地物波谱组合识别的神经网络模型

Model of Landmark Spectrum Recognition Based on Neural Network

遥感探测是一个多参数、高混合度的综合识别问题.利用BP神经网络的全局训练特性,直接利用教师信号来训练,建立了一个多组合的地物识别模型.实例验证了这种组合识别模型具有简单性、实用性和准确分辨等特点.

Telemetering is a comprehensive recognition with a lots of parameter and big confusion. This thesis use the overall drill of BP, and use the teacher signal to drill the weight directly, it become quick and precise to find a recognition model for landmarks. The analysis of cases shows that this recognition model has the properties of simplicity,practicability and pricision.