基于动态RBF神经网络的遥感图像中烟辨识

针对森林火灾的突发性和分散性特点,根据森林火灾中的烟影像的特征,结合RBF神经网络分类器通用性的特点,采用PCA方法降低图像的维数和Fisher线性判别方法提取遥感图像中烟的影像特征,用特征值数代替训练样本数,并作为RBF网络输入数,在对样本训练的过程中提出一种改进型的混合算法。仿真实验表明,该算法对烟图像识别的准确率较高。

Smoke Identification of Remote Sensing Images Based on Dynamic RBF Neural Network

Aiming at the characteristics and the dispersion sudden of the forest fires,and according to the characteristic of smoke in the forest fire,combined with RBF neural network classifiers universal characteristics,uses the PCA method reduce the dimension of the image and the Fisher linear discriminant method to extract imaging features of the smoke of the remote sensing images,uses the number of eigenvalues instead of the number of training samples,and as the RBF network input number,proposes an improved hybrid algorithm in the training sample process.Simulation result shows that the algorithm of smole image recognition has high accuracy.