Dissertation
Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device

Study on Depth Estimation of Vehicle Infrared Images

Author ZouFangYu
Tutor SunShaoZuo
School Donghua University
Course Detection Technology and Automation
Keywords NV images Depth estimation Vehicular infrared Monocular cue
CLC TP391.41
Type Master's thesis
Year 2013
Downloads 58
Quotes 0
Download Dissertation

Nowadays, vehicle infrared is used more and more widely, but the images have no color and depth perception, it is not fit to understand the scene. If we can restore the scene color and estimate its depth, combine the depth with hue and saturation, which will greatly enhance the ability of understanding the scene, research contents of this article is to estimate the depth of vehicle infrared image. Depth is the distance of scene which is far from the viewer, methods of depth estimation have two categories which are monocular and binocular at home and abroad. This article is for vehicle infrared image, the focus of research is methods of depth estimation.The main content is divided into two parts, the first part is that depth estimation based on the image content, improving depth estimation based on the image content by OTSU, and depth estimation of infrared image based on line detection; the second part is that depth estimation of vehicle infrared video. Specific as follows:1. Depth estimation of the single vehicle infrared image based on image content. Because of the specific scene, scene classification used the method of region growing segmentation, and estimated their depth, giving scene color by combining the depth information with hue and saturation. The experimental results shown that the image has stereo perception.2. We proposed a depth estimation algorithm to improve the last algorithm. In the first, using top-hat transform to preprocess infrared image. Then we extracted the foreground by the OTSU. Finally, estimating the depth information of the background and foreground. We researched a method of depth estimation which based on the line detection, detecting straight lines on the image, then determining prospect depth information by the line information.3. We researched the method of depth estimation on vehicle infrared video, proposed a video depth estimation algorithm which is based on frame difference region. In the first, estimating depth information which based on frame difference of sky and ground area of vehicle infrared video, Frame difference is result from front and back frames. Then generate a reference depth map which based on the trees region of previous frame, correcting the depth of latter frame according to coordinate information of the trees.

Related Dissertations
More Dissertations