CAJ | 학술논문

利用最优化云变换HOT检测云区位置, 再经过云区偏移和膨胀计算检测云影. 对于多时相遥感图像的辐射特征存在非线性关系问题, 采用BP神经网络对非线性函数进行模拟, 预测仿真后各像元的灰度值, 实现参考影像与目标影像的光谱匹配. 根据云区及其阴影的检测图, 将BP神经网络辐射校正后的参考影像镶嵌入目标影像, 达到影像修复的目的. 实验结果表明, BP神经网络对于Landsat数据具有很好的函数逼近效果, 去云后图像质量得到较大改善.
이용최우화운변환HOT검측운구위치, 재경과운구편이화팽창계산검측운영. 대우다시상요감도상적복사특정존재비선성관계문제, 채용BP신경망락대비선성함수진행모의, 예측방진후각상원적회도치, 실현삼고영상여목표영상적광보필배. 근거운구급기음영적검측도, 장BP신경망락복사교정후적삼고영상양감입목표영상, 체도영상수복적목적. 실험결과표명, BP신경망락대우Landsat수거구유흔호적함수핍근효과, 거운후도상질량득도교대개선.
An approach using BP ANN for radiometric calibration in multi-temporal remote sensing images is proposed in this paper.First of all, the cloud is detected by haze optimized transform method, and the shadow is computed by offsetting and expanding the area of cloud .Then the spectral nonlinear relation can be solved by BP ANN to simulate the nonlinear function and predict each pixel gray value, so that the spectral features of two images can be unified.Finally the target image would be repaired by replacing the pixels of color corrected im-age.Results show that for Landsat data, BP ANN has a good effect on function approximation, and the image quality can be greatly improved after employing this method.