Automatic obstacle avoiding via Haar-like features and cascade classifier on wheeled mobile robots

• Main Authors: LEE, CHUNG-JUNG, 李振榮
• Other Authors: TSAI, CHUN-MING
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/78693196021232115402

碩士 === 臺北市立大學 === 資訊科學系 === 104 === When robot is working indoor, it is important to detect and avoid obstacles. A lot of robots use infrared or ultrasonic to detect obstacles, but it does not determine the position of obstacles. The purpose of this study is using image processing to let wheeled mobile robot to recognize the existence of cones as warning and avoid it. Besides using features like colors and edges to detect cones, we use Haar like feature to obtain the cascade classifier to detect cones. We got images of the cones from the camera of the WMR and sent the image by through RF to the PC. Then we collected the positive samples with cones and negative samples without cones. Next, we used these positive samples and negative samples to extract Haar like features. And we used Adaboost algorithm to improve the cascade classifier. The result of the experiment shows that wheeled mobile robot can detect cones efficiently and avoid them.