基于视觉感知的车门开启提示系统设计
关键词 Lucas-Kanade 法，Horn-Schunck 法，Brox 等人的 Warping 法，灰度化
处理，自适应阈值分割，形态学滤波，目标查找
Title Visual perception based car door opening security-reminder system
Abstract
Lucas-Kanade Method, Horn-Schunck Method and Warping Method developed by
Brox etc. are introduced in this article respectively. These methods have
been adopted to compute the optical flow of image sequences. And the
resulted images will be processed by graying, adaptive thresholding
segmentation, morphological filtering and target searching successively.
Moving objects that affect vehicle’s door open will be marked in the image
sequences. We tested these optical flow methods with two different image
datasets and evaluated both their pros and cons. Furthermore, we tested
their performance by using the above mentioned algorithms to search moving
target and analyzed their influence on the visual perception based car door
opening security-reminder system. At last, suggestions have been provided
based on above test results.
Keywords Lucas-Kanade method, Horn-Schunck method, Warping method
developed by Brox etc., graying, adaptive thresholding segmentation,
morphological filtering, target searching
目 次
1 绪论 .. 1
1．1 光流法简述 . 1
1．2 光流法发展 . 1
1．3 光流法所存在的问题 . 2
1.3.1 大位移问题 2
1.3.2 鲁棒性问题与精确性问题 2
1.3.3 实时性问题 2
2 光流场求解 4
2．1 有关假设与方法 . 4
2.1.1 灰度值恒定假设 4
2.1.2 梯度恒定假设 .. 4
2.1.3 平滑性假设 4
2.1.4 金字塔方法 5
2．2 Lucas-Kanade 方法 ... 6
2．3 Horn-Schunck 方法 ... 8
2．4 Brox 等人的 Warping 方法 . 9
3 运动目标的获取 ... 13
3．1 光流场的灰度化处理 13
3．2 自适应阈值分割 13
3．3 形态学滤波 14
3．4 查找和获取目标 15
4 评估 . 16
4．1 光流算法的评估 16
4.1.1 相关评估量 ... 16
4.1.2 算法性能评估 . 17
4．2 整个提示系统的评估 26
4.2.1 Lucas-Kanade 法产生光流场的系统 ... 26
4.2.2 Horn-Schunck 法产生光流场的系统 ... 27
4.2.3 Brox 等人的 Warping 法产生光流场的系统 . 28