基于自编码神经网络的航空物探遥感数据分类方法研究

Research on classification method of airborne geophysical remote sensing data based on self-coding neural network

摘要:

航空物探遥感数据的采集过程中受到电磁波辐射等外界因素的影响,导致航空物探遥感数据分类准确率较低,为此提出基于自编码神经网络的航空物探遥感数据分类方。根据航空物探对象的基本特征,设置遥感数据的分类标准。通过辐射校正、几何纠正、噪声消除等步骤,完成航空物探遥感数据的预处理。构建自编码神经网络,利用自编码神经网络算法,从光谱、形状、纹理等方面提取遥感数据特征,通过特征匹配确定航空物探遥感数据的所属类型。通过分类性能测试实验得出结论:所提方法的全局遥感数据分类成功率和错误率的平均值分别为99.8%和0.6%,局部遥感数据分类的成功率和错误率的平均值分别为99.8%和0.3%,即所提方法在分类性能方面具有明显优势。

Abstract:

The acquisition process of airborne geophysical remote sensing data is affected by external factors such as electromagnetic wave radiation, resulting in low classification accuracy of airborne geophysical remote sensing data. Therefore, a classification method for airborne geophysical remote sensing data based on self coding neural network is proposed. Set classification standards for remote sensing data based on the basic characteristics of aerial geophysical exploration objects. Complete the preprocessing of airborne geophysical remote sensing data through radiation correction, geometric correction, noise elimination, and other steps. Construct a self coded neural network, use self coded neural network algorithms to extract features of remote sensing data from aspects such as spectrum, shape, texture, and determine the type of airborne geophysical remote sensing data through feature matching. Through classification performance testing experiments, it is concluded that the proposed method has an average classification success rate and error rate of 99.8% and 0.6% for global remote sensing data, and an average classification success rate and error rate of 99.8% and 0.3% for local remote sensing data, respectively, indicating that the proposed method has significant advantages in classification performance.