NavneetDalal关于HOG行人识别的博士论文

Navneet DALAL

R´esum´e

Cette th

images et vid

inantes de formes visuelles, qui permettent de d

algorithme de type machine

e enti

notre approche se base sur des descripteurs d’apparence et de mouvement de bas niveau sans

utiliser de mod

trons sur la d

dans applications come l’analyse de film et de vid

´

D’abord, nous pr

statiques : les grilles d’histogrammes d’orientations de gradients d’image (en anglais, HOG –

Histogrammes of Oriented Gradients). Les histogrammes sont

´

spatiaux, avec une forte normalisation locale. Cette structure assure

`

act

sition, d’orientation spatiale, d’illumination locale et de couleur. Nous montrons que la combi-

par rapport aux m

´

descripteurs de mouvement bas

l’approche pr

statique les diff

r

eduit le taux de faux positifs d’un ordre de grandeur par rapport

esentons un travail en cours sur la fac¸on de cr

Abstract

This thesis targets the detection of humans and other object classes in images and videos. Our

dimensional feature vectors that support high accuracy object/non-object decisions. To test our

objects. As a test case we focus on person detection as people are one of the most challenging

object classes with many applications, for example in film and video analysis, pedestrian de-

tection for smart cars and video surveillance. Nevertheless we do not make any strong class

specific assumptions and the resulting object detection framework also gives state-of-the-art

performance for many other classes including cars, motorbikes, cows and sheep.

Histograms of Oriented Gradients (HOG) as descriptors for object detection in static images.

image gradient orientation features extracted at fixed resolution and gathered into a high-

dimensional feature vector. They are designed to be robust to small changes in image contour

locations and directions, and significant changes in image illumination and colour, while re-

maining highly discriminative for overall visual form. We show that unsmoothed gradients,

fine orientation voting, moderately coarse spatial binning, strong normalisation and overlap-

ping blocks are all needed for good performance. Secondly, to detect moving humans in videos,

multiple overlapping detections, that takes into account the number of detections, their confi-

dence scores and the scales of the detections. Lastly, we present work in progress on a parts

based approach to person detection that first detects local body parts like heads, torso, and legs

and then fuses them to create a global overall person detector.