Campus Event Calendar
Event Entry
What and Who
Pedestrian Detection and Online Tracking from Monocular Images
Yang He
Beijing Institute of Technology, China
PhD Application Talk
Master's student at Beijing Institute of Technology, China
AG 1, AG 2, AG 3, AG 4, AG 5, SWS, RG1, MMCI
Public Audience
English
Note: We use this to send email in the morning.
Date, Time and Location
Monday, 23 February 2015
09:00
120 Minutes
E1 4
R024
Saarbrücken
Abstract
Pedestrian detection and online tracking are key techniques in intelligent transportation system. A successful pedestrian detection and tracking system depends on good features for detection as well as a reliable tracking method. Firstly, the appearance of pedestrians has large variations and the background is always complex, increasing the difficulty of detecting a pedestrian. Secondly, it is easy to cause drift problem because of occlusions and inaccurate tracking results. For the above problems, this study intends to obtain a powerful feature for pedestrian detection without considering the depth information, and an effective template update strategy to alleviate the drift problem during tracking.
Recently, benefitting from the prowess of feature learning, numerous computer vision tasks have improved a lot. Besides, channel features are very useful for pedestrian detection, however, channel features are a kind of low-level features, which are lack of semantic information. In this work, we apply feature learning on channel features to obtain an effective mid-level feature, and combine the learnt feature with channel features to detect a pedestrian.
In the aspect of tracking, an alternate template update strategy is proposed to boost an online tracker by alleviating the drift problem. The goal of this strategy is to develop a robust way of updating an adaptive appearance model. There are several modules in the appearance model, and the alternate template update strategy is the case that only one module is updated with newly added data and the others keep stable. It makes proposed tracker be updated while keeping historical information. When occlusions or inaccurate tracking results happen, they affect only one module in the tracker. Therefore, the proposed method can track the object by other modules. What’s more, in order to fuse these modules, this paper desired a criterion with considering the appearance similarity and motion information between two consecutive frames.
Lastly, inspired by the effectiveness of Convolutional Neural Networks (ConvNets) with random filters in image classification task, a visual tracker with simple random filter-based ConvNets is proposed. We generate a group of simple rectangle filters randomly to build a two-stage ConvNets model, and apply the model to extract features. With the two-stage feature, we train an online Naive Bayes classifier to build the appearance representation. According to a serious of experiments on challenging image sequences, it is shown that the proposed tracker achieves considerable results.
Recently, benefitting from the prowess of feature learning, numerous computer vision tasks have improved a lot. Besides, channel features are very useful for pedestrian detection, however, channel features are a kind of low-level features, which are lack of semantic information. In this work, we apply feature learning on channel features to obtain an effective mid-level feature, and combine the learnt feature with channel features to detect a pedestrian.
In the aspect of tracking, an alternate template update strategy is proposed to boost an online tracker by alleviating the drift problem. The goal of this strategy is to develop a robust way of updating an adaptive appearance model. There are several modules in the appearance model, and the alternate template update strategy is the case that only one module is updated with newly added data and the others keep stable. It makes proposed tracker be updated while keeping historical information. When occlusions or inaccurate tracking results happen, they affect only one module in the tracker. Therefore, the proposed method can track the object by other modules. What’s more, in order to fuse these modules, this paper desired a criterion with considering the appearance similarity and motion information between two consecutive frames.
Lastly, inspired by the effectiveness of Convolutional Neural Networks (ConvNets) with random filters in image classification task, a visual tracker with simple random filter-based ConvNets is proposed. We generate a group of simple rectangle filters randomly to build a two-stage ConvNets model, and apply the model to extract features. With the two-stage feature, we train an online Naive Bayes classifier to build the appearance representation. According to a serious of experiments on challenging image sequences, it is shown that the proposed tracker achieves considerable results.
Contact
IMPRS-CS Office
0681 9325 1800
--email hidden
passcode not visible
logged in users only
Tags, Category, Keywords and additional notes
Stephanie Jörg, 02/20/2015 08:57 -- Created document.