Human detection for autonomous vehicles

In today´s world the need for autonomous vehicles is growing as there is a need for having greater reliability and much faster reaction time compared to human drivers. This paper proposes vision based solution for autonomous vehicles for detecting humans, obstacles to avoid collisions. This solution facilitates timely halting of an autonomous vehicle in case there are obstacles in the vicinity of the vehicle. This algorithm supports detection of 16 different human postures in a radius of 5 meters from the vehicle. Vision based system overcomes the range limitations as observed while using proximity sensors. This algorithm would use HOG (Histogram Oriented Gradients) features for classification of objects as human or obstacles. To reduce errors, thermal image analysis can be used for detection. Multiple algorithms can run in parallel on the same image to increase confidence factor of detection. This will facilitate human/obstacle detection in day as well as at night time. It also makes it independent of environmental conditions such as dust or fog.