人工神经网络在高分影像防沙带识别中的应用  

Application of Artificial Neural Network in Anti-desertification Belt Identification in High-resolution Images

作　　者：尼加提·穆合塔尔 周迪 余洁[1,2,3,4,5] 朱琳 Nijat Muhtar(不详)

机构地区：[1]首都师范大学资源环境与旅游学院,北京100048 [2]首都师范大学城市环境过程与数字模拟国家重点实验室培育基地,北京100048 [3]首都师范大学地球空间信息科学与技术国际化示范学院,北京100048 [4]首都师范大学三维信息获取与应用教育部重点实验室,北京100048 [5]首都师范大学资源环境与地理信息系统北京市重点实验室,北京100048

出　　处：《地理空间信息》2021年第1期13-17,I0005,共6页Geospatial Information

基　　金：国家自然科学基金资助项目(41671417);科技创新服务能力建设——基本科研业务费(科研类)资助项目(19530050175)。

摘　　要：基于高分一号(GF-1)遥感影像,利用人工神经网络方法对新疆和田洛浦县沙漠边缘村庄周围的防沙带进行识别,并与目视解译的防沙带结果进行对比,验证识别精度。将人工神经网络方法的识别结果与传统的最大似然法进行对比分析,并利用图像识别领域中的准确率、召回率和F值3个指标对两种方法进行精度评价。结果表明,人工神经网络方法识别防沙带的准确率为79%、召回率为83%、F值为81%,最大似然法的准确率为74%、召回率为82%、F值为78%,人工神经网络方法的3个评价指标均高于最大似然法。In this study,based on the remote sensing images of GF-1 satellite,we used the artificial neural network method to identify the anti-desertification belts which were around the desert edge villages in Luopu County,compared with the results of visual interpretation for anti-desertification belts,and verified the accuracy of the identification.Then,we compared the recognition results of artificial neural network method with the traditional maximum likelihood method and evaluated the accuracy of the two methods by using the accuracy rate,the recall rate and F value in the field of image recognition.The results show that accuracy rate of artificial neural network method is 79%,the recall rate is 83%,and F value is 81 %.And the accuracy rate of the maximum likelihood method is 74 %,the recall rate is 82%,and F value is 78 %.All three evaluation indexes of artificial neural network method are higher than the maximum likelihood method.

关 键 词：人工神经网络方法 GF-1号 最大似然法 防沙带 

分 类 号：P237[天文地球—摄影测量与遥感]