An example of a project to evaluate the appearance of agri-food products through color image analysis

Abstract

The artificial vision is particularly well-adapted to product quality control in the agro-food industry, since it makes it possible to establish a quick, repeatable, objective, non-destructive measurement without contact with the product. Within this context we have worked on research project to assess aspect and quality of animal food products. This paper presents a project of the evaluation of aspect of turkey carcass sets in slaughter plants. After an analysis of industrial problems, the aims of the vision system were defined, the image acquisition system was devised and treatments were developped to detect and identify aspect defects on turkey carcass. These treatments are based on a segmentation method using multidimensionnal histogram of color images. The treatments and measurements were validated on industrial sets to assess the efficiency of the system in an industrial environment. This example shows the benefit of use of numerical vision in the case of complex objects with very unsaturated color defects and under hard industrial conditions and with high speed of plant.