Research On Crucible Defect Detection Method Based On Machine Vision

In order to solve the shortcomings that existed in the manual detection method of the traditional fused silica crucible(hereinafter refers to as the crucible):it is ineffective due to the labors' experience,degree of fatigue and so on.This paper designs a robotic-based crucible detection system which is based on machine vision.The system detects crucible defects and extracts "3 points" of circumscribed circle "3 points" by image processing technology,and through the establishment of the reference coordinate system and the robot hand-eye calibration and other coordinate system conversion,the pixel coordinates are converted into the robot base coordinates,and at the same time,the posture angle of the robot hand is defined,and the translation and rotation of the three axes are combined.When the coordinates were sent to the robot system,according to the order of a circular arc to send,that is:to send three coordinates to define three attributes 1,2,3,and then the robot activates the coordinates of the corresponding properties to perform a point-by-point movement to complete a full circle,finally,the robot visually marks the detected defectsBecause the experiment only adopts a single camera,the depth of the camera's optical axis can not be obtained.Therefore,a reference coordinate system is established as a bridge between the coordinate systems.This paper adopts image preprocessing methods such as image grayscale,image intensity inversion,background subtraction and then adopts threshold segmentation and feature extraction to detect the shortcomings.Among them,the background subtraction algorithm is used to make different grayscale correction according to the uneven lighting conditions,so that the grayscale value of the non-defect location is close to 0 and the location of the defect has a certain grayscale value.According to all the above processes,it achieve the goal to highlight the foreground of the pictures.At the same time,an image mosaic algorithm based on translation vector is proposed.This method can deal with the situation that a defect is segmented by several image.At the same time,with the template matching algorithm to avoid the same defects in different images detected repeatedly or the occurrence of marked confusion,the problem of insufficient number of feature points in the image fusion splicing method is solved,and the points at the edges of the image splicing are theoretically the same point.In addition,this paper write a upper computer vision system platform.This platform can capture images,grab a frame of images and also do the image rectifications,image intensity inversion,background subtraction,threshold segmentation,morphology,feature extraction and a series of processes to extract the final image defects for display.When the system automatically monitors defects and marks the defects,the number of monitored defects and the coordinates of the image are returned to the visual interface.In addition,the software platform also includes some basic settings of the system,such as the internal parameters of the camera,the parameters of the robot's hand-eye calibration,the intermediate conversion parameters and the other reference coordinate system,etc.The parameters enable the system to accurately do the detection system in different situations.The results show that the crucible detection system designed in this paper can meet the actual needs,and also has the following advantages:rapidity,stability and practicability.All in all,it has a great application prospect.