Research on pedestrian occlusion detection based on SSD algorithm

Abstract

As a basic identification technology, pedestrian detection provides technical support for many areas such as security monitoring and autonomous driving, and has a wide range of application scenarios. Based on the Single Shot MultiBox Detector (SSD) target detection algorithm, this paper trains a pedestrian detection system based on the SSD target detection framework with a self-built occlusion pedestrian dataset for the specific target of occlusion pedestrians. The test set and the re-annotated INRIA test set were used to compare the HOG+SVM based pedestrian detection system and the trained SSD model in OpenCV. The experimental results show that the detection effect of the SSD model is significantly better than the traditional pedestrian detection system based on HOG+SVM. The features learned by the deep convolutional neural network are more robust.