Search results

(1 - 2 of 2)

Title

PEDESTRIAN DETECTION AND TRACKING FOR ADVANCED DRIVER ASSISTANCE SYSTEMS

Creator

Mesmakhosroshahi, Maral

Date

2017, 2017-05

Description

In an effort to reduce driver errors in being the major cause of traffic accidents, there is a lot of research being conducted into the...
Show moreIn an effort to reduce driver errors in being the major cause of traffic accidents, there is a lot of research being conducted into the development of advanced driver assistance systems (ADAS). ADAS is a system aimed at helping the driver in tasks such as pedestrian and vehicle detection, traffic sign recognition and lane detection. Pedestrian detection is one of the major tasks in advanced driver assistance systems (ADAS). Most of the stereo based pedestrian detection algorithms include three major steps: 1. Ground plane estimation 2. Region of interest (ROI) generation 3. Pedestrian classification In this thesis, we present a stereo-based pedestrian detection framework for advanced driver assistance systems by exploiting both color and depth information obtained from a stereo camera installed in a vehicle. In our proposed framework, we first use the vertical gradient of the dense depth map to estimate and discard the ground plane. The boundaries of the ground plane are then searched to detect the pedestrians and the depth values of the boundaries are used to compute the size of the detection windows for detecting pedestrians at different scales. In addition, a depth-based multi-scale ROI extraction method has been proposed to reduce the computation time of ROI extraction. For classifying ROIs to pedestrian and non-pedestrian classes, Histogram of Oriented Gradients (HOG)/Linear support vector machine (SVM) and Integral Channel Features (ICF)/Adaboost are used. To recover the missed pedestrians and improve the detection rate, an ROI tracking algorithm is proposed which incorporates the ROIs extracted from the current frame with theROIs tracked from a reference frame. For additional reduction in search space, we propose a novel algorithm to reduce the number of candidate windows extracted as ROI by taking advantage of the temporal correlation between the adjacent frames. We also propose a method to improve the accuracy of the pedestrian classifi- cation using the aggregated channel features. In this approach, we use the aggregated channel features as our baseline detector and improve it's accuracy using the tanh normalization and Gabor filter. After classification using Adaboost, we use a posi- tive subset of the bounding boxes to classify them again using Convlutional Neural Network to finalize the detection.
Ph.D. in Electrical Engineering, May 2017
Show less

Title

Exploiting contextual information for deep learning based object detection

Creator

Zhang, Chen

Date

2020

Description

Object detection has long been an important research topic in computer vision area. It forms the basis of many applications. Despite the great...
Show moreObject detection has long been an important research topic in computer vision area. It forms the basis of many applications. Despite the great progress made in recent years, object detection is still a challenging task. One of the keys to improving the performance of object detection is to utilize the contextual information from the image itself or from a video sequence. Contextual information is defined as the interrelated condition in which something exists or occurs. In object detection, such interrelated condition can be related background/surroundings, support from image segmentation task, and the existence of the object in the temporal domain for video-based object detection. In this thesis, we propose multiple methods to exploit contextual information to improve the performance of object detection from images and videos.First, we focus on exploiting spatial contextual information in still-image based object detection, where each image is treated independently. Our research focuses on extracting contextual information using different approaches, which includes recurrent convolutional layer with feature concatenation (RCL-FC), 3-D recurrent neural networks (3-D RNN), and location-aware deformable convolution. Second, we focus on exploiting pixel-level contextual information from a related computer vision task, namely image segmentation. Our research focuses on applying a weakly-supervised auxiliary multi-label segmentation network to improve the performance of object detection without increasing the inference time. Finally, we focus on video object detection, where the temporal contextual information between video frames are exploited. Our first research involves modeling short-term temporal contextual information using optical flow and modeling long-term temporal contextual information using convLSTM. Another research focuses on creating a two-path convLSTM pyramid to handle multi-scale temporal contextual information for dealing with the change in object's scale. Our last work is the event-aware convLSTM that forces convLSTM to learn about the event that causes the performance to drop in a video sequence.
Show less