혼성 매칭에 의한 다중 이동 물체의 동시 추적과 분류

최연성, 김선우, 최용원 et al. · 한국전자통신학회 학술대회지 · 2008

In this paper, the multiple objects tracking methods for smart surveillance system are proposed. In the proposed system, the object with salient motion is detected by a temporal differential image to be applied weight. After extracted the information for detected objects from multiple features, the object is tracked. The proposed whole-body object tracking algorithms is more efficient to track the moving object due to the synthesis of the spatial position, shape, intensity and correspondence features. The proposed histogram-based correspondence matching approach recognizes the identities of objects after occlusion. We proposed a novel object classification algorithm based on the object shape similarity. This algorithm gives promising results in categorizing object types. But much efforts to generate a template database for all possible silhouettes of different classes is required. In experiment, We confirmed that the success rate of classification by the proposed method is higher about 20~30% than that of the existing method.

Cited by 0 works.

Authors