Search

CN-122023453-A - Self-adaptive CCD visual analysis system for special-shaped hardware

CN122023453ACN 122023453 ACN122023453 ACN 122023453ACN-122023453-A

Abstract

The invention relates to the technical field of computer vision and discloses a self-adaptive CCD vision analysis system for special-shaped hardware, which comprises a dynamic optical calibration module, an edge detection module, a dynamic region of interest segmentation module, a self-adaptive parameter analysis module, a coordinate construction mapping module and a global grabbing pose determination module; the method comprises the steps of collecting an initial image of a hardware, carrying out optical calibration, carrying out edge detection on the calibrated image to obtain a complete contour of the hardware, carrying out region segmentation on the complete contour to obtain a dynamic region of interest, carrying out self-adaptive parameter analysis on the dynamic region of interest to obtain accurate characteristic parameters, constructing a workpiece coordinate system, mapping the accurate characteristic parameters onto the workpiece coordinate system to obtain an associated characteristic data set, and determining global grabbing pose information based on grabbing point definition rules and the associated characteristic data set.

Inventors

  • LI SHULIN

Assignees

  • 东莞骏伟塑胶五金有限公司

Dates

Publication Date
20260512
Application Date
20260414

Claims (10)

  1. 1. The self-adaptive CCD visual analysis system for the special-shaped hardware is characterized by comprising a dynamic optical calibration module, an edge detection module, a dynamic region-of-interest segmentation module, a self-adaptive parameter analysis module, a coordinate construction mapping module and a global grabbing pose determination module, wherein: The dynamic optical calibration module is used for acquiring an initial image of the special-shaped hardware, and carrying out dynamic optical calibration on the initial image to obtain a calibrated image of the special-shaped hardware; The edge detection module is used for carrying out edge detection on the calibrated image to obtain the complete outline of the special-shaped hardware; the dynamic region of interest segmentation module is used for carrying out region segmentation on the complete contour based on curvature mutation points of the complete contour to obtain a dynamic region of interest of the special-shaped hardware; the self-adaptive parameter analysis module is used for carrying out self-adaptive parameter analysis on the dynamic region of interest to obtain accurate characteristic parameters of the special-shaped hardware; The coordinate construction mapping module is used for constructing a local workpiece coordinate system of the special-shaped hardware, and mapping the accurate characteristic parameters to the local workpiece coordinate system to obtain an associated characteristic data set of the special-shaped hardware; The global grabbing pose determining module is used for determining global grabbing pose information of the special-shaped hardware based on a preset grabbing point definition rule and the associated characteristic data set.
  2. 2. The adaptive CCD vision analysis system for special-shaped hardware as set forth in claim 1, wherein the dynamic optical calibration module is configured to, when performing acquisition of an initial image of the special-shaped hardware and performing dynamic optical calibration on the initial image to obtain a calibrated image of the special-shaped hardware: Acquiring an initial image of the special-shaped hardware based on a CCD camera; Performing pixel gray level statistics on the initial image to generate a gray level histogram, and extracting the integral gray level distribution characteristics of the initial image from the gray level histogram; Dividing the initial image into local blocks, and determining local reflection characteristics based on gray variance of the local blocks; Generating equipment adjusting instructions of the special-shaped hardware according to the integral gray level distribution characteristics and the local reflection characteristics; based on the equipment adjusting instruction, adjusting the light source brightness of the special-shaped hardware and adjusting the exposure time of the CCD camera; And acquiring the calibrated image of the special-shaped hardware under the adjusted light source brightness and exposure time.
  3. 3. An adaptive CCD vision analysis system for special-shaped hardware as claimed in claim 2, wherein said dynamic optical calibration module, when performing said dividing of said initial image into local areas and determining local reflectance characteristics based on the gray variance of said local areas, is specifically configured to: dividing the initial image into rectangular local blocks according to a preset grid size; calculating the variance of the gray values of the pixel points in the local block to obtain the gray variance value of the rectangular local block; Normalizing the gray variance value to obtain a normalized variance value of the rectangular local block; comparing the normalized variance value with a preset light reflection threshold, and judging that the rectangular local block is a strong light reflection area when the normalized variance value is larger than the light reflection threshold, or else, is a weak light reflection area; And integrating the number and the distribution positions of the strong light reflecting areas into the local light reflecting characteristics of the local area blocks.
  4. 4. The adaptive CCD vision analysis system for a special-shaped hardware as set forth in claim 1, wherein the edge detection module, when performing edge detection on the calibrated image, is specifically configured to: Performing Gaussian filtering on the calibrated image to obtain a smoothed image of the special-shaped hardware; performing edge extraction on the smoothed image to obtain an initial edge image of the special-shaped hardware; Performing morphological closing operation on the initial edge image to obtain continuous edge images of the special-shaped hardware; and carrying out contour tracking on the continuous edge images, and extracting the closed contour of the outermost layer to obtain the complete contour of the special-shaped hardware.
