Research Article
Image Recognition Technology of UAV Tracking Navigation Path Based on ResNet
@INPROCEEDINGS{10.1007/978-3-031-50574-4_3, author={Lulu Liu and Degao Li and Junqiang Jiang and Shibai Jiang and Linan Yang and Xinyue Chen}, title={Image Recognition Technology of UAV Tracking Navigation Path Based on ResNet}, proceedings={Multimedia Technology and Enhanced Learning. 5th EAI International Conference, ICMTEL 2023, Leicester, UK, April 28-29, 2023, Proceedings, Part II}, proceedings_a={ICMTEL PART 2}, year={2024}, month={2}, keywords={ResNet UAV Tracking Navigation Path Image Recognition}, doi={10.1007/978-3-031-50574-4_3} }- Lulu Liu
Degao Li
Junqiang Jiang
Shibai Jiang
Linan Yang
Xinyue Chen
Year: 2024
Image Recognition Technology of UAV Tracking Navigation Path Based on ResNet
ICMTEL PART 2
Springer
DOI: 10.1007/978-3-031-50574-4_3
Abstract
The image of UAV tracking navigation path is easily affected by noise, and there are problems of low image recognition accuracy and poor image enhancement effect. Therefore, a ResNet-based image recognition technology for UAV tracking navigation path is proposed. Extract image features and analyze image color and shape. The Laplacian operator is used to process the image to enhance the edge of the image. Using the bilinear interpolation method, the image is scaled and grayscale is processed, and the noise is processed by combining the wavelet transform. Build a ResNet-based recognition model, use a multi-resolution octree hierarchical structure, render each node, and output any node image coordinates. Perform global pooling on the input feature map to improve image degradation. The gradient image is binarized using the binarization method. Fully consider the characteristics of the UAV tracking and navigation path, and use the statistical averaging method to obtain the average interference amplitude and phase, and calculate the interference characteristic distance. The iterative threshold selection method is used to obtain image recognition results. It can be seen from the experimental results that this technology can extract comprehensive image information, and has a high signal-to-noise ratio, which can achieve the purpose of image enhancement, and the highest image recognition accuracy obtained is 0.96, with accurate recognition results.
