M. S. Thesis Abstract
Model-Based Vision-Guided Automated Cutting of Natural Products
Melissa C. Sandlin
Often in industry, objects are fixtured rigidly so that processes can be performed from a "home" position. In addition, the geometry and structure of the object is known to a high level of accuracy. However, in the areas of inspection and sensing for motion control as they apply to the medical and food processing industries, operations are directly dependent on the internal structure and surface features of the object of interest. Also, these objects vary in size and orientation and cannot be fixtured in a traditional manner. Therefore, it is desired to develop a procedure to correlate information between the internal structure of the object and the surface data or externally observable features.
One very labor-intensive process in the poultry industry is the deboning process. There have been several efforts to improve the deboning process. Empirical evidence shows that the initial cut, currently performed manually, is the most critical step in this process. Observation of poultry industry workers indicates that the initial cut requires good coordination and wrist flexibility as well as some prior knowledge of the internal structure. Although there are several existing techniques for the deboning of meat products, the case for poultry is unique in that the product surface is flexible and there is more variation in size of products. This thesis will develop a Model-Based Vision-Guided Method which uses calibration and transformation techniques on x-rays and digital images to obtain a template/model which accurately locates the shoulder joint with respect to the surface features of the product. Consequently, a cutting path trajectory which an automated instrument can follow will be determined and added to the model.
This research seeks to improve the existing deboning process through the application of machine vision. At the same time general principles will be developed for use in the design of systems to ensure wider applicability. This should benefit similar operations in a broader range of industries. These principles will be particularly applicable in areas such as medicine as well as other food industries. For future research in poultry deboning, an automated system should be designed that utilizes the template by comparing it to images of each product on the line and determining a corresponding cutting path based on scaling and orientation of the template.