Visual detection method for appearances

Abstract

The invention provides a visual detection method for appearances which comprises the following steps: an optical guiding step: calculating the similarity of images to be compared by taking a referenceimage as a standard, so as to ensure the optical imaging consistency of the images acquired by the same batch of products on different machines; and in the visual guidance step, when small part products need to reach preset high precision, otherwise, indicating that a mechanical arm cannot conduct normal feeding; acquiring the deviation angle, the X position and the Y position of a product through a visual guidance algorithm before product feeding, and notifying the machines to conduct adjustment to ensure the product feeding precision. According to the invention, a deep learning and traditional image processing machine vision detection method is adopted, so that the requirement of similarity comparison in an image acquisition stage exists, and the invention is used for ensuring that consistent images are acquired, thereby ensuring accuracy of subsequent data for depth model detection and the accuracy of image detection.

Description

technical field [0001] The present invention relates to the field of graphic detection, in particular to a visual appearance detection method. Background technique [0002] Traditional optical guidance mainly relies on the subjective experience of the implementation engineer, and controls the optical imaging quality by adjusting the camera focal length, aperture, and working distance. The effect is average, and the imaging optical consistency of different machines is poor. [0003] The traditional visual guidance process is generally as follows: a loading mechanism (manipulator, suction cup, etc.) with an industrial camera takes a photo before each grabbing of the material, and calculates the information of the position deviation and angle deviation of the material to be grasped through the visual software to ensure It can meet the blanking accuracy when blanking. The main defect is that the accuracy cannot be guaranteed. [0004] Traditional target detection algorithms of...

AI Technical Summary

Problems solved by technology

[0005] Traditional target detection algorithms often use a sliding window strategy to traverse the entire image, and then use Haar, SIFT, HOG and other feature extractors to extract target objects, and then use SVM, Adaboost and other classifiers to classify the extracted targets. Although this exhaustive strategy includes all possible positions of the target, its disadvantages are also obvious: the time complexity is too high, and too many redundant windows are generated, which also seriously affects the speed and performance of subsequent feature extraction and classification.

Images