Design and Implementation of Campus Pedestrian Detection based on Unmanned Robot “Sweeper”

Abstract

Based on the unmanned autonomous sanitation robot “Sweeper” introduced by Shanghai Sanda University, the target detection technology is studied by adopting the idea of blob analysis to improve the target detection ability and ground cleaning function of “Sweeper” with more than 300 images of campus pedestrians. After HSV thresholding binarization, image morphology processing and image contour detection fitting, erosion operation with expansion operation and different rectangular structured element adjustment, the noise reduction of the surrounding environment and the extraction of campus pedestrian contours were achieved and campus pedestrian image detection is obtained. The test shows that the system can achieve 100% success rate for single pedestrian detection in any environment, and 65%-75% for multiple pedestrian images, and the detection success rate of stable natural environment will be better than that of complex natural environment.