SIVA: Automatic Visual Inspection System Applied to Egg Quality Control in Poultry Industries

This work addresses the problem of automatic control of egg quality in egg processing plant, by using image analysis techniques. In recent years, the increasing throughput of modern egg grading machines, which frequently grade up to 100.000 eggs per hour, has become the visual inspection of eggs by humans a critical bottleneck in the egg sorting chain. In this scenario, to assure a high and consistent egg quality, novel sensor-based technologies have been developed. Unlike most existing methods, which are frequently based on the use of mechanical techniques or spectroscopic principles (e.g., near-infrared, mid-infrared and fluorescence spectroscopies), this work proposes an automatic visual inspection system based on computer vision algorithms to detect, track, count and classify eggs in image sequences captured by stationary cameras installed in the industrial production line. The vision system was designed to classify different levels of very common eggshell defects (gross cracks, hairline cracks, cage marks and stained eggs) as well as blood spots inside the eggs. The system was successfully validated in a prototype of an industrial production line, achieving an egg counting accuracy of 100 percent and a classification accuracy of eggshell defects as high as 78.5 percent.

Publications

See also

Prof. Flávio Cardeal.

Douglas Machado.

Prof. José Fernandes.