Multi-Pedestrian Tracking with GM-PHD Filters in an embedded Heterogeneous Parallel Processing Platform with Sensor Fusion

Real-Time Pedestrian detection and tracking methods are the state-of-the-art techniques in present driver assistance systems. However, pedestrian detection and tracking methods that exploit the parallel processing capabilities of heterogeneous high performance computing devices such as FPGAs (or GPUs) with sensor fusion(camera and Lidar), a technology that potentially will replace ECUs in a coming generation of cars, are a rare subject of interest. In this research a pedestrian detection and tracking algorithm is developed and implemented, especially designed to incorporate one or many, and even heterogeneous, hardware accelerators in the first phase. In the second phase it will incorporate Lider and use the data fusion technique to gain better accuracy and precision in real-time. Pedestrian detection is done using Histogram of Oriented Gradients (HOG) for human detection. Parallel implementation of HOG people detection had given a very good real-time performance and robustness. For tracking pedestrian we have used Guassian Mixture of Probability Hypothesis Density filter, which is very suitable for sensor fusion because of its output set based. The code of the parallel implementation of HOG and sequential implementation is tested on 2 GPU – the NVIDIA GeForce 940 M and NVIDIA GeForce GTX 660 TI. Tests on GPU show a significant improvement in performance and memory consumption. Detection performance is improved by 10x to 15x on an average by these above mentioned GPU. The algorithm is tested with different benchmark dataset and performed very well with speed and accuracy. Sequential implementation of GMPHD filter performance heavily decreased with increasing number of target. Parallel implementation of some of the block of GMPHD filter achieved 10x performance speedup.