Multi-cue Pedestrian Detection and Tracking from a Moving Vehicle

Abstract

Abstract. This paper presents a multi-cue vision system for the real-time detection and tracking of pedestrians from a moving vehicle. The detection component involves a cascade of modules, each utilizing complementary visual criteria to successively narrow down the image search space, balancing robustness and efficiency considerations. Novel is the tight integration of the consecutive modules: (sparse) stereo-based ROI generation, shape-based detection, texture-based classification and (dense) stereo-based verification. For example, shape-based detection activates a weighted combination of texture-based classifiers, each attuned to a particular body pose.Performance of individual modules and their interaction is analyzed by means of Receiver Operator Characteristics (ROCs). A sequential optimization technique allows the successive combination of individual ROCs, providing optimized system parameter settings in a systematic fashion, avoiding ad-hoc parameter tuning. Application-dependent processing constraints can be incorporated in the optimization procedure.Results from extensive field tests in difficult urban traffic conditions suggest system performance is at the leading edge.

Dense Stereo-Based ROI Generation for Pedestrian Detection

“…Significant speed-ups can be obtained by including application-specific constraints such as flat-world assumption, ground-plane based objects and common geometry of pedestrians, e.g. object height or aspect ratio [9,17].…”

“…[9,1]) because of lower processing cost. However, with recent hardware advances, real-time dense stereo has become feasible [12] (here we use a hardware implementation of the semi-global matching (SGM) algorithm [7]).…”

“…The aim of this paper thus is to investigate the advantages of dense vs. sparse disparity maps when detecting generic obstacles in the early stage of a pedestrian detection system [9]. We are interested in both the ROC performance (trade-off correct vs. false detections) and in the processing cost.…”

Pedestrian Detection by Probabilistic Component Assembly

“…In place of the constant histogram selection [2], Zhu et al [19] use a variable selection made by an AdaBoost cascade, which achieves better results. Gavrila and Munder [7] recognize pedestrians with local receptive fields and several neural networks.…”

Tracking and intercepting pedestrians: a robotic approach to surveillance of critical infrastructures

“…This histogram is sent to a linear Support Vector Machine (SVM) classifier, that performs the final detection. A different approach is called Shape-based methods, they use template matching algorithms in order to detect the appearance of human silhouettes on the image [18]. Yet another approach is the part-based representations.…”