At the same time, the surface of the tooling to be deburred is not a smooth metal structure, but has a certain frosted texture, which will cause a certain deviation in the acquired corner points. Therefore, the acquired corner points cannot be used directly and need further screening.

[0062] The grayscale of the fixture surface to be deburred is inherently uneven. Acquiring corner points on this uneven surface can result in excessive amounts of useless corner point information, impacting subsequent matching accuracy. When matching captured corner points, it's necessary to filter the corner point information and select valid corner points for subsequent matching between images acquired at different locations.

[0063] Next, any surveillance image is taken as an example for description. Other surveillance images can also be processed using the method provided in this embodiment.

[0064] Specifically, any surveillance image is designated as a candidate surveillance image. For any corner point in the candidate surveillance image, the total number of times that corner point is identified as a corner point in all scale maps corresponding to the candidate surveillance image is recorded as the scale eigenvalue of the corner point. Using this method, the scale eigenvalue of each corner point in the candidate surveillance image can be obtained. Based on the scale eigenvalues of all corner points in the candidate surveillance image, the maximum inter-class variance method is used to classify all corner points in the candidate surveillance image into two categories. The maximum inter-class variance method is a state-of-the-art method and will not be described in detail here. The average scale eigenvalues of all corner points in each category are calculated and recorded as the average scale eigenvalue of each category. The category with the largest average scale eigenvalue is designated as the first category, and the category with the smallest average scale eigenvalue is designated as the second category. Corner points in the first category are stable points with obvious scale characteristics, i.e., structural corner points of the tooling to be deburred or large burr corner points. Corner points in the second category are scale-unstable points, i.e., structural corner points of the tooling to be deburred or small burr corner points.

[0065] This embodiment performs a Gaussian pyramid scale transformation on each monitoring image, which can smooth out the uneven grayscale areas in the monitoring image of the tooling fixture to be deburred. However, when the burr is too small, the scale change will cause it to be smoothed out and the feature points in the burr area may not be obtained. When the burr is too large, it will lead to the acquisition of feature points at the edge of the tooling fixture to be deburred itself. Therefore, only the correspondence between the corner points in the scale map can be used to preliminarily screen the acquired corner points, that is, only part of the data can be screened out, and it cannot directly extract the feature points of the burr area and the feature points caused by the structural changes of the part itself.

[0066] When performing subsequent matching, we can first obtain the texture features of the corner points in images of different scales, and then construct the initial matching features of each corner point for matching by combining the texture features and the above-mentioned scale features, that is, the spatial texture change features in the image.

[0067] Burrs can range in size from tiny to large, and their shapes can be irregular, with sharp edges or protruding corners. In an image, burrs appear as edge discontinuities, with protrusions or depressions appearing where an edge should be smooth. Burrs can also alter the texture of a local area; for example, a previously smooth surface may appear rough due to the presence of burrs.

[0068] The Canny edge detection algorithm is used to perform edge detection on each surveillance image and the corresponding scale map of each surveillance image to obtain edge lines in each image. The Canny edge detection algorithm is an existing technology and will not be described in detail here.

[0069] Next, this embodiment is described by taking any corner point in any scalogram as an example. The method provided in this embodiment can be used to process other corner points in the scalogram or all corner points in other scalograms.

[0070] Specifically, any corner point in any scalogram is recorded as a candidate corner point, and the corner points directly connected to the candidate corner point by an edge line are recorded as the reference corner points of the candidate corner point. The degree of dispersion of the slopes of all pixels on the edge line between the candidate corner point and each of its reference corner points is calculated and recorded as the first eigenvalue between the candidate corner point and each of its reference corner points. In this embodiment, the variance of the slopes of all pixels on the edge line connected to the candidate corner point and each of its reference corner points is used as the degree of dispersion of the slopes of all pixels on the edge line between the candidate corner point and each of its reference corner points. The texture representation value of the candidate corner point is obtained based on the Euclidean distance between the candidate corner point and all of its reference corner points and the first eigenvalue. The Euclidean distance is negatively correlated with the texture representation value, while the first eigenvalue is positively correlated with the texture representation value.

[0071] Among them, a positive correlation relationship indicates that the dependent variable will increase as the independent variable increases, and the dependent variable will decrease as the independent variable decreases. It can be an additive relationship, a multiplicative relationship, etc., which is determined by actual application; a negative correlation relationship indicates that the dependent variable will decrease as the independent variable increases, and the dependent variable will increase as the independent variable decreases. It can be a subtractive relationship, a division relationship, etc., which is determined by actual application.

