基于人工神经元网络技术的土地利用/覆盖变化探测

Land Use/Cover Change Detection Based on Artificial Neural Network

针对现有的一些土地利用 /覆盖变化探测方法存在的某些不足 ,提出了利用人工神经元网络 (artificialneuralnetwork ,ANN)进行土地利用 /覆盖变化探测的方法 ,并对ANN网络的输入输出、网络结构和不同的网络模型进行了深入研究 ,充分利用已有的基础地理信息和高分辨率遥感影像辅助选取了ANN训练样本。试验结果表明 ,利用ANN总体上可提高土地利用

ANN has been introduced in land use/cover change detection to improve the change detection results.In this paper,the input and output,the structure and reasonable settings of ANN have been studied and compared.Different ANN models and algorithms have been introduced to improve the performance of ANN.The results have shown that LVQ and MAALR(momentum_ [FK(W7?40ZQ]adaptive adjust ment of learning rate) have turned to be more efficient in land use/cover change detection than BPNN because they take less ANN training time and have no local minimum. The experiments based on TM satellite images of different time have shown that ANN method is practical and efficient for the change detection.The accuracy of it is higher than those of the traditional methods,and it can provide both changed areas and categories at the same time.Besides,it is easy to integrate multi_source data because of low demand for data distribution.