Integrating appearance and edge features for on-road bicycle and motorcycle detection in the nighttime

Abstract

It is critical to detect bicycles and motorcycles on the road because collision of autos with those light vehicles becomes major cause of on-road accidents nowadays especially in the nighttime. Therefore, a vision-based nighttime bicycle and motorcycle detection method relying on use of a camera and near-infrared lighting mounted on an auto vehicle is pro posed in this paper. Generally, the foreground objects in front of the auto, not the far-away background, will reflect near-infrared lighting in the nighttime environments. However, some components of the bicycles and the motorcycles absorb most infrared lighting and thus make the bicycles and motor cycles hardly recognizable. To cope with this problem, the aforementioned detection method is part-based, which com bines the two kinds of features related to the characteristics of bicycles and motorcycles. Also, the information about the geo metric relation among all the parts and the object centroid is learned off-line. Due to high computation load, Adaboost algo rithm is used to select effective parts with better geometric in formation for detection. To validate the proposed results, sev eral experiments are conducted to show that the developed system is reliable in detecting bicycles and motorcycles in the nighttime.

AutoBiVision - A Novel Vision-Based Bicycle Warning System

“…The vision-based bicycle detection has been a hot research topic in last decade. The feature of the bicycles can be employed in the detection procedure, e.g., ellipse approximation [2], edge features of the bicycle [3], and deformable part model [4]. Alternatively some approaches employ the features of the cyclists, such as MSC-HOG feature [5] and RealAdaBoost [6].…”

Intelligent co-detection of cyclists and motorcyclists based on an improved deep learning method

“…In order to identify bicycles and the motorcycles, Messelodi et al [19] firstly divided the bicycle and motorcycle images into a side view group and a front view/back view group; and then, a set of features was extracted from the images to classify the bicycles and motorcycles by a support vector machine (SVM). Han et al [20] used part models to describe the bicycles and motorcycles based on the appearance-based features and edgebased features, and then a SVM classifier was used to distinguish them. Wu et al [21] represented the pedestrians, bicycles, motorcycles and other vehicles using multiple types of objects, including the image pixel and HOG features; then a deep confidence network was used to classify the ARUs.…”