2005 Fiscal Year Final Research Report Summary
Vision System for Real-time Road Sign Detection.
| Project/Area Number |
15300051
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Perception information processing/Intelligent robotics
|
|---|
| Research Institution | University of Tsukuba |
|---|
Principal Investigator |
HIRAI Yuzo University of Tsukuba, Graduate School of Systems and Information Engineering, Professor, 大学院・システム情報工学研究科, 教授 (80114122)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
SAKAI Ko University of Tsukuba, Graduate School of Systems and Information Engineering, Associate Professor, 大学院・システム情報工学研究科, 助教授 (80281666)
|
|---|
| Project Period (FY) |
2003 – 2005
|
| Keywords | ITS / Road sign symbol / HSV color coodinate / Opponent-color filter / Discriminant analysis / Car mounted video camera |
|---|
| Research Abstract |
To assist safe driving, a vision system for real-time recognition of road sign symbols has been developed. This research focuses on six road signs : Speed limits to 30km,40km and 50km, No parking, No parking and standing, and No overtake. The process of road sign recognition consists of four stages : (1)Detection of red object, (2)detection of red circular object, (3)segmentation of sign symbol surrounded by a read annulus, and (4)classification of the symbol. Detection of red object is carried out by using HSV color coordinate system. Circularity of the detected red object is checked by fitting a circle equation to the outer boundary of the object. In order to detect partially occluded signs correctly, one of the four directional states, increase or decrease in x and y coordinates, is assigned to the boundary edge pixels and three consecutive partial sequences within a complete circular state sequence are accepted as a circular object. Segmentation of sign symbol is carried out by applying red-green opponent color filter and the filter output of each pixel is assigned to either foreground or background according to discriminant analysis. After the segmentation of symbol part structural features are detected, and road sign is classified by a decision tree.
The performance of the system has been investigated by 2 hours video data, and it has shown that the recognition rate is 97%. The processing speed is about 10 frames/sec by 2.8GHz Pentium processor. We are continuing the research to recognize "Stop" sign, for example, oversight of which will lead to fatal car accidents.
|
Research Products
(14 results)
