Pedestrian detection for driving assistance systems: single-frame classification and system level performance

Shashua, Amnon; Gdalyahu, Yoram; Hayun, G.

doi:10.1109/ivs.2004.1336346

Cited by 265 publications

(183 citation statements)

References 16 publications

Supporting

Mentioning

174

Contrasting

Unclassified

Order By: Relevance

“…Shashua et al (2004), for instance, extract a feature vector from each of 9 fixed sub-regions. Other approaches attempt to directly identify certain body parts.…”

Section: Previous Workmentioning

confidence: 99%

Multi-cue Pedestrian Detection and Tracking from a Moving Vehicle

Gavrila

Munder

2006

Int J Comput Vision

467

328

View full text Add to dashboard Cite

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.

show abstract

“…Shashua et al (2004), for instance, extract a feature vector from each of 9 fixed sub-regions. Other approaches attempt to directly identify certain body parts.…”

Section: Previous Workmentioning

confidence: 99%

Multi-cue Pedestrian Detection and Tracking from a Moving Vehicle

Gavrila

Munder

2006

Int J Comput Vision

467

328

View full text Add to dashboard Cite

show abstract

“…Mohan et al (2001) divide human body into four parts: head-shoulder, legs, left arm, and right arm. Shashua et al (2004) divide human body into nine overlapping sub-regions. Some of the sub-regions correspond to natural human body parts, such as head, torso, arms, and legs; some do not.…”

Section: Detection Of Multiple Occluded Objectsmentioning

confidence: 99%

“…After obtaining the part detection responses, some early methods (e.g. Mohan et al 2001;Shashua et al 2004;Mikolajczyk et al 2004) do the part combination independently for each human. The combination is usually based on majority voting (e.g.…”

Section: Detection Of Multiple Occluded Objectsmentioning

confidence: 99%

Detection and Segmentation of Multiple, Partially Occluded Objects by Grouping, Merging, Assigning Part Detection Responses

Nevatia

2008

Int J Comput Vis

135

View full text Add to dashboard Cite

We propose a method that detects and segments multiple, partially occluded objects in images. A part hierarchy is defined for the object class. Both the segmentation and detection tasks are formulated as binary classification problem. A whole-object segmentor and several part detectors are learned by boosting local shape feature based weak classifiers. Given a new image, the part detectors are applied to obtain a number of part responses. All the edge pixels in the image that positively contribute to the part responses are extracted. A joint likelihood of multiple objects is defined based on the part detection responses and the object edges. Computation of the joint likelihood includes an inter-object occlusion reasoning that is based on the object silhouettes extracted with the whole-object segmentor. By maximizing the joint likelihood, part detection responses are grouped, merged, and assigned to multiple object hypotheses. The proposed approach is demonstrated with the class of pedestrians. The experimental results show that our method outperforms the previous ones.

show abstract

“…Typical accuracy values for recently-published detection systems are still far from optimal (percent hit / percent false alarm): 75-90% / 100% (Nanda and Davis, 2002), 85% / 3% (Zhao and Thorpe, 1999), 84% / 19% and 92% / 3% per frame (Fang et al, 2003), 93% / 5% (Xu, Liu, and Fujimura, 2005), 70% / 20% (Bertozzi et al, 2004), and 85-93% / 0.01-0.1% (Shashua, Gdalyahu, and Hayun, 2004). (Methods of calculation differ among experiments.…”

Section: Pedestrian Detection Systemsmentioning

confidence: 99%

Pedestrian detection with night vision systems enhanced by automatic warnings

Tsimhoni¹,

Flannagan²,

Minoda³

2006

PsycEXTRA Dataset

View full text Add to dashboard Cite

This experiment compared pedestrian detection using far-infrared (FIR) and near-infrared (NIR) night vision systems, combined with automatic warnings at one of two distances or no warning at all. Sixteen subjects (eight younger than 30 years and eight older than 60 years) pressed a button as soon as they saw a pedestrian on a night vision system in the center console of a vehicle simulator. In addition, they performed a concurrent simulated steering task that required almost continuous viewing of the forward scene, similar to real driving.As in a previous experiment (Tsimhoni, Bärgman, Minoda, and Flannagan, 2004), detection distances with FIR systems were substantially greater than with NIR systems. Detection distances with both systems were shorter than in the previous experiment by about 20 m, probably because of the addition of simulated steering in the present experiment. The automatic visual warning was a blue rectangle that zoomed in on the pedestrian in the video display. In the long-distance condition, it was presented when the pedestrian was 150 m away. Detection distance and accuracy for both night vision systems increased, but the effects were more prominent for the NIR system. Automatic warnings at 75 m improved performance with NIR but worsened performance with FIR, perhaps because in some trials subjects waited for the automatic warning before responding. Subjective ratings of mental workload and of effort were higher for NIR than for FIR, but the addition of automatic warnings did not decrease perceived workload significantly. Overall, automatic visual warnings based on image processing were effective in increasing accuracy and detection distance for pedestrians except when short-distance warnings were used with the FIR system.17.

show abstract

Pedestrian detection for driving assistance systems: single-frame classification and system level performance

Cited by 265 publications

References 16 publications

Multi-cue Pedestrian Detection and Tracking from a Moving Vehicle

Multi-cue Pedestrian Detection and Tracking from a Moving Vehicle

Detection and Segmentation of Multiple, Partially Occluded Objects by Grouping, Merging, Assigning Part Detection Responses

Pedestrian detection with night vision systems enhanced by automatic warnings

Contact Info

Product

Resources

About