城市绿地提取及动态变化监测方法

Analysis and study on urban green space extracting and dynamic monitoring method

选择西安市城区三环以内作为研究区域,使用多时相LANDSAT TM影像,分别采用基于简单规则的决策树分类和支持向量机(SVM)分类法提取城市绿地信息并对其精度进行评价。针对TM影像绿地信息提取中存在的混合像元问题,将模糊C均值法(FCM)引入到绿地提取中。研究结果表明:SVM分类法相比于简单规则的决策树分类法,平均分类精度提高了15%,更有利于城市绿地信息的提取,然而对城区中心的绿化带、行道树等小面积绿地信息提取仍然不全面;引入FCM算法后,可根据像元中各类别的不同隶属度,进行更加精细和准确的分类,城市中面积较小的绿地信息都能被很好地提取出来,分类精度得到进一步提高,该算法很好地解决了绿地信息提取中的混合像元问题。

In order to overcome the disadvantages of the existing method,we extract green space information and solve the problem that consume much time and labor when changing detection of urban green space.Taking Xi’an urban green space as researching object,using the LANDSAT TM satellite images,we employ decision tree classifier based on simple rules and the classification method of support vector machine（SVM） respectively to extract the urban green space information and evaluate its accuracy.Then,the paper applies the fuzzy C-means method（FCM） into the extraction of green space to solve the problems of the mixed pixel.The introduced algorithm can calculate the fuzzy membership value of the pixel in each classification category.The research results show as following.（1） Compared to the decision tree classification method which is based on the simple rules,the classification accuracy of SVM is increased by about 15%.It is more conducive to extract urban green space information.However,the extraction of small area of green space information is still incomplete,such as the information of the green belt and street trees.（2） Using the FCM algorithm,we can do a more refined and accurate classification result according to the pixel to different categories of membership.Small area of green space information can be extracted first-rate and the classification accuracy is improved well.This proposed algorithm can solve the problem of mixed pixel in the green information extraction.