人工神经网络在高分影像防沙带识别中的应用

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。

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

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.