Enhancing the Prediction Techniques for Night Vision Environment in Vehicle Automation

Abstract

An autonomous car is a vehicle capable of sensing its environment and operating without human involvement. Autonomous vehicle is the exponential growing topic in the research filed due its huge demand in the current industry where object detection and the distance estimation play the major role to take any major decision. Object detection in the normal vision yields the excellent performance due to deep learning models but object detection in the night vision does not yield the good performance due to various challenges including 1. Due to uneven intensity in the night vision images foreground and background treated as the same which leads to segmentation issues. 2. More information like vehicle, road, traffic light and person will be lost due to poor lighting. 3 No accurate object is detected due to motion blurring, blurred images or partial occlusion. 4 Night images do not receive the sufficient light due to which there will be more noise resulting in the failure of the multi object detection. Extracting or detecting object information in the night image is not possible without applying the most feasible pre processing technique. By going through all these challenges, proposing the three main objectives for working in the night vision environment for autonomous vehicle. First objective aims to improving the night image quality by applying the novel GHE -Ensemble model which improves the image quality, reduces noise and uniform distribution of pixel throughout the image. Second objective focuses on the multi object detection in the night vision where the proposed model EnPreNiNet resulted in the better accuracy precision than the baseline methods. Third objective focused on the distance estimation between the two-vehicle detected using DisNet model which resulted the error rate 0-3 Meters in estimating the distance when compared to the existing models for the distance estimation.