基于BP神经网络的地图数字化误差纠正

The BP neural network method for errors correction in digital maps

由于在数字化采集过程中不可避免地会引入系统误差和异常误差,因此消除和削弱这些误差的影响是提高空间数据质量的关键。然而由于图纸变形不均匀,扫描误差又极其复杂,用常规的多项式拟合技术只能消除部分有规律的系统误差,很难完全消除它们对地图数字化坐标的影响。BP神经网络是一个高度非线性映射系统,能以任意精度逼近。结合地图数字化坐标改正的特点,本文给出了基于BP神经网络地图数字化坐标误差纠正的方法,并通过实例验证了该方法的有效性。

Due to some systematic errors and outliers introduced in the process of digital, it is a key issue to weaken or eliminate these influences. As the map distortion is not the same on the map or in the process of scanning, polynomial fitting can only eliminate some regular systematic errors and cannot completely eliminate their influences. BP neural network is a very complicated system which is used to approach at any precision. Combining the characteristic of the coordinate for map digitizing, the method of errors correction in digital maps is given, and effectiveness in improving the quality of digital maps is proved by an actual example.