| With the rapid development of robot technology, computer technology, image processing and pattern recognition technology, etc. Vision-based industrial robots have been widely used. Among them, machine vision is industrial robot learns an important means of environmental information, can enhance the industrial robot autonomy and flexibility. In the automobile production line, the traditional quality inspection of the solder joints of the vehicle body is carried out by the technical personnel through manual sampling, hand held testing equipment to detect the quality of the solder joint. This method is inefficient, seriously affect the vehicle production capacity. In view of the above problems, it is urgent to develop a set of quality inspection robot system based on machine vision. In this thesis, the industrial robot target recognition and location algorithm for the identification and positioning of the body solder joints were studied in detail. Mainly carried out the following work.Firstly, according to the working principle and technical requirements of the white body solder joint quality inspection robot system, the overall scheme of the system is designed. Use the MOTOMAN-MH12 type robot Yaskawa company as the industrial robot, and the Smartek GC1392C industrial camera and Advantech IPC as the main hardware of machine vision. Construction of the robot system software and hardware framework, build the body solder joint quality inspection robot system platform.Secondly, the image of body solder joints should be preprocessed in order to enhance image quality. The main preprocess ways are the image of the gray, median filtering and image sharpening. The edge feature of the solder joint is extracted, and several classical edge detection operators are analyzed and compared. And the wavelet maximum edge detection operator is used to obtain more superior to the detection results.Thirdly, Random Hough transform (RHT) algorithm was used to realize the detection of circular arc. And it was used to identified the central location of the joints on the image. In order to realize the on-line intelligent identification and classification, the image characteristics of the welding spot are extracted. Based on the analysis of the characteristics of the solder joint image and the actual situation of the project, the image features of solder joints are described by using the characteristic parameters of gray level co-occurrence matrix (GLCM). About body solder joint identification. Based on the traditional support vector machine (SVM), the body solder joint classifier is designed by combining particle swarm fused kernel fuzzy C-means (P-KFCM) and SVM, it can distinguish the solder joints from other interference factors successfully.Finally, the mathematical model of Yaskawa MH-12 robot, hand-eye imaging model and the coordinate transformation were analyzed. Then, the relative position of the coordinate position of solder joint in image and ultrasonic probe on the end of manipulator in the base coordinate system of the robot was calculated. VC++6.0, OpenCV image function library and other software was used in the white body solder joint quality inspection robot system platform to complete the relevant experiment. Based on the proposed algorithm, the experiment that machine vision achieved the body solder joint real-time online positioning was completed In the vehicle body solder joint quality inspection robot system platform. |
PDF Full Text Request