We develop a deep learning-based matching method between an RGB and an infrared image that were captured from satellite sensors. A densely-connected CNN is designed to extract common features from different spectral bands. The network consists of a series of densely-connected convolutions to make full use of low-level features and an augmented cross entropy loss to avoid model overfitting. The network takes band-wise concatenated RGB and infrared images as the input and outputs the similarity score of the RGB and infrared image pair.
The proposed method has been also demonstrated to have high performance and generalization ability applying to multi-temporal remote sensing images and close-range images.
Citing us if you find it is helpful in your study (Zhu, R., Yu, D., Ji, S., & Lu, M. (2019). Matching RGB and Infrared Remote Sensing Images with Densely-Connected Convolutional Neural Networks. Remote Sensing, 11(23), 2836.
This code contains a Keras implementation of "Matching RGB and Infrared Remote Sensing Images with Densely-Connected Convolutional Neural Networks" and the original paper is published in Remote Sensing. The proposed network was trained and tested on a single NVIDIA GeForce GTX 1080TI GPU with 11 GB RAM.
python dense_image_matching_network.py