Browsing by Subject "Detection and identification"
Now showing 1 - 2 of 2
Results Per Page
Sort Options
Item Estimation of Vehicle's Lateral Position via the Lucas-Kanade Optical Flow Method(Intelligent Transportation Systems Institute, Center for Transportation Studies, University of Minnesota, 2012-09) Yang, Jiann-ShiouThe use of rumble strips on roads has proven to be an effective means of providing drivers lane departure warning (LDW). However, rumble strips require an infrastructure and do not exist on a majority of roadways. Furthermore, rumble strips present a difficult issue of where to establish the rumble-strip distance threshold. To develop an effective virtual rumble-strip LDW system where the rumble-strip threshold is allowed to vary according to the risk of the vehicle departing the road, it is essential to know the vehicle’s lateral characteristics; in particular, the vehicle’s lateral position and speed. In this report, we use image processing via an in-vehicle camera to estimate the vehicle’s lateral position and speed. The lateral position is estimated by determining the vehicle’s heading angle via a homography and the Lucas-Kanade optical flow techniques; while the lateral speed is determined via the heading angle and the vehicle’s On Board Diagnostic (OBD)-II forward speed data access. The detail of our approach is presented in this report together with our findings. Our approach will only need the minimal set of information to characterize the vehicle lateral characteristics, and therefore, makes it more feasible in a vehicle application.Item Thermal Image-Based Deer Detection to Reduce Accidents Due to Deer-Vehicle Collisions(Intelligent Transportation Systems Institute, Center for Transportation Studies, University of Minnesota, 2013-01) Zhou, DebaoDeer-vehicle collision (DVC) is one of the most serious traffic issues in the Unite States. To reduce DVCs, this research developed a system using a contour-based histogram of oriented gradients algorithm (CNT-HOG) to identify deer through the processing of images taken by thermographic cameras. The system is capable of detecting deer in low visibility. Two motors are applied to enlarge the detection range and make the system capable of tracking deer by providing two degrees of freedom. The main assumption in the CNT-HOG algorithm is that the deer are brighter than their background in a thermo image. The brighter areas are defined as the regions of interest, or ROIs. ROIs were identified based on the contours of brighter areas. HOG features were then collected and certain detection frameworks were applied to the image portions in the ROIs instead of the whole image. In the detection framework, a Linear Support Vector Machine classifier was applied to achieve identification. The system has been tested in various scenarios, such as a zoo and roadways in different weather conditions. The influence of the visible percentage of a deer body and the posture of a deer on detection accuracy has been measured. The results of the applications on roadside have shown that this system can achieve high detection accuracy (up to 100%) with fast computation speed (10 Hz). Achieving such a goal will help to decrease the occurrence of DVCs on roadsides.