[0072] In this embodiment, a specific calculation formula for the texture representation value is given. The texture representation value of the j-th corner point in any scale map can be expressed as:

[0073]

[0074] Among them, B j represents the texture representation value of the j-th corner point in the scale map, A represents the number of reference corner points of the j-th corner point in the scale map, ε j,a Indicates the discrete degree of the slope of all pixels on the edge line between the jth corner point and its ath reference corner point, l j,a represents the Euclidean distance between the jth corner point and its ath reference corner point.

[0075] The greater the discreteness of the slopes of all pixels on the edge line between the jth corner point and its ath reference corner point, and the smaller the Euclidean distance between the jth corner point and its reference corner point, the greater the texture representation value of the jth corner point in the scale map.

[0076] By using the above method, the texture representation value of each corner point in each scale map can be obtained.

[0077] Step S3: combining the texture representation value of each corner point in the corresponding scale map in all monitoring images under the same viewing angle to obtain the texture feature value of each corner point in the monitoring image; and using the texture feature value to filter the burr area for each category.

[0078] The texture representation value obtained in step S2 of this embodiment only reflects the texture variation characteristics of the corresponding position in the plane. However, edge features may also exist in the texture of the product surface, making it difficult to distinguish between texture and burrs. Since the burr structure is a raised portion in the product and changes in shape in space, it can be seen that during the rotation of the tooling to be deburred, the texture variation is small and the burrs may show significant differences. Based on the above characteristics, this embodiment will then combine the texture representation value of each corner point in the corresponding scale map of all monitoring images under the same viewing angle to obtain the texture feature value of each corner point in the monitoring image.

[0079] Next, we will still use a corner point as an example to illustrate. Specifically, any corner point in any monitoring image is recorded as the corner point to be analyzed. Calculate the variance of the texture representation values of the corresponding positions of the corner point to be analyzed in all scale maps. It should be noted that the corner point to be analyzed may only exist in part of the scale maps. Therefore, when calculating the variance, only the variance of all texture representation values of the position in the scale map where the position exists is calculated. Record the variance of the texture representation values of the corresponding positions of the corner point to be analyzed in all scale maps as the correction coefficient. Use the correction coefficient to correct the texture representation values of the corresponding positions of the corner point to be analyzed in all monitoring images at the same viewing angle in all scale maps to obtain the texture feature value of the position of the corner point to be analyzed in the monitoring image. Specifically, the cumulative sum of the products of the texture representation values of the corresponding positions of the corner point to be analyzed in all monitoring images at the same viewing angle in all scale maps and the correction coefficient is determined as the texture feature value of the position of the corner point to be analyzed in the monitoring image.

[0080] In this embodiment, a specific calculation formula for the texture feature value is given. The texture feature value at the i-th corner point position can be expressed as:

[0081]

[0082] Among them, T i Indicates the texture feature value of the i-th corner point position, B i ′ ,z,x represents the texture representation value of the i-th corner point in the z-th scale map corresponding to the x-th surveillance image under the same viewing angle, q i represents the correction coefficient of the i-th corner point position, S represents the number of surveillance images under the same viewing angle, and m represents the number of scale maps corresponding to the x-th surveillance image under the same viewing angle.

[0083] It should be noted that for any position in the monitoring image, after the monitoring image is transformed, the position may not appear in some scale maps, and the texture feature value of the position in the scale map where the position does not appear is 0.

[0084] Using the above method, the texture feature value of each corner point in the monitoring image is obtained.

[0085] For any surveillance image: based on the texture eigenvalues of all corner points in each category in each surveillance image, the maximum inter-class variance method is used to divide all corner points in each category in each surveillance image into two subcategories again; the average texture eigenvalues of all corner points in each subcategory are calculated and recorded as the average texture eigenvalue of each subcategory; the subcategory with the largest average texture eigenvalue in the first category is recorded as the first subcategory, and the subcategory with the largest average texture eigenvalue in the second category is recorded as the second subcategory; all corner points in the first subcategory and the edge lines between the corner points enclose the first burr area; all corner points in the second subcategory and the edge lines between the corner points enclose the second burr area. The first burr area is the large burr area, and the second burr area is the small burr area.

[0086] So far, the burr area has been screened out using the method provided in this embodiment.

