• Proceedings of the Korean Information Science Society Conference
• /
• 2003.04c
• /
• pp.196-198
• /
• 2003

본 논문에서는 스네이크 모델의 에너지 최소화 알고리즘을 개선하여 속도와 정확도에 대한 문제를 해결한다. 개선된 알고리즘은 스네이크를 이루는 정점들의 적합성에 따라 탐색 윈도우를 가변적으로 확장시킴으로써 빠르고 정확하게 윤곽선을 추출한다. 또한 정점의 정렬과정을 통해 정점이 지역적 최소점에 빠지는 것을 방지하며 스네이크의 연속성과 완만성을 보존한다.

• The KIPS Transactions:PartB
• /
• v.10B no.5
• /
• pp.527-534
• /
• 2003

An active contour model called snake is powerful tool for object contour extraction. But, conventional snakes require exhaustive computing time, sometimes can´t extract complex shape contours due to the properties of energy function, and are also heavily dependent on the position and the shape of an initial snake. To solving these problems, we propose in this paper an improved snake called "shaking snake", based on a greedy algorithm. A shaking snake consist of two steps. According to their appropriateness, we in the first step move each points directly to locations where contours are likely to be located. In the second step, we then align some snake points with a tolerable bound in order to prevent local minima. These processes shake the proposed snake. In the experimental results, we show the process of shaking the proposed shake and comparable performance with a greedy snake. The proposed snake can extract complex shape contours very accurately and run fast, approximately by the factor of five times, than a greedy snake.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.13 no.5
• /
• pp.937-943
• /
• 2009

In this paper, the algorithm discriminating the existence and nonexistence of bolts and nuts using image processing in an automobile T-Bar influencing the vibration of a frame, is introduced. To distinct whether bolts and nuts exist or not, the features of bolts and nuts are learned, and then these feature values are matched using a statistical pattern matching algorithm. Furthermore, the minimum and maximum variation rate of pixel values are used since the matching rate is low with the large variation of pixel values of bolts and nuts in each image. The proposed method in this paper is very efficient in the automation of inspection requiring real time since the inspection time is significantly reduced compared with the conventional methods.