In remote sensing image processing, image segmentation requires not only the accuracy of segmentation, but also requires the lightweight of the network model as a way to improve the training speed to achieve faster response time. Therefore, in the pursuit of efficient real-time application requirements, this paper improves the lightweight network structure ICNet. Firstly, the efficient channel attention (ECA) module is added to ICNet, and the ECA module improves the feature extraction capability and ensures the simplicity of the model through the weighting operation of the key channel information. Then the joint pyramid upsampling (JPU) module is also introduced, which is integrated into the original ICNet's upper, middle, and lower branching structure for processing feature information. In the subsequent experiments, the proposed EJICNet network structure is trained and evaluated in depth, and the experimental results clearly show that EJICNet significantly reduces the computational complexity while maintaining a higher segmentation accuracy compared to the existing network structure. This proves that the optimisation method proposed in this study balances efficiency and accuracy, and satisfies the network's need for real-time performance.
The purpose of pansharpening is to fuse low-resolution multispectral image (LRMS) and high-resolution panchromatic image (PAN) to obtain high-resolution multispectral (HRMS). In response to the shortcomings of traditional remote sensing image fusion algorithms causing spectral distortion, more and more deep learning algorithms are utilized, and this paper proposes a new deep network structure, two-branch Self-Attentive DenseNet network. In terms of maintaining high spatial resolution, the image feature information is extracted by different inch-scale convolutional kernels, and the effective feature information is enhanced to suppress the invalid image information by using DenseNet network model and introducing Self-Attention, and the fused image spectral information is enhanced by using hopping connection to maintaining the spectral structure. Experiments show that the proposed method of this paper has improved image quality evaluation metrics compared with the previously existing traditional fusion algorithms and deep network algorithms.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.