A new benchmark for stereo-based pedestrian detection

Abstract

Abstract-Pedestrian detection is a rapidly evolving area in the intelligent vehicles domain. Stereo vision is an attractive sensor for this purpose. But unlike for monocular vision, there are no realistic, large scale benchmarks available for stereobased pedestrian detection, to provide a common point of reference for evaluation. This paper introduces the Daimler Stereo-Vision Pedestrian Detection benchmark, which consists of several thousands of pedestrians in the training set, and a 27-min test drive through urban environment and associated vehicle data. The data, including ground truth, is made publicly available for non-commercial purposes. The paper furthermore quantifies the benefit of stereo vision for ROI generation and localization; at equal detection rates, false positives are reduced by a factor of 4-5 with stereo over mono, using the same HOG/linSVM classification component.

Generación de Regiones con Potencial de Contener Peatones usando Reconstrucción 3D No Densa a partir de Visión Monocular

“…Tanto es así que se han implementado propuestas ajustadas a las distintas condiciones de iluminación, día y noche, usando visión monocular (2D) y/o visión estéreo (3D) (Zhao et al, 2009;Yuan et al, 2015;Kohler et al, 2015;Min et al, 2013;Tetik and Bolat, 2011;Keller et al, 2011;Mesmakhosroshahi et al, 2014;Wang et al, 2014;Flores-Calero et al, 2015).…”

“…En los próximos tres trabajos la base de datos Daimler pedestrian benchmark (Keller et al, 2011) ha sido empleada para generar los resultados experimentales, así, Min et al (Min et al, 2013) han obtenido una tasa de detección del 73 % a 10 −4 FPPW, Mesmakhosroshahi et al (Mesmakhosroshahi et al, 2014) han alcanzado una tasa de detección del 98,8 % y Zhang et al (Zhang et al, 2015a) han complementado sus experimentos con imágenes capturadas alrededor de la ciudad de Chicago Tetik y Bolat (Tetik and Bolat, 2011) han desarrollado sus experimentos sobre las bases de datos Nicta (Overett et al, 2008) para entrenamiento, y Penn Fudan (Wang et al, 2007) para validación. Esta propuesta alcanza una tasa de detección del 84,4 %.…”

“…Esto convierte a un sistema de detección de peatones (SDP) en una alternativa necesaria para evitar accidentes graves y fatalidades (Yuan et al, 2015;Kohler et al, 2015;Min et al, 2013;Tetik and Bolat, 2011;Keller et al, 2011;Wang et al, 2014).…”

The Benefits of Dense Stereo for Pedestrian Detection

“…Most benchmark studies have dealt with monocular pedestrian detection. Recently, Keller et al [22] have introduced a large publicly available stereo-based pedestrian data set, involving a 27-min test drive through urban environment, including vehicle data. In terms of methods, previous work has mostly followed a module-based strategy comprising generation of possible pedestrian locations (ROIs), followed by pedestrian classification and tracking.…”

Preceding Vehicle Detection and Tracking with Motion Estimation by Radar-vision Sensor Fusion

“…자차의 움직임은 그림 2(b)와 같이 모델링 [8] 하고 자 차의 종방향 움직임을 보상하여 주행 중인 차량과 고정 과 제안하는 방법의 성능 비교 Table 3. Performance comparison between the proposed method and Liu et al method [5] .…”