Performance Evaluation of Multi-Target Object Trackers in IOT based autonomous intelligent vehicles

The crucial challenge in autonomous intelligent vehicles is multi-object tracking, classifying, and locating in a variety of traffic conditions. Car makers and academic researchers are tackling this problem using a variety of sensors, including camcorders, light detection and ranging, interferometric synthetic aperture radars, infrared sensors, global positioning systems, and automotive technologies. Deep learning networks are playing a key role in resolving this challenge. In this paper some tracking algorithms have been proposed. The prototype-tracker based on Monte-Carlo technique utilizes a rectangular shaped target to assess all detections compared to the predictable track state. This assists the tracker in determining the shape and angle more precisely. However, the procedure of ray-intersections, linked with sequential Monte Carlo methods are computationally more complex and involves several calculations leading to delay in taking decisions timely.