Motion Object Detection and Tracking Optimization in Autonomous Vehicles in Specific Range with Optimized Deep Neural Network

Abstract

Autonomous Vehicles have become increasingly popular around the world in recent years. The potential of this technology is clear and transportation is expected to change dramatically over what is known today. The advantages of Autonomous Vehicles are pollution reduction in urban areas due to improved driving and fuel efficiency to help control traffic flow and parking problems. In addition, Autonomous Vehicles accelerate people and cargo transportation, as well as reducing human errors. There are a variety of issues in the field of Autonomous Vehicles which one of them is the issue of detecting and tracking motion objects as obstacles. In this article, we presented a novel method to optimizing motion objects detection and tracking from the KITTI data set in Autonomous Vehicles in a specific range in between 50 to 80 meters. This approach proposes a real-time and simultaneous structure for motion detection and tracking, so that the data fully enter the combined method called CRF-based Deep Spiking Neural Network with Probabilistic Particle Filter (PPF-DSNN). In fact, CRF-based deep spiking neural network is used to train and test data to extract features and probabilistic particle filtering methods with the aim of detecting and tracking these moving objects. The results represent that proposed approach is highly efficient in comparison to recent methods.