摘要
基于改进BPNN的遥感图像分类,先利用ERDASImagine的Classifier模块对原始遥感图像进行聚类分析并对比,确定分类类别数。后在原始图像上采集各类别的训练样本和测试样本,通过输入训练样本集训练BP神经网络分类器。最后输入整幅遥感图像到训练好的分类器,得到分类图像。试验表明,其精度优于传统统计分类方法。
The remote sensing image classification system is based on improved BPNN. At first, the Classifier module in ERDAS Imagine was used to carry out the non-monitoring classification and contrast of original remote sensing images; and the class identification numbers of classification were ensured. Then, the training samples and testing samples of varied classes were gathered in remote images; the BP neural network classifier was trained through inputting training samples. At last, the complete remote sensing image was inputted into trained classifier to get classification images. The test proved that the precision was higher than traditional statistical classification method.
出处
《兵工自动化》
2006年第4期28-29,共2页
Ordnance Industry Automation