SVM 기법을 이용한 색상 인식 알고리즘 = Color Classification Based on the Support Vector Machine Method

다국어 초록 (Multilingual Abstract)

We present a vehicle color classification method from outdoor vehicle images. Although the vehicle color recognition is important especially for the newest applications including ITS(intelligent transportation system), we have no significant previous result at least to our knowledge. In this paper, first, we classify all the pixels of a vehicle image into seven vehicle color classes using Bayesian Maximum Likelihood method. Then, we construct the low dimensional feature vector, by using the histograms for the classes. As a final step, we use the SVM(support vector machine) method to classify these feature vectors into seven vehicle color classes: black, blue, green, red, silver, white and yellow. Our implementation result shows 92.74% of success rate for 700 outdoor vehicle images. And the classification speed was 36 times faster compared to the case of constructing the feature vectors in a brute-force way.

목차 (Table of Contents)

• 1. 서 론 1
• 2. 특징 벡터의 구성 6
• 2.1 차량의 색상 특징 6
• 2.2 특징 벡터의 차원 축소 9
• 3. Support Vector Machine 19
• 4. 실험 방법 및 결과 21
• 4.1 베이지안 최대 우도 기법을 이용한 색상 인식 22
• 4.2 Brute-force 방법을 이용한 색상 인식 24
• 4.3 수행 시간 측정 29
• 5. 결론 및 향후 과제 31
• 참고문헌 33
• 영문초록 35