TRAFFIC LIGHT RECOGNITION FOR COMPLEX SCENE WITH FUSION DETECTIONS

Abstract

Traffic light recognition is one of the important tasks in the studies of intelligent transport system. In this paper, a robust traffic light recognition model based on vision information is introduced for on-vehicle camera applications. Our contribution mainly includes three aspects. First, in order to reduce computational redundancy, the aspect ratio, area, location, and context of traffic lights are utilized as prior information, which establishes a task model for traffic light recognition. Second, in order to improve the accuracy, we propose a series of improved methods based on an aggregate channel feature method, including modifying the channel feature for each types of traffic light and establishing a structure of fusion detectors. Third, we introduce a method of inter-frame information analysis, utilizing detection information of previous frame to modify original proposal regions, which makes the accuracy further improved. In the comparison of other traffic light detection algorithms, our model achieves competitive results on the complex scene VIVA data set. Furthermore, an analysis of small target luminous object detection tasks is given.

Existing System

Aggregate Channel Features (ACF), Deformable Parts Model (DPM)

Proposed System

In this paper, a TLR proposal method is proposed. Our detection model combination prior feature and inter-frame analysis with feature learning algorithms. Through statistics of shapes, location and context of traffic lights, only candidate regions with physical meaning are selected. Meanwhile, the fusion detection model introduces structural features of different types of traffic lights, our specific model for traffic light significantly improves the performance of standard ACF method. While the introduction of adjacent frame information solves the drawbacks when detecting the fake flashing luminous objects by single image frame. Furthermore, to solve the task of detecting small luminous objects in complex scenes, our methods for TLR are also of reference. For a specific object recognition task, it is a good way to establish a model which combines prior feature with statistic learning.

Architecture

BLOCK DIAGRAM