PROJECT TITLE :

Fast Detection of Multiple Objects in Traffic Scenes With a Common Detection Framework

ABSTRACT:

Traffic scene perception (TSP) aims to extract correct real-time on-road surroundings info, which involves 3 phases: detection of objects of interest, recognition of detected objects, and tracking of objects in motion. Since recognition and tracking often rely on the results from detection, the flexibility to detect objects of interest effectively plays a crucial role in TSP. In this paper, we specialize in three important categories of objects: traffic signs, cars, and cyclists. We tend to propose to detect all the three important objects in a very single learning-primarily based detection framework. The proposed framework consists of a dense feature extractor and detectors of 3 vital categories. Once the dense options have been extracted, these features are shared with all detectors. The advantage of using one common framework is that the detection speed is much faster, since all dense features would like solely to be evaluated once in the testing phase. In contrast, most previous works have designed specific detectors using totally different options for each of those three categories. To enhance the feature robustness to noises and image deformations, we introduce spatially pooled options as a half of aggregated channel features. In order to additional improve the generalization performance, we propose an object subcategorization methodology as a means of capturing the intraclass variation of objects. We have a tendency to experimentally demonstrate the effectiveness and potency of the proposed framework in 3 detection applications: traffic sign detection, automobile detection, and cyclist detection. The proposed framework achieves the competitive performance with state-of-the-art approaches on many benchmark knowledge sets.