  5. 5. The adaptive CCD vision analysis system for a special-shaped hardware as set forth in claim 1, wherein the dynamic region of interest segmentation module is configured to, when executing the curvature mutation point based on the complete contour, perform region segmentation on the complete contour to obtain the dynamic region of interest of the special-shaped hardware: performing spline curve fitting on the complete profile to obtain a smooth profile curve of the special-shaped hardware; equidistant sampling is carried out on the profile curve to obtain profile points, and a curvature sequence of the profile curve is constructed based on the curvature value of the profile points; detecting a local extremum of the curvature sequence, and taking a curvature point with a curvature value larger than a curvature threshold value in a local extremum detection result as a curvature abrupt change point based on a preset curvature threshold value; taking the curvature abrupt change points as dividing points, and dividing the complete contour into continuous contour segments; And carrying out geometric feature recognition on the profile section to obtain the shape feature of the profile section, and carrying out region classification on the profile section based on the shape feature to obtain a dynamic region of interest, wherein the dynamic region of interest comprises a plane region, a bending region and a hole site region.
  6. 6. The adaptive CCD vision analysis system for special-shaped hardware as set forth in claim 5, wherein the dynamic region-of-interest segmentation module, when executing the local extremum detection for the curvature sequence and based on a preset curvature threshold, uses a curvature point with a curvature value greater than the curvature threshold in the local extremum detection result as a curvature abrupt change point, specifically: carrying out smooth filtering on the curvature sequence to obtain a smooth curvature sequence; taking the local maximum value point and the local minimum value point on the smooth curvature sequence as candidate extreme points of the smooth curvature sequence; Determining the curvature variation amplitude of the candidate extremum point based on the peak Gu Chazhi of the adjacent candidate extremum point; And marking the candidate extreme points with the curvature change amplitude larger than a preset amplitude threshold as curvature abrupt points.
  7. 7. The adaptive CCD vision analysis system for a special-shaped hardware as set forth in claim 1, wherein the adaptive parameter analysis module is configured to, when performing adaptive parameter analysis on the dynamic region of interest to obtain the accurate feature parameter of the special-shaped hardware: Extracting image sub-blocks of the dynamic region of interest; performing pixel gray gradient calculation on the image sub-blocks to obtain a gradient amplitude diagram of the image sub-blocks; According to the statistical distribution of the gradient amplitude diagram, dynamically determining an adaptive segmentation threshold of the dynamic region of interest; binarizing the image subblocks based on the self-adaptive segmentation threshold to obtain a binarized image; performing morphological processing on the binarized image, and performing edge refinement on a morphological processing result to obtain accurate geometric characteristics of the binarized image; and taking the precise geometric characteristic as the precise characteristic parameter of the special-shaped hardware.
  8. 8. The adaptive CCD vision analysis system for special-shaped hardware of claim 7, wherein the adaptive segmentation threshold is calculated as: Wherein, the For the adaptive segmentation threshold value in question, For the arithmetic average of the gradient magnitudes of all pixels in the dynamic region of interest, The coefficients are adjusted for a predetermined positive real number, For the standard deviation of the gradient magnitudes of all pixels in the dynamic region of interest, The coefficients are adjusted for a predetermined positive real number, And (5) a gradient direction consistency coefficient of the dynamic region of interest.
  9. 9. The adaptive CCD vision analysis system for a special-shaped hardware as set forth in claim 5, wherein the coordinate construction mapping module, when executing construction of a local workpiece coordinate system of the special-shaped hardware and mapping the precise feature parameters to the local workpiece coordinate system, is specifically configured to: Selecting a characteristic point of the plane area from the accurate characteristic parameters as a reference point, and selecting a normal vector of the plane area as a reference direction; constructing a local workpiece coordinate system of the special-shaped hardware by taking the datum point as a coordinate origin and the datum direction as a Z-axis direction; converting the image coordinates of the feature points in the accurate feature parameters to the local workpiece coordinate system to obtain converted feature point coordinates; and integrating the converted feature point coordinates and corresponding feature attributes into an associated feature data set of the special-shaped hardware.
  10. 10. The adaptive CCD vision analysis system for a special-shaped hardware according to claim 1, wherein the global capture pose determination module is configured to, when executing the determination of the global capture pose information of the special-shaped hardware based on a preset capture point definition rule and the associated feature dataset: Traversing the plane areas in the associated feature data set, and taking the largest plane area in the plane areas as a grabbing reference plane; Extracting a boundary point coordinate set of the grabbing reference surface from the characteristic data of the grabbing reference surface, and carrying out coordinate averaging on the boundary point coordinate set to obtain grabbing center coordinates of the grabbing reference surface; traversing the hole sites in the associated characteristic data set, and taking the hole site with the minimum aperture as a grabbing positioning hole; extracting the center point coordinates of the grabbing positioning holes based on the characteristic data of the grabbing positioning holes, and extracting the normal vectors of the grabbing positioning holes from the characteristic data of the grabbing reference surface; And determining the gesture angle of the grabbing tool of the special-shaped hardware by taking the central point coordinate of the grabbing positioning hole as a gesture reference point and the normal vector of the grabbing reference plane as a gesture main direction, and integrating the grabbing central coordinate and the gesture angle into global grabbing gesture information of the special-shaped hardware.

