基于顶帽-底帽变换和二次图像分割的目标提取

Target extraction based on the hat-bottom cap transformation and second image segmentation

文中分别采用最大内间方差法、迭代法、二维最大熵法、KFCM单次阈值分割法和灰度均值区域生长法对坦克和卡车从背景中进行提取。针对单次阈值分割法和灰度均值区域生长法在提取目标时出现不能分割出车身投影或者大量丢失车体等问题,本文结合单次阈值分割法和区域生长法的优势,提出了基于顶帽-底帽变换和二次图像分割的目标提取算法。对比试验表明,该算法在较完整保留车体部分同时能去除车体的投影,很大程度上抑制了目标的形心漂移。

This paper use the method of maximum amount of variance, iterative, two-dimensional maximum entropy, KFCM single threshold segmentation and grayscale average region to put the tanks and trucks segmented from the background growing respectively. To solve the problem of single threshold segmentation method and the grayscale average region growing method being not able to segment the car or lost some part in extracting target projection, this paper has combined the advantage of single threshold segmentation method and the region growing method, and has put forward a target extraction method based on the hat-bottom cap transformation and the second image segmentation. Comparison test shows that the algorithm can remove the bodywork projection while remaining the intact body, which can inhibit the target centroid drift to a large extent.