基于微种群遗传算法和自适应BP算法的遥感图像分类

介绍了采用微种群遗传算法和自适应BP算法相结合的混合遗传算法来训练前向人工神经网络(BPNN)的方法。即先用微种群遗传学习算法进行全局训练,再用自适应BP算法进行精确训练,以达到加快网络收敛速度和避免陷入局部极小值的目的。将此算法用于遥感图像分类,网络的训练速度及分类结果表明,该算法收敛速度较快,预测精度较高。

Classifying remote sensing images based on micro-group genetic algorithm and adaptive back propagation algorithm

An image classification method is presented, which trains the back-propagation neural network (BPNN) by combining micro-group genetic algorithm and adaptive back propagation algorithm. The project use micro-group genetic algorithm training the BPNN to get an approximate optimal solution in the whole area, and then use adaptive back propagation algorithm to modify the solution to a better one. This hybrid genetic back-propagation algorithm can speedup the convergence and avoid getting into local optimal solutions. When use it to classify a remote sensing image, the result shows that it is superior to the corresponding uncombined versions in the convergence speed and classification accuracy.