Keywords: Cyclist, Risk Assessment, Focus of Expansion, Optical Flow, Smartphone Data, Computer Vision
Due to their zero pollution emissions, health improvements benefits and ease of access bicycles are gaining an increasing popularity as a mean of transportation in today’s world. However, traffic accidents involving bikes are not decreasing, as well as fatalities. Thus, it is important to assess cyclists’ safety in urban scenarios to allow city planners to develop better infrastructures that foster better protection for cyclers. Therefore, from smartphone captured data and video, we propose a video-based framework to assess dangerous situations for bicyclists. We take advantage of motion estimation (optical flow) to estimate the Focus of Expansion on a set of images and then use this to define risk areas on the image. We then use the defined areas on the image to create a risk descriptor on the given situation and given the detected objects on the image. Our framework enables the assessment of risk on different criteria (Path Occupation and Proximity) based on our risk descriptor. Finally, we test our framework on real data gathered from the improved developed smartphone application and achieve promising results.
Author: Miguel Costa
Supervisors: Manuel Marques, João Paulo Costeira