在形状约束下进行目标区域生成  

The Production of Objective Region with the Shape Factor

作　　者：管佳颖[1] 李德仁[1] 关泽群[1] 

机构地区：[1]武汉测绘科技大学信息工程学院,湖北武汉430079

出　　处：《遥感技术与应用》2001年第1期49-51,共3页Remote Sensing Technology and Application

基　　金：高等学校骨干教师资助计划资助

摘　　要：针对感兴趣的目标从遥感图像上自动圈出进行了探讨。主要包括使用矢量 Prewitt算子获取梯度方向 ,将其引入 Snake算法中进行目标轮廓逼近 ,并在轮廓逼近中附加形状约束。在实现过程中 ,对求梯度方向的算法提出了改进算法 ,即无需求出梯度方向的精确角度 ,直接判断该像素的梯度方向 ,减少了计算量。另外还对阈值选取问题进行了探讨。In the processing of the remote sensing imagery interpretation, the needs of selecting the objects that we are interested become more and more important. But much of the work is still accomplished by handwork presently. In order to change this situation, to improve the automatic analysis of the computer, this paper makes some discussion of the objects' automatic detection that we are interested.In color edge detection, we used to applying the same method as monochrome images processing, thus we can't take good advantage of much color information. The idea of using vector order Prewitt operators to edge detectors can help us make full use of the color information. On the other hand, many Remote Sensing images are multi\|wave\|band. Therefore, using vector order Prewitt operator is more useful than conventional Prewitt operators in processing the imagery of Remote Sensing. The “Snake” algorithm is a famous active contour model, which is often used when we need a precise contour of an object in image processing field. This method will be more reliable if we consider the shape factor\|the relationship among all the Snake points in the procedure of finding the minimum energy position.According to these approaches, this paper introduces the concept of “vector order Prewitt operators” into the snake algorithm with the shape factor. During the processing, it makes some improvement in the gradient direction algorithm. That is, it can judge the orientation of a pixel's gradient directly, without computing the exact degree of the gradient. Consequently this helps to reduce the amount of computation. In addition, we make some argument about the threshold. We found that make threshold in different sequences will have different effects. The effects are shown in this paper. The effectiveness of the proposed method and its value of practicality can be confirmed by the experiments.

关 键 词：矢量Prewitt算子 Knake算法 形状结束 边缘检测 遥感图像解释 

分 类 号：TP753[自动化与计算机技术—检测技术与自动化装置]