We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Target detection algorithm based on super-resolution color remote sensing image reconstruction.
- Authors
Zhihong Wang; Chaoying Wang; Yonggang Chen; Jianxin Li
- Abstract
An improved generative adversarial network model is adopted to improve the resolution of remote sensing images and the target detection algorithm for color remote sensing images. The main objective is to solve the problem of training super-resolution reconstruction algorithms and missing details in reconstructed images, aiming to achieve high-precision detection of medium and low-resolution color remote sensing targets. First, a lightweight image super-resolution reconstruction algorithm based on an improved generative adversarial network (GAN) is proposed. This algorithm combines the pixel attention mechanism and up-sampling method to restore image details. It further integrates edge-oriented convolution modules into traditional convolution to reduce model parameters and achieve better feature collection. Then, to further enhance the feature collection ability of the model, the YOLOv4 object detection algorithm is also improved. This is achieved by introducing the Focus structure into the backbone feature extraction network and integrating multi-layer separable convolutions to improve the feature extraction ability. The experimental results show that the improved target detection algorithm based on super resolution has a good detection effect on remote sensing image targets. It can effectively improve the detection accuracy of remote sensing images, and have a certain reference significance for the realization of small target detection in remote sensing images.
- Subjects
IMAGE reconstruction; REMOTE sensing; GENERATIVE adversarial networks; IMAGE reconstruction algorithms; OBJECT recognition (Computer vision); FEATURE extraction
- Publication
Journal of Measurements in Engineering, 2024, Vol 12, Issue 1, p83
- ISSN
2335-2124
- Publication type
Article
- DOI
10.21595/jme.2023.23510