22 June 2023 in IEEE Access
To improve the accuracy of feature extraction and description of various scales in traditional Oriented FAST and Rotated BRIEF (ORB) feature matching algorithm, this paper proposes an ORB feature matching algorithm based on multi-scale feature description fusion and feature point mapping error correction. Firstly, when establishing the feature scale pyramid, the method of using the same patch-size for description on each feature layer is adopted instead of using different patch-sizes on a unified feature layer in the original algorithm, which enhances the robustness of the descriptor and improves the matching accuracy. Secondly, FAST-SCORE maps are established on different scales, and the coordinates of high-level feature points mapped to the bottom layer are corrected to further improve the positioning accuracy of feature points. The algorithm is verified in remote sensing images, autonomous driving, and industrial automation fields. Experimental results show that when resisting theoretical interference, the average matching accuracy of the proposed algorithm is 67.9%, which is about 2.0 times that of the ORB algorithm, and the average stability is 14.0, which is about 1.5 times that of the ORB algorithm. After correcting the feature point mapping, the matching accuracy can be further improved by 19.2%, indicating that the improved algorithm has excellent robustness when resisting interference. In the experiments on the KITTI and custom datasets, the matching accuracy of the proposed algorithm reached 88.70% and 96.88%, respectively, which is an improvement of 10.15% and 1.2% compared to the ORB algorithm. At the same time, the matching time was reduced by 17.34% and 24.30%, ensuring the accuracy and real-time performance of the algorithm in practical scenarios.
27 June 2023 in IEEE Access
The Cobb angle is an important indicator for evaluating the severity of spinal curvature. Currently, there are various segmentation models used to segment the spine on X-ray images, but there is still room for improvement in segmentation performance. In addition, existing automatic methods for detecting the Cobb angle require manual cropping or bounding of the dataset, and cannot achieve fully automatic detection. Therefore, this paper proposes an automatic detection method based on RetinaNet and W-Net, which detects the ROI of the spine and performs segmentation to achieve automatic detection of the 17 required vertebrae. Experimental results show that the MIoU of W-Net is 0.9084±0.0032, the Dice Coefficient is 0.9442±0.0161, and the Precision is 0.9433±0.0176, which is better than other typical segmentation models. In addition, a feasible post-processing method is proposed to solve the problem of adjacent vertebrae adhesion in the segmentation image. The reliability of the Cobb angle automatic detection method was evaluated, and the intra-group correlation coefficients (ICC) with two observers were 0.902 and 0.915, respectively, both greater than 0.9, and the mean absolute deviation (MAD) was 3.08° and 2.91°, respectively, both less than 3.5°. Therefore, the proposed method can achieve automatic detection of the Cobb angle without manual cropping or additional human intervention, and has good reliability.