Description

Self-adaptive CCD visual analysis system for special-shaped hardware Technical Field The invention relates to the technical field of computer vision, in particular to a self-adaptive CCD (charge coupled device) vision analysis system for special-shaped hardware. Background In the field of automatic precise assembly, visual positioning and grabbing of special-shaped hardware are one of core procedures. CCD visual analysis provides grabbing pose information for a feeding device by collecting workpiece images and extracting geometric features, and the identification accuracy of the device directly determines the automation level and the yield of an assembly line. Along with miniaturization and structural complexity of electronic products, the contour complexity of special-shaped hardware is remarkably improved, and higher requirements are put on the self-adaption capability and anti-interference capability of a visual analysis system. In the prior art, a CCD visual analysis system mostly adopts fixed light source parameters and preset image processing threshold values, and is difficult to adapt to dynamic changes of the reflection characteristics of the surface of the special-shaped hardware along with the angle and material difference of the materials, so that incomplete edge extraction or profile fracture is caused. Meanwhile, the traditional method relies on a pre-established template library for feature matching, and the workpiece is identified by comparing the real-time image with standard features in the template library. In addition, in the prior art, when extracting features of complex structures such as bending areas, hole sites and the like, a fixed binarization threshold is generally adopted, and the segmentation parameters cannot be dynamically adjusted according to local gradient distribution features, so that the extraction precision of the complex structure features is insufficient, and the requirement of high-precision grabbing is difficult to meet. Therefore, development of a CCD vision analysis system capable of adaptively adjusting imaging parameters, dynamically dividing an interested region and intelligently analyzing characteristic parameters is needed to solve the problems of low recognition rate, poor adaptability and insufficient positioning accuracy in the prior art, and improve the reliability and efficiency of automatic feeding of special-shaped hardware. Disclosure of Invention In order to achieve the above purpose, the invention provides a self-adaptive CCD vision analysis system for special-shaped hardware, which is characterized in that the system comprises a dynamic optical calibration module, an edge detection module, a dynamic region of interest segmentation module, a self-adaptive parameter analysis module, a coordinate construction mapping module and a global grabbing pose determination module, wherein: The dynamic optical calibration module is used for acquiring an initial image of the special-shaped hardware, and carrying out dynamic optical calibration on the initial image to obtain a calibrated image of the special-shaped hardware; The edge detection module is used for carrying out edge detection on the calibrated image to obtain the complete outline of the special-shaped hardware; the dynamic region of interest segmentation module is used for carrying out region segmentation on the complete contour based on curvature mutation points of the complete contour to obtain a dynamic region of interest of the special-shaped hardware; the self-adaptive parameter analysis module is used for carrying out self-adaptive parameter analysis on the dynamic region of interest to obtain accurate characteristic parameters of the special-shaped hardware; The coordinate construction mapping module is used for constructing a local workpiece coordinate system of the special-shaped hardware, and mapping the accurate characteristic parameters to the local workpiece coordinate system to obtain an associated characteristic data set of the special-shaped hardware; The global grabbing pose determining module is used for determining global grabbing pose information of the special-shaped hardware based on a preset grabbing point definition rule and the associated characteristic data set. In a preferred embodiment, the dynamic optical calibration module is specifically configured to, when performing acquisition of an initial image of the special-shaped hardware and performing dynamic optical calibration on the initial image to obtain a calibrated image of the special-shaped hardware: Acquiring an initial image of the special-shaped hardware based on a CCD camera; Performing pixel gray level statistics on the initial image to generate a gray level histogram, and extracting the integral gray level distribution characteristics of the initial image from the gray level histogram; Dividing the initial image into local blocks, and determining local reflection characteristics based on gray variance of the lo