SVM 的光栅成像光谱仪图像畸变校准方法

成像光谱仪是一种图谱合一的光学遥感仪器。光栅成像光谱仪在获取数据立方体时，由于色散元件本身光谱展开的非线性，导致获取的条带像出现畸变，致使采样频率与拼接方式合理匹配时，获取的图像依然会出现畸变。利用边缘到中心灰度值渐变原理和遗传算法，更准确的提取畸变特征点，选择合适的参数，建立支持向量机回归数学模型，对畸变图像进行校正。与其常规畸变校正方法相比，该方法能够有效的兼顾全局校正法中存在的局部误差，提高校正精度。实验验证校正检验误差可以控制在0.5 个像素之内。

Image distortion calibration of imaging spectrometer with grating by SVM

Imaging spectrometer was a kind of optical remote sensing instruments which combined image with spectrum. Grating imaging spectrometer can acquire data cube. When the sampling frequency and splicing was reasonable, because of the nonlinear about dispersive elements making spectral expansion, lead to the distortion of strip image, as well as the distortion of spliced image. The distortion feature points were extracted more accurate by using the gray gradient from the edge to the center and genetic algorithm. By choosing appropriate parameters, we established the support vector machine regression mathematical model to correct the distortion. Compared with the conventional distortion correction method. This method was able to balance effectively the error from the global and the local and improve the accuracy of correction. The calibration of error can be controlled within 0.5 pixels by experimental verification.