Predict Vehicle Collision by TTC From Motion Using a Single Video Camera

Abstract

The objective of this work is the instantaneous computation of Time-to-Collision (T T C) for potential collision only from the motion information captured with a vehicle borne camera. The contribution is the detection of dangerous events and degree directly from motion divergence in the driving video, which is also a clue used by human drivers. Both horizontal and vertical motion divergence are analyzed simultaneously in several collision sensitive zones. The video data are condensed to the motion profiles both horizontally and vertically in the lower half of the video to show motion trajectories directly as edge traces. Stable motion traces of linear feature components are obtained through filtering in the motion profiles. As a result, this avoids object recognition and sophisticated depth sensing in prior. The fine velocity computation yields reasonable T T C accuracy so that the video camera can achieve collision avoidance alone from the size changes of visual patterns. We have tested the algorithm for various roads, environments, and traffic, and shown results by visualization in the motion profiles for overall evaluation.M. Kilicarslan was with the

Predicting Hazardous Driving Events Using Multi-Modal Deep Learning Based on Video Motion Profile and Kinematics Data

“…Driving videos, usually processed with computer vision techniques, provide environment factors during a vehicle trip. Works in [7,8,9] estimate TTC from motion in driving videos without applying vehicle recognition and depth measuring in prior. Computer vision technique has limited performance in recognizing far road traffic due to a low resolution, and a cut in of other vehicles in complex traffic scenes due to the difficulty in explicit modeling of dynamic environments.…”

“…To bridge the video signal to the classification of hazardous events and avoid influence caused by complex traffic environment, a data representation that reflects the motion trajectory information more than one video frame is implemented, since the divergence of trajectories can be linked to TTC [9]. Temporal driving video is converted to a spatial-temporal map so that the time, distance, position, and speed of surrounding scenes can be included.…”

“…Another merit of it is the data reduction for both bigvideo learning and on-line real-time event detection during driving. We employ the motion profile [7,8,9] to record the motion of surrounding traffic.…”

“…The earlier time happening (trace) in the far motion profile (b channel) may be less critical in causing a crash, while any happening in the close motion profile (r channel) may cause danger immediately. The approach of a vehicle has its size expanding in the profile, i.e., its trajectory diverges [9]. A constantly approaching vehicle from far to close corresponds to a transition of trajectory from high motion profile (b channel) to low motion profile (r channel), which is a serious case that requires precaution.…”

A Forward Collision Warning System Using Driving Intention Recognition of the Front Vehicle and V2V Communication

“…Some of these FCW systems were based on collision warning models with fixed parameters. Kilicarslan and Zheng [8] used the timeto-collision (TTC) model with a fixed threshold and a single camera to evaluate the potential collision danger. Nagatani [9] proposed a time-headway (THW) model, which not only considered the speed of the following vehicle, but also the distance between the two cars.…”