Preventing the False Negatives of Vehicle Object Detection in Autonomous Driving Control Using Clear Object Filter Technique

Autonomous vehicle also known as “Self-Driving” vehicle uses handful of algorithms and decision-making controls that can operate vehicle with little to no effort from human involvement. The key aspects of autonomous driving controls and/or any driver assist control systems shall provide utmost comfort for driver as well making decisions which are safer when driver presence on reactions counted as negligible. This research shall talk about driver assist feature in assisting driving controller of “Ego” vehicle to clearly detect the presence of “Leader” vehicle object when in lane. Vehicle object detections are key aspect in any vehicle that occupied with driver assist feature whose decision of control shall depend on image detections that can happen positively. This research shall also address most common problems in image detection technique that can occur when a vehicle object detection notice is sent to driving controller “Negatively” also called as “not detecting a vehicle even the presence of object exists” or also commonly noted as “False Negatives” of “Leader” Vehicle Detection by “Ego” Vehicle. “False Negative” detections more unsafe and can cause hazardous events such as front-end collision of “Ego” or “Follower” vehicle and risk of involving in accidents. This research uses low-cost computer vision-based vehicle object image detection and noise cancellation of “Leader” vehicle object using a sliding-window method of inducing “Median” filter to detect corrupted “Leader” vehicle object followed by deep-learning neural network of machine learning.