[0087] This embodiment first performs a Gaussian pyramid scale transform on each monitoring image of the tooling fixture to be deburred to obtain a corresponding scale map, then performs corner point detection on each monitoring image and scale map, and then divides the corner points in the monitoring image into two categories based on the presentation of the corner points in the scale map. One category is stable points with obvious scale features, that is, structural corner points or large burr corner points of the tooling fixture itself to be deburred, and the other category is points with unstable scale features, that is, structural corner points or small burr corner points of the tooling fixture itself. Combined with the relative distance and edge change between each corner point in the scale map and other corner points, the texture distribution of the corner points in the scale map is evaluated to obtain a texture representation value. Combined with the texture representation value of each corner point in the corresponding scale map in all monitoring images from the same perspective, the texture feature value of each corner point in the monitoring image is obtained. The texture feature value is used to again screen the corner points in each category to determine the burr area, thereby implementing subsequent deburring processing, making the positioning of the burr area more accurate and improving the burr removal effect.

[0088] An embodiment of a deburring device for a tooling fixture:

[0089] like Figure 2 As shown, the figure shows a structural block diagram of a deburring device for a tooling fixture, the device includes an image acquisition module, a texture representation value determination module, a burr area screening module and a deburring module;

[0090] The image acquisition module is used to obtain several monitoring images of the tooling fixture to be deburred at each viewing angle;

[0091] The texture representation value determination module is used to perform Gaussian pyramid scale transformation on each surveillance image and perform corner detection. The corner points in each surveillance image are divided into two categories based on their presentation in the scale map. The texture representation value of each corner point in the scale map is obtained by combining the relative distance and edge changes between each corner point and other corner points in the corresponding scale map of each surveillance image.

[0092] The burr area screening module is used to combine the texture representation value of each corner point in the corresponding scale map of all monitoring images under the same viewing angle to obtain the texture feature value of each corner point in the monitoring image; and use the texture feature value to screen the burr area for each category;

[0093] The deburring module is used to screen the obscured structural feature points on the tooling to be deburred based on the relative distance between each position of the tooling to be deburred and the burr area and the fact that each position is judged as a corner point in the monitoring image under the same viewing angle; reconstruct the three-dimensional model of the tooling to be deburred by combining the obscured structural feature points and all corner points in all monitoring images under different viewing angles; and plan the deburring route of the deburring equipment and perform deburring processing based on the three-dimensional model and the position of the burr area.

[0094] The image acquisition module, texture expression value determination module and burr area screening module of the deburring equipment of the tooling fixture are used to execute the steps in the embodiment of the burr automatic positioning method. The deburring module of the deburring equipment of the tooling fixture is used to execute the following steps: based on the relative distance between each position of the tooling fixture to be deburred and the burr area and the situation that each position is judged as a corner point in the monitoring image under the same perspective, the obscured structural feature points on the tooling fixture to be deburred are screened; based on the obscured structural feature points and all corner points in all monitoring images under different perspectives, a three-dimensional model of the tooling fixture to be deburred is reconstructed; based on the three-dimensional model and the position of the burr area, a deburring route of the deburring equipment is planned and deburring processing is performed.

[0095] The purpose of extracting corner points from surveillance images is to match the content between surveillance images taken by different cameras. Corner points can be used to match images. When matching images taken at different positions, the area where the large burrs are located may block the structural feature points at the original position. At the same time, since the two cameras cannot fully capture all the information of the tooling fixture, the tooling fixture needs to be rotated during the shooting process. By rotating the camera at the next position, image information of different areas can be captured. When performing the above matching, the image information captured by the camera itself can be matched.

[0096] For any position of the tooling to be deburred:

[0097] Calculate the minimum value of the distance between the position and all corner points in the first subcategory; count the number of times the position is determined to be a corner point in each surveillance image under the same viewing angle; and obtain the probability that the position in each surveillance image belongs to an occluded structure point based on the minimum value and the number of times the position is determined to be a corner point in each surveillance image under the same viewing angle.

[0098] In this embodiment, a specific calculation formula for the probability of belonging to an occluded structure point is given. The probability that the r-th pixel position in any monitoring image belongs to an occluded structure point can be expressed as:

[0099]

[0100] Among them, Z r Indicates the probability that the rth pixel position in the monitoring image belongs to the occluded structure point, min(L r,x ) represents the minimum distance between the rth pixel position in the surveillance image and all corner points in the first subclass, S represents the number of surveillance images under the same viewing angle, and D rx represents the corner point feature value corresponding to the r-th pixel position in the monitoring image, min() represents the minimum function, λ represents the preset adjustment parameter, and norm() represents the normalization function.

[0101] The process of obtaining the corner point feature value corresponding to the r-th pixel position in the monitoring image is as follows: if the r-th pixel position in the monitoring image is determined to be a corner point, then its corresponding corner point feature value is 1; if the r-th pixel position in the monitoring image is not determined to be a corner point, then its corresponding corner point feature value is 0, that is, the corner point feature value represents the number of times the corresponding position is determined to be a corner point.

