Turn Signal Detection During Nighttime by CNN Detector and Perceptual Hashing Tracking

Abstract

Detecting vehicle turn signals at night is critical for both assistant driving systems and autonomous driving systems. In this paper, we propose a novel method that consists of detection and tracking modules to achieve a high level of robustness. For nighttime vehicle detection, a Nakagami-image-based method is used to locate the regions containing vehicle lights. At the same time, a set of vehicle object proposals is generated using a region proposal network based on convolutional neural network (CNN) feature maps. Then, the light regions and proposals are combined to generate the regions of interest (ROIs) for the further detection. Vehicle candidates are extracted from the ROIs using a softmax classifier with CNN-based features. For the tracking module, we propose a perceptional hashing algorithm to track these vehicle candidates. During the tracking, turn signals are detected by analyzing the continuous intensity variation of the vehicle box sequences.

Existing System

Dense Spatio-Temporal Context Learning Method.

Proposed System

The proposed method consists of two major steps, vehicle detection and tracking at nighttime and turn signal detection. For nighttime vehicle detection, we propose a novel region of interest (ROI) selection algorithm for the following classification instead of the traditional sliding window method. For an image, Nakagami image combing with an HSI segmentation method is utilized to locate vehicle light areas. At the same time, a selection of vehicle object proposals is obtained using RPN based on CNN feature maps. Then, the two results are combined to generate the ROIs by eliminating most non-vehicle lights such as street lights and billboard lights. We represent each ROI using a CNN-based feature and classify this feature using a softmax classifier. The flashing of turn signals is a temporal dynamic process that requires continuous observation of the detected vehicle. Consequently, we propose a fast vehicle tracking method based on a perceptual hashing algorithm. During tracking steps, turn signals are recognized by analysing the continuous intensity variation of the vehicle box sequences.

Architecture

BLOCK DIAGRAM