Multi-Camera Calibration And Image Mosaic Method For Detecting Parts

As a non-contact optical sensor measurement system,the machine vision system integrates the computer software and hardware technology and image processing technology.It can automatically obtain the information of the detected object from the collected images to replace the human eye Measurement and judgment.Therefore,It is widely used in automatic production line condition monitoring,product testing and quality control and other fields to improve the flexibility and automation of the production line.However,in the use of machine vision to detect large objects,if a single camera to obtain a complete image of the detection object,often because of the large shooting range,we usually cannot get clear images.At this time,we must use multiple cameras to complete the detection task.so in recent years,multi-camera visual inspection has become a research focus,but multi-camera calibration and image stitching process is often more complex.In this thesis,multi-camera calibration and image mosaic method for large-scale part detection are researched.First of all,as to multi-camera calibration,according to the industrial field detection environment,this thesis designed a suitable multi-camera mechanical splicing program.Then,for the position and task area of the multi-camera,a calibrator with obvious characteristics was designed.The feature points on the calibrator are extracted by using the Hough transform so that the target area of the camera is obtained and and the regional calibration of the multi-camera is realized.Secondly,in the aspect of image mosaic,this thesis introduces the principle and process of SIFT algorithm to extract image features in the first place.Then uses K-D tree algorithm to characterize the extracted feature points.After that,it uses distance ratio method and RANSAC algorithm to eliminate false matching.In order to make the image transition smoothly,this thesis uses the weighted average method to fuse the image to realize the accurate and smooth image splicing process.Finally,this thesis uses the programmed multi-camera calibration program to carry out the calibration experiment.The image processing target area of each camera is well extracted.The multi-camera image splicing experiment was carried out by using the developed multi-camera image mosaic software,and a good panoramic image was obtained.And the multi-camera calibration and image splicing method is applied to the optical fiber preform test project with the frame size of 600 mm.The calibration and stitching effect are obtained,which proves that the method proposed in this thesis is feasible and effective.