论文部分内容阅读
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。
Imaging spectrometer is a kind of “spectrogram ” optical remote sensing instrument. When the grating imaging spectrometer acquires the data cube, the obtained image will appear distorted due to the non-linearity of the spectrum of the dispersive element itself. As a result, the acquired image will still be distorted when the sampling frequency matches the splicing method reasonably. The edge-to-center grayscale gradation principle and genetic algorithm are used to extract the distorted feature points more accurately and to select the appropriate parameters. The support vector machine regression mathematical model is established to correct the distorted images. Compared with the conventional method of distortion correction, this method can effectively balance the local errors existing in the global correction method and improve the correction accuracy. Experimental verification Calibration test error can be controlled within ± 0.5 pixels.