[0102] The smaller the minimum distance between the rth pixel and all corner points in the first subclass, and the more times the rth pixel is identified as a corner point, the greater the probability that the rth pixel in the surveillance image is an occluded structure point. It should be noted that the probability of a pixel at the same position in all surveillance images from the same viewing angle being an occluded structure point is the same.

[0103] Using this method, we can determine the probability of each location in the surveillance image belonging to an occluded structure point. If the probability is greater than a probability threshold, the location in the surveillance image is determined to be an occluded structure point, effectively filtering out multiple occluded structure points. In this embodiment, the probability threshold is 0.75; in specific applications, the implementer can set this threshold based on specific circumstances.

[0104] The occluded structural points and corner points are used to match the images captured by different cameras at the same time. The specific matching operation is an existing method and will not be described in detail here. The successfully matched point pairs are used to calculate the disparity, that is, the horizontal position difference of the same point in different images. The depth information of each position is calculated based on the disparity and the parameters of the binocular camera (such as focal length, baseline distance, etc.). The feature points in the two-dimensional image coordinate system are converted to the three-dimensional world coordinate system to obtain the three-dimensional coordinates of each position, that is, the three-dimensional model of the tooling fixture to be deburred is constructed. The reconstruction method of the three-dimensional model is an existing technology and will not be described in detail here. The location of the burr in the three-dimensional model of the tooling fixture to be deburred is marked based on the location of the burr area in the image.

[0105] Trajectory planning involves planning a linear trajectory in Cartesian space for the end of the robotic arm based on the requirements of the polishing equipment to meet the actual task requirements. The MoveIt Kinematics and Dynamics Library (KDL) plug-in is used to obtain the inverse kinematic solution for the end position. For example, given the starting coordinates of the end of the robotic arm are (x1, y1, z1), and the end coordinates are (x2, y2, z2), it moves along a straight line from the starting point to the end point, with N interpolation times. The KDL plug-in is used to calculate the joint values corresponding to each segment of the linear trajectory, and the joint information is passed to the controller. The deburring trajectory is planned based on the obtained corner point information, thus completing the deburring route planning for the deburring equipment and then performing the deburring process.

[0106] It should be understood that Figure 2 The block diagram of the deburring device for a tooling fixture and its modules shown can be implemented in various ways. For example, in some embodiments, the device and its modules can be implemented using hardware, software, or a combination of software and hardware. The hardware portion can be implemented using dedicated logic; the software portion can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or specially designed hardware. Those skilled in the art will understand that the above-mentioned methods and systems can be implemented using computer-executable instructions and / or contained in processor control code, such as provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The system and its modules of this specification can be implemented not only by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips and transistors, or programmable hardware devices such as field programmable gate arrays and programmable logic devices, but can also be implemented by software, such as executed by various types of processors, or by a combination of the above-mentioned hardware circuits and software (e.g., firmware).

[0107] For more details about the above modules, please refer to other places in this manual and will not be repeated here.

[0108] In other embodiments, a medium is also provided, which stores at least one program executable by a computer. When the at least one program is executed by a computer, the computer executes the steps of the automatic burr location method in the above embodiment. The medium may be a computer-readable storage medium.

[0109] Among them, the provided equipment and media are used to execute the corresponding methods provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding methods provided above, and will not be repeated here.

[0110] It should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A burr automatic positioning method, characterized in that: The method comprises the following steps: Acquire several monitoring images of the tooling to be deburred at each viewing angle; Each surveillance image is scaled using a Gaussian pyramid and corner detected. The corners in each surveillance image are divided into two categories based on how they appear in the scale map. The texture representation value of each corner in the scale map is obtained by combining the relative distances and edge changes between each corner in the corresponding scale map. Combining the texture representation value of each corner point in the corresponding scale map of all monitoring images under the same viewing angle, the texture feature value of each corner point in the monitoring image is obtained; using the texture feature value to filter the burr area for each category; Get the texture representation value of each corner point in the scale map, including: The corner point directly connected to the candidate corner point through the edge line is recorded as the reference corner point of the candidate corner point; the discrete degree of the slope of all pixels on the edge line between the candidate corner point and each of its reference corner points is calculated and recorded as the first eigenvalue between the candidate corner point and each of its reference corner points; Obtaining a texture representation value of the candidate corner point based on the Euclidean distance between the candidate corner point and all reference corner points and the first eigenvalue, wherein the Euclidean distance is negatively correlated with the texture representation value, and the first eigenvalue is positively correlated with the texture representation value; The candidate corner point is any corner point in any scale map; Get the texture feature value of each corner point in the monitoring image, including: The variance of the texture representation values of the corresponding positions of the corner points to be analyzed in all scale images is recorded as the correction coefficient; The texture feature value of the corner point to be analyzed in the surveillance image is determined as the cumulative sum of the products of the texture representation values of the positions corresponding to the corner point to be analyzed in all surveillance images under the same viewing angle in all scale images and the correction coefficient; The corner point to be analyzed is any corner point in any monitoring image.

