Research on Reference Target Detection of Deep Learning Framework Faster-RCNN

Abstract

In recent years, with the continuous development and progress of computer-related facilities and equipment, deep learning has been widely used in computer fields such as image classification and target detection. Many researchers have proposed a new learning objective -- detection algorithm on the basis of traditional computer research. In this paper, by comparing the target detection algorithm with the traditional manual feature extraction and feature design algorithm, it can be found that the convolutional neural network algorithm has good effects in feature expression ability, semantic expression ability, robustness and other aspects. The deep learning target detection algorithms mainly involved in this paper include R-CNN, Fast-RCNN, Faster-RCNN, YOLO and SSD. The research mainly considers the detection speed and the sensitivity of small target detection comprehensively, and realizes apple detection under natural light through improving the based model of Faster-RCNN.