Abstract
Pedestrian fatalities and injuries most likely occur in vehicle-pedestrian crashes. Meanwhile, engineers have tried to reduce the problems by developing a pedestrian detection function in Advanced Driver-Assistance Systems (ADAS) and autonomous vehicles. However, the system is still not perfect. A remaining problem in pedestrian detection is the performance reduction at nighttime, although pedestrian detection should work well regardless of lighting conditions. This study presents an evaluation of pedestrian detection performance in different lighting conditions, then proposes to adopt multispectral image and deep neural network to improve the detection accuracy. In the evaluation, different image sources including RGB, thermal, and multispectral format are compared for the performance of the pedestrian detection. In addition, the optimizations of the architecture of the deep neural network are performed to achieve high accuracy and short processing time in the pedestrian detection task. The result implies that using multispectral images is the best solution for pedestrian detection at different lighting conditions. The proposed deep neural network accomplishes a 6.9% improvement in pedestrian detection accuracy compared to the baseline method. Moreover, the optimization for processing time indicates that it is possible to reduce 22.76% processing time by only sacrificing 2% detection accuracy.
Highlights
Several deep neural network (DNN) algorithms have been tested for applying fusion datasets, such as Single Shot Detector (SSD) [34], Region Proposal Network (RPN) combined with Boosted Decision Trees [35], and the combination of convolutional neural network (CNN) and Support Vector Networks (SVR) [36]
The first contribution shows that the multispectral image can be the best solution to solve pedestrian detection problems, especially in lowlight conditions
The second contribution of this research is proposing an optimization of You Look Only Once (YOLO) v3 for improving the detection accuracy
Summary
Pedestrian fatalities and injuries are significant issues in modern traffic systems. In the United States 16% of the total traffic fatalities recorded in 2017 were pedestrians [1]. A system called Advanced Driving Assistance System (ADAS) has been developed for car users to prevent unexpected accidents from happening. This system is equipped with many features to support the safety of passengers, drivers, and the surroundings. Four scenarios were examined with no ambient light conditions and when only a low-beam car light was used for lightening the environment In those experiments, four different cars did not decelerate correctly in any of the four tests. Four different cars did not decelerate correctly in any of the four tests As a result, this indicates that pedestrian detection does not behave as expected under nighttime conditions, conceptually it should perform well under any lighting conditions
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have