Non-Rigid Object Segmentation Based on Distance Transformation
As the meat intake increases, the demand for broiler poultry also increases every year. Broilers used for cooking are shipped before they are fully grown, and there is small weights difference depending on the kind of dish. Thus the breeder must check from time to time whether poultry are grown to the target weight. However, frequent contact reduces the growth and quality of poultry, and pollutants in enclosed spaces adversely affect the health of the breeder. Therefore, it is necessary to measure the weight without directly contacting with poultry. However, it is difficult to know the average weight of the poultry because there is a problem to segment each object due to the nature of the poultry which is always massed. In this paper1, we propose a non-rigid algorithm for segmenting gathered poultry. This algorithm detects the number of poultry by calculating the center of poultry through Mean-shift clustering, and it is expected to be used for poultry automation with accuracy of 91.17%.
Index Terms - Poultry automation, Object detection, Object segmentation, Mean-shift