2. The method for automatically locating burrs according to claim 1, characterized in that: According to the presentation of the corner points in each surveillance image in the scale map, the corner points in each surveillance image are divided into two categories, including: For any corner point in the candidate monitoring image, the total number of times the corner point is judged as a corner point in all scale images corresponding to the candidate monitoring image is recorded as the scale feature value of the corner point; Based on the scale feature values of all corner points in the candidate monitoring image, the maximum inter-class variance method is used to classify all corner points in the candidate monitoring image into two categories; the candidate monitoring image is any monitoring image.

3. The method for automatically locating burrs according to claim 1, characterized in that: Calculating the discrete degree of the slopes of all pixels on the edge line between the candidate corner point and each of its reference corner points includes: taking the variance of the slopes of all pixels on the edge line connecting the candidate corner point and each of its reference corner points as the corresponding discrete degree.

4. The method for automatically locating burrs according to claim 2, wherein: The method of using the texture feature value to divide and filter the burr area into each category includes: For any surveillance image: Calculate the average of the scale characteristic values of all corner points in each category and record it as the average scale characteristic value of each category; record the category with the largest average scale characteristic value as the first category, and the category with the smallest average scale characteristic value as the second category; Based on the texture feature values of all corner points in each category in each surveillance image, the maximum inter-class variance method is used to divide all corner points in each category in each surveillance image into two subcategories again; Calculate the average value of the texture feature values of all corner points in each subclass and record it as the average texture feature value of each subclass; record the subclass with the largest average texture feature value in the first class as the first subclass, and record the subclass with the largest average texture feature value in the second class as the second subclass; A burr area is determined according to the first subcategory and the second subcategory.

5. The method for automatically locating burrs according to claim 4, characterized in that: The determining of the burr area according to the first subcategory and the second subcategory includes: All corner points in the first subclass and edge lines between the corner points form a first burr area; All corner points in the second subclass and the edge lines between the corner points form a second burr area; The burr region includes a first burr region and a second burr region.

6. A deburring device for a tooling fixture, characterized in that: The burr area screening module of the device is used to implement the method according to any one of claims 4 to 5. The device includes: An image acquisition module is used to obtain several monitoring images of the tooling to be deburred at each viewing angle; The texture representation value determination module is used to perform Gaussian pyramid scale transformation on each surveillance image and perform corner detection. The corner points in each surveillance image are divided into two categories based on their presentation in the scale map. The texture representation value of each corner point in the scale map is obtained by combining the relative distance and edge changes between each corner point and other corner points in the corresponding scale map of each surveillance image. The burr area screening module is used to combine the texture representation value of each corner point in the corresponding scale map of all monitoring images under the same viewing angle to obtain the texture feature value of each corner point in the monitoring image; and use the texture feature value to screen the burr area for each category; The deburring module is used to screen the obscured structural feature points on the tooling to be deburred based on the relative distance between each position of the tooling to be deburred and the burr area and the fact that each position is judged as a corner point in the monitoring image under the same viewing angle; reconstruct the three-dimensional model of the tooling to be deburred by combining the obscured structural feature points and all corner points in all monitoring images under different viewing angles; and plan the deburring route of the deburring equipment and perform deburring processing based on the three-dimensional model and the position of the burr area.

7. The deburring device for a tooling fixture according to claim 6, characterized in that: The method of screening the obscured structural feature points on the tooling to be deburred according to the relative distance between each position of the tooling to be deburred and the burr area and whether each position is determined to be a corner point in the monitoring image at the same viewing angle includes: For any position of the tooling to be deburred: Calculate the minimum value of the distance between the position and all corner points in the first subclass; count the number of times the position is determined to be a corner point in each surveillance image under the same viewing angle; and obtain the probability that the position is an occluded structure point in each surveillance image based on the minimum value and the number of times the position is determined to be a corner point in each surveillance image under the same viewing angle; If the probability is greater than the probability threshold, it is determined that the position in the corresponding monitoring image belongs to an obscured structure point.

Images