A Neural-Network-Based Visual Servoing System for Shoe Spraying Trajectory Planning

• Main Authors: Kai-Min Zhang, 張凱閔
• Other Authors: Shou-Tao Peng
• Format: Others
• Language: zh-TW
• Published: 103
• Online Access: http://ndltd.ncl.edu.tw/handle/94n33p

碩士 === 南台科技大學 === 機械工程系 === 102 === This research investigates the use of Neural-Network-Based Visual Servoing System to construct the glue path of the shoes. The research adopts the ”Eye-in-hand” visual servo configuration in which the camera is installed on the end of the actuator in order to get the image with higher resolution and to prevent vision processing from obstruction. The un-calibrated Self-Organizing Maps (SOM) neural network is introduced to build the map of relationship between image position and machine movement. This research adopts the digital image processing and Hough transform to detect and recognize the glue path of shoes. Also, it uses the mapping model of SOM to drive the ” Eye-in-hand” visual servoing mechanism and makes the mechanism move and follow the glue path. The experimental result shows that after sufficiently repetitions of learning, the system can achieve accurate tracking effect of glue path processing.