A robust pedestrian detection approach based on shapelet feature and Haar detector ensembles

Automated Detection of Hands and Objects in Egocentric Videos, for Ambient Assisted Living Applications

“…We evaluated the performance of our object detection models using the mean Average Precision (mAP) metric from the PASCAL VOC 2007 detection benchmarks [15,16]. This evaluation measure is algorithm-independent, unlike for example the Detection Error Trade-off used for evaluating pedestrian detectors [17,18] which is only suitable for sliding window methods. Furthermore, a comparison of the mAP measure and the "area under curve" (AUC) measure [19] on PASCAL VOC2006 showed that the mAP measure highlighted differences between methods to a greater extent [16].…”

Pedestrian Detection Technique’s – A Review

“…There is a vast study is done by different researcher on pedestrian detection using various techniques.Like S.Wang,Li Zhu et al [15] proposed a multiscale handling method for the fast pedestrian detection in which the tactics detection from sparse to dense is used.Where as Pawan Sinha,Tomaso A.Poggio et al [14]used wavelet templates for pedestrian detection.In which the wavelet template interpret the shape of an object entitled as a subset of the wavelet coefficient of the image.Therefore they proves that the invariant properties and computational efficiency of the wavelet template becomes an successful tool for object detection.A.Broggi,M.Bertozzi et at. [12]proposed a pedestrian detection system based on shape,in which a vision based algorithms used.This algorithm is rooted on the localization of human shape,based on symmetry,size,ratio and shape.From the combination of shapelet feature and Haar detector,Wentao Yao et al [13] gives another approach.In this double stage algorithm, using shapelet feature the non pedestrian objects are removed so at second stage only some objects are need to be recognize. On the contrary Bastian Leibe,Edgar Seemann et al [6] presents a series of iterative evidence aggregation steps for a pedestrian detection in a crowded scene ,where they combine local and global cues by means of an automatically computed top down segmentation.…”

A new algorithm for the establishing data association between a camera and a 2-D LIDAR

“…They have large field of view and provide an accurate 3-D model of the environment. However, the color and texture information, which is very important for object detection and recognition [1][2][3][4][5] , cannot be acquired from them. On the other hand, cameras capture purpose of extrinsic calibration is to obtain these parameters.…”