红外目标的边界检测和标注

Boundary detection and object labeling in infrared object detection

边界检测和标注是红外目标探测和识别的一个重要环节。自然背景复杂的纹理特征和红外图像中的噪声影响，以及红外目标模糊的边缘，给红外目标的边界检测和分割带来一定的困难。分水岭算法作为一种数学形态学的图像分割方法，为进行目标特征分析和识别提供了可能，然而，分水岭算法产生的过度分割，又为背景和噪声抑制增加了难度。文中在分水岭算法的基础上，结合模糊逻辑方法，对自然背景下的红外人工目标的边界检测和分割以及红外背景的抑制进行了探讨。

Boundary detection and object labeling is key point in infrared object detection and recognition. The effect of natural background and noise in infrared image, fuzzy edge of infrared object, make it difficult to segment and label artificial object in natural background. It′s convenient to using watershed transform to analyze texture of target and label target, but over\|segment produced by watershed transform bring additional difficulties in background noise suppression. In this paper, an algorithm based on watershed transformation and fuzzy logical catchment basins combination is proposed to suppress or remove the feature of natural background, and applied to infrared object boundary detection and labeling. \;