A Novel Algorithm for Detecting Pedestrians on Rainy Image.

Yuhang Liu,Yuchen Wang,Jianxiao Ma,Chenhong Zong

doi:10.3390/s21010112

Abstract

Pedestrian detection is widely used in cooperative vehicle infrastructure systems. Traditional pedestrian detection methods perform sufficiently well under sunny scenarios and obtain trustworthy traffic data. However, the detection drastically decreases under rainy scenarios. This study proposes a pedestrian detection algorithm with a de-raining module that improves detection accuracy under various rainy scenarios. Specifically, this algorithm determines the density information of rain and effectively removes rain streaks through the de-raining module. Then the algorithm detects pedestrians as a pair of keypoints through the pedestrian detection module to solve the problem of occlusion. Furthermore, a new pedestrian dataset containing rain density labels is established and used to train the algorithm. For the scenarios of light, medium, and heavy rain, extensive experiments on synthetic datasets demonstrate that the proposed algorithm increases AP (average precision) of pedestrian detection by 21.1%, 48.1%, and 60.9%. Moreover, the proposed algorithm performs well on real datasets and achieves improvements over the state-of-the-art methods, which reveals that the proposed algorithm can significantly improve the accuracy of pedestrian detection in rainy scenarios.

Highlights

It can be observed that the algorithm effectively removes rain streaks and accurately detects pedestrians while maintaining image details
Most pedestrians on rainy days wear raincoats or carry umbrellas, so there is a lot of occlusions among pedestrians
In view of the particularity of pedestrian detection on rainy days, this paper proposed a novel algorithm with a de-raining module for detecting pedestrians

Summary

Introduction

The existing synthetic datasets lack the scale and density of rain streaks corresponding to each synthetic rainy image, and there are few pedestrian targets in the image. These datasets cannot meet the requirement of training pedestrian detection algorithm on rainy days. There are a lot of occlusions among pedestrians To solve these problems, we propose a pedestrian detection algorithm for the scenario of rainy days. Fu et al [16] first introduced deep learning methods into the problem of de-raining, breaking down rainy images into low-frequency and high-frequency sections, and mapping high-frequency portions to the rain streak layer.

Pedestrian Detection Methods

The Proposed Algorithm

Synthetic Dataset

Results on the Real-World Images

Conclusions