Segmentation Line Construction Method for Pig Carcass Musculature Based on Blurred and Distorted X-ray Images

What is it about?

The precise construction of the muscle-bone segmentation line is the basis for the on-demand boning of the pig carcass segmentation robot. However, pigs' skeletons include intricate, multi-layered structural elements that include bone block inlay, muscle attachment, and soft tissue attachment. In addition, there is noise and the X-ray images are blurry. This situation makes the images of pig bones acquired by X-ray produce blurring and distortion of geometric boundaries. Thus, it is not possible to accurately bone the pig carcasses on demand. To address the above difficulties, this paper proposes a segmentation line construction method for pig carcass musculature based on blurred and distorted X-ray images. Firstly, the blurred X-ray image features are extracted using a multi-scale feature pyramid feature extraction network. Secondly, the distorted image edge features are corrected by the feature contrast correction method. Finally, the method is used to segment the pig musculoskeletal X-ray image by example to obtain a more accurate location and extent of the pig bones, and the segmentation line is obtained using the segmentation contour information. The experimental results show that the experimental results of IoU, Pixel Precision (PP), and segmentation quality (SQ) on the X-ray image dataset are 76.42%, 85.68% and 91.43%, respectively, which are 2.60%, 2.31% and 1.86% better than the comparison methods.

Featured Image