采用边界评价的红外视频运动目标时空域分割方法

Spatiotemporal segmentation method of moving-object using boundary evaluation in infrared video

为了在红外视频中准确分割运动目标,提出了一种基于边界评价的红外运动目标时空域分割的新方法。首先,利用运动目标在时域差分图像中的"空洞"效应,提取出最有意义运动目标种子点。重点是运动目标的空间分割,利用种子区域整体与局部的关系,在提取出的种子上进行区域生长,可以得到不同生长阈值下的运动目标分割掩膜。为确定最佳生长阈值,提出了一种无需先验知识的红外目标分割掩膜边界评价准则,并采用"分割-评价-再分割-再评价"的循环迭代模式,利用"由粗到精"的搜索方法,找出最佳的生长阈值,同时得到最佳的运动目标分割掩膜。实验证明,所提出的方法能在红外视频中准确分割出运动目标区域,效果良好,性能鲁棒。

In this paper, a new method was presented for spatiotemporal segmentation of moving-object using boundary evaluation in infrared video. At first, the ideal seeds of every moving object were extracted based on the ＂holes＂ effect of temporal difference, respectively. The wok focus was spatial segmentation. On the basis of the relationship between the global and local standard deviation of seeds, the segmented masks could be grown form the ideal seeds by using different growing thresholds. For determination of the best growing threshold, a criterion was constructed for evaluating the boundary of the segmented mask of infrared moving-object without prior knowledge. According to the proposed criterion, an iterative model which was ＂segmentation-evaluation-segmentation-evaluation＂ and the search method called as ＂coarse to fine＂ were applied to find the best growing threshold. Meanwhile the best segmented mask was obtained too. The experiment results show that the proposed method is superior and effective on segmentation of moving object in infrared video.