Research on Human Target Recognition Algorithm of Home Service Robot Based on Fast-RCNN

According to the needs of users, Home Service Robots gradually work outside. As a result, new requirements for the detection and recognition performance of Home Service Robots are put forward. Compared with indoor environment, outdoor environment is more complex, which brings difficulties to detect objects. But extracting features by Histogram of Oriented Gradient (HOG) method can not work well in complex environment. To solve this problem, depth learning method was introduced. In this paper, a New Region Proposal Network (NRPN) algorithm was presented and a multi-layer Convolution Neural Network(CNN) was built to achieve a better result. A human target detection system was also built based on the Fast-RCNN algorithm combined with the VGG network model on the open source toolkit platform Caffe. The results show that testing on the comprehensive database established in this paper, the new algorithm has improved the detection rate and recognition rate of the target.