论文部分内容阅读
针对面状水体识别过程中面状水体数据特征不宜提取、伪洼地易与面状水体混淆的问题,通过分析面状水体的面积、深度和潜在出水口等基本DEM数据特征,构建了面状水体识别模型,并将面状水体的三个数据特征和面状水体识别模型的计算结果作为输入输出神经元,利用RBF神经网络对建立的面状水体识别模型进行了仿真验证。从全国1∶250 000 DEM数据中选取150组洼地数据作为样本数据,采用减聚类算法对RBF神经网络进行训练,训练时样本的最小平均相对误差为2.75%,仿真的准确率为98%,表明面状水体识别模型可解决面状水体和伪洼地难以区分的问题,并提高了面状水体识别的准确率。
Aiming at the problem that the characteristics of surface water body data should not be extracted and the pseudo-depression is easy to be confused with the surface water body, the paper analyzes the characteristics of surface water body such as area, depth and potential outlet, The model was identified and the three data features of planar water body and the water surface identification model were used as the input and output neurons. The established water body recognition model was validated by RBF neural network. Select 150 sets of depression data from 1: 250 000 DEM data in the whole country as the sample data, and use the subtractive clustering algorithm to train the RBF neural network. The minimum average relative error of the training samples is 2.75%, the simulation accuracy rate is 98% It shows that the face water recognition model can solve the difficult distinction between the face water and the pseudo depression, and improves the recognition accuracy of the face water.