Research On Subspace Mode Extracted From Traffic Sign Shape Features

Along with the socio-economic development, Chinese road transport systemdevelops rapidly, it brings convenience to people’s lives, but also brings troubles topeople’s travel. The traffic sign recognition system provides one auxiliary method tosolve those troubles; therefore it attracts scholar’s widespread attention.Under natural conditions, color information is sensitive to external environment.Relatively, shape information is more stable. Although the images will have somedeformations due to the changes in acquisition angle, this deformation can be correctedby affine transformation, because the traffic signs are rigid objects. As a means of featureextraction, subspace methods can effectively reduce the dimension of the original images,and extract the most important information. This paper studies the subspace modes oftraffic sign shape features, mainly has completed following several parts of work.Starting from the principle and constructing method, this paper conducted acomprehensive analysis of subspace methods, and summarized their characteristics andapplication range; in this paper, the traffic sign shape features obtained by binarizationfirstly. Then, features extracted from traffic sign images using PCA,2DPCA andPC-LDA. Finally, it uses the minimum range sorter to recognize the features; By Matlabon the PC, using the MPEG7CE Shape-1Part-B computer graphics database, standardtraffic sign images and natural traffic sign images, it completes the authentication.Because the gray information is more simply in binary images, and the featureextracted by PC-LDA has distinct advantage in both object description and objectclassification. The results show that extracting features from binary images usingPC-LDAcan improve the traffic sign recognition system.