Artificial intelligence in the modernization of the inspection process for the poultry industry

• Grant number: 22/08221-5
• Support Opportunities: Research Grants - Innovative Research in Small Business - PIPE
• Start date: December 01, 2022
• End date: November 30, 2024
• Field of knowledge: Agronomical Sciences - Veterinary Medicine - Products of Animal Origin Inspection
• Agreement: SEBRAE-SP
• Principal Investigator: Carolina Cesarino Coutinho
• Grantee: Carolina Cesarino Coutinho
• Company: Ecotrace Tecnologia da Informação S/A
• CNAE: Desenvolvimento de programas de computador sob encomenda ; Desenvolvimento e licenciamento de programas de computador customizáveis ; Consultoria em tecnologia da informação
• City: Vinhedo
• Pesquisadores principais: Diego Rafael Lucio
• Associated researchers: Eric Cristiano de Souza Luque ; Fernando Henrique Pimenta de Moura ; Rebeca Garcia Lemes
• Associated research grant(s): 22/15307-3 - Artificial intelligence in the modernization of the inspection process for the poultry industry, AP.PIPE
• Associated scholarship(s): 23/12375-0 - Research in Artificial Intelligence and Convolutional Neural Network Models, BP.TT; 23/02483-0 - Research in Artificial Intelligence and Convolutional Neural Network Models, BP.TT; 23/02974-4 - Artificial intelligence in the modernization of the inspection process for the poultry industry, BP.TT; 23/03863-1 - Artificial intelligence in the modernization of the inspection process for the poultry industry, BP.TT

Abstract

Nowadays, the verification of lesions in poultry carcasses is totally dependent on the macroscopic evaluation of a human. The data generated during this inspection are recorded manually, using an abacus, which makes the process susceptible to failures, increasing operating costs due to the low efficiency of the process.Post-mortem inspection of chicken carcasses is performed by an evaluator visually, that is, macroscopically. Knowing that human vision provides low resolution of details when an object is in motion in relation to its point of view, the low efficiency of the human evaluator is notorious. Another challenge is the recording of data from this activity. In practice, hand tools are used to record data on injury incidence and bird condemnation rates, such as the abacus and printed forms. This information is entered into the Brazilian System of Management Information of the Federal Inspection Service, since this process is manual, it provides, in addition to the possibility of errors in the data, an increase in the estimated time to perform the task. Noting that during the inspection, not all lesions present on the carcass are recorded, only those that stand out, thus evidencing a possible occurrence of under or overestimation of the markings of the lesions present in the carcasses and consequently the lack of accuracy in the data generated from this record.In order to ensure good management, the quality of chicken meat and the safety of food for consumers, it is necessary to insert a modern inspection system, based on accurate and reliable data, based on scientific knowledge. With the creation of new technologies in the slaughter lines, the federal inspection process will gain a great facilitator within the slaughterhouses.Given the above, the objective of this project is to use computer vision and artificial intelligence to identify lesions present in poultry carcasses (dermatitis, dermatitis, scratches, arthritis, dorsal and frontal hematoma, skin rupture, fracture and foot callus) in a automatic and assertive. (AU)