적외선 영상의 효율적인 비표적 제거 방법

다국어 초록 (Multilingual Abstract)

In this paper, we propose a new method for eliminating non-targets which have low correlation between the target and candidate patches in infrared images. First, we generate an average target template using various target patches. Then we calculate an adaptive threshold value in the average target template and the Patch Binary Ratio. The Patch Binary Ratio means the ratio between the patch area and the candidate target area. Next, we apply a threshold to the input infrared image and extract candidate patches. Then we calculate the Patch Binary Ratio of all candidate patches. Finally, we compare the Patch Binary Ratio of the average target template with the Patch Binary Ratio of each candidate patch, then classify candidate patches into non-targets if the Patch Binary Ratio is over a certain value range. The proposed method runs in 0.83ms for 37,000 candidate patches and eliminates 90.4% of non-target patches on average.

참고문헌 (Reference)

1 김병익, "적외선 영상에서 다층 구조 요소 NWTH 변환을 이용한 소형 표적 검출 방법" 한국정보기술학회 9 (9): 57-64, 2011

2 H. Bay., "SURF : Speeded Up Robust Features" 404-417, 2006

3 P. Viola., "Robust real-time face detection" 57 (57): 137-154, 2004

4 P. Viola., "Rapid Object Detection using a Boosted Cascade of Simple Features" 511-518, 2001

5 Z. Kalal., "P-N learning : Bootstrapping binary classifiers by structural constraints" 49-56, 2010

6 J. P. Lewis, "Fast Normalized Cross-Correlation" Vision Interface 120-123, 1995

7 M. I. C. Murguia, "Document Segmentation Using Texture Variance and Low Resolution Images" 164-167, 1998

8 Y. Fang., "Comparison between Infrared-imagebased and Visible-image-based Approaches for Pedestrian Detection" 505-510, 2003

9 M. Leung., "A region based approach for human body motion analysis" 321-339, 1987