An Image-Assisted Rod-Platform Weighing System for Weight Information Sampling of Broilers

摘要

The average weight and flock uniformity of broilers in group housing is important information that allows producers to know the flock growth conditions and determine the selling time. However, gathering weight information of chickens is not only labor-intensive for humans but also frightening for the birds. In this study, an image-assisted rod-platform weighing system was developed to automatically monitor the average weight and flock uniformity of broilers in chicken houses. This weighing system consists of a computer and several weighing scales. Each weighