In this paper we present an algorithm for path planning in a fixed range-only beacon field. We define and calculate entropy values for regions of interest and provide a method for finding "safe," low-entropy...
详细信息
The University of Toronto is one of eight teams competing in the SAE AutoDrive Challenge a competition to develop a self-driving car by 2020. After placing first at the Year 1 challenge [1], we are headed to MCity in ...
详细信息
ISBN:
(纸本)9781728118383
The University of Toronto is one of eight teams competing in the SAE AutoDrive Challenge a competition to develop a self-driving car by 2020. After placing first at the Year 1 challenge [1], we are headed to MCity in June 2019 for the second challenge. There, we will interact with pedestrians, cyclists, and cars. For safe operation, it is critical to have an accurate estimate of the position of all objects surrounding the vehicle. The contributions of this work are twoli)ld: First, we present a new object detection and tracking dataset (UoITPed50), which uses GPS to ground truth the position and velocity ()I' a pedestrian. To our knowledge, a dataset this type for pedestrians has not been shown in the literature before. Second, we present a lightweight object detection and tracking system (aUToTrack) that uses vision, LIDAR, and GPS/IMU positioning to achieve state-of-theart performance on the KITTI Object Tracking benchmark. We show that aUToTrack accurately estimates the position and velocity of pedestrians, in real-time, using CPUs only. aUToTrack has been tested in closed-loop experiments on a real self-driving car (seen in Figure 1), and we demonstrate its performance on our dataset.
Road quality assessment is a crucial part in municipalities' work to maintain their infrastructure, plan upgrades, and manage their budgets. Properly maintaining this infrastructure relies heavily on consistently ...
详细信息
ISBN:
(纸本)9781538628188
Road quality assessment is a crucial part in municipalities' work to maintain their infrastructure, plan upgrades, and manage their budgets. Properly maintaining this infrastructure relies heavily on consistently monitoring its condition and deterioration over time. This can be a challenge, especially in larger towns and cities where there is a lot of city property to keep an eye on. We review road quality assessment methods currently employed, and then describe our novel system, which integrates a collection of existing algorithms, aimed at identifying distressed road regions from street view images and pinpointing cracks within them. We predict distressed regions by computing Fisher vectors on local SIFT descriptors and classifying them with an SVM trained to distinguish between road qualities. We follow this step with a comparison to a weighed contour map within these distressed regions to identify exact crack and defect locations, and use the contour weights to predict the crack severity. Promising results are obtained on our manually annotated dataset, which indicate the viability of using this cost-effective system to perform road quality assessment at the municipal level.
Fire detection using video offers a novel way of detecting fire in spaces where conventional smoke-based fire detectors tend to exhibit high false alarm behavior. This paper explores a simple algorithm for flame detec...
详细信息
We describe the development and deployment of a system for long-distance remote observation of robotic operations. The system we have developed is targeted to exploration, multi-participant interaction, and tele-learn...
详细信息
Realistic rendering in real-time augmented reality applications leads one to consider physical interactions between real and virtual worlds. One of these interactions is mutual occlusions in the rendered viewpoint. Th...
详细信息
The processing of colour images to improve sharpness is nearly always been realized in RGB colour space. This paper explores the effects of using different colour spaces on the application of image sharpening algorith...
详细信息
Model initialisation is an important component of object tracking. Tracking algorithms are generally provided with the first frame of a sequence and a bounding box (BB) indicating the location of the object. This BB m...
详细信息
ISBN:
(纸本)9781538664810
Model initialisation is an important component of object tracking. Tracking algorithms are generally provided with the first frame of a sequence and a bounding box (BB) indicating the location of the object. This BB may contain a large number of background pixels in addition to the object and can lead to parts-based tracking algorithms initialising their object models in background regions of the BB. In this paper, we tackle this as a missing labels problem, marking pixels sufficiently away from the BB as belonging to the background and learning the labels of the unknown pixels. Three techniques, One-Class SVM (OC-SVM), Sampled-Based Background Model (SBBM) (a novel background model based on pixel samples), and Learning Based Digital Matting (LBDM), are adapted to the problem. These are evaluated with leave-one-video-out cross-validation on the VOT2016 tracking benchmark. Our evaluation shows both OC-SVMs and SBBM are capable of providing a good level of segmentation accuracy but are too parameter-dependent to be used in real-world scenarios. We show that LBDM achieves significantly increased performance with parameters selected by cross validation and we show that it is robust to parameter variation.
The Nomad 200 and the Nomad SuperScouts are among the most popular platforms used, for research in robotics. Built in the early 1990's they were the base of choice for many mobile robotics researchers. Unfortunate...
详细信息
As vision plays a central role in the operation of autonomous cars, one key challenge is that the limited dynamic range of camera sensors can only capture a certain portion of the scene radiance. This can lead to loss...
详细信息
ISBN:
(纸本)9781728198910
As vision plays a central role in the operation of autonomous cars, one key challenge is that the limited dynamic range of camera sensors can only capture a certain portion of the scene radiance. This can lead to loss of information from images, which affects the performance of autonomous cars. To address this, we present an implementation of an exposure compensation method from the literature to auto-adjust camera exposure for the cameras mounted on a self-driving car. Furthermore, we extend this algorithm to incorporate gain compensation. The algorithm dynamically changes camera exposure time and gain settings with the intent to maximize image gradient information. The algorithm was evaluated in both indoor and outdoor environments, and experimental results demonstrate the effectiveness of our implementation. An open-source implementation of our technique is provided.
暂无评论