To solve the issue of poor accuracy in remote sensing image fusion,a method of oblique photography image fusion based on parameter adaptive convolutional neural network(CNN)was studied and constructed.Firstly,an image fusion model based on deep convolutional neural network was built.Then,channel similarity attention was introduced to enable adaptive learning of network features.The results indicate that the proposed algorithm performs well in spatial correlation coefficient,with an index of 0.758 and a general image quality of 0.57.In addition,the algorithm has less image distortion and a peak signal-to-noise ratio of 23.75 dB.In the Pavia Center dataset,the proposed algorithm still shows good spatial detail information retention ability,image fusion accuracy,matching effect,and image fusion effect.It also has good image fusion performance on the CAVE dataset,with the highest correlation coefficient of 0.997,spectral angular reflectance and relative global synthesis error of 4.440 and 2.617,respectively.The experimental results demonstrate that the proposed model has good ability to preserve detailed information in image fusion.The research results can provide certain technical support for the image fusion of oblique photography and promote the development of oblique photography and remote sensing technology.
维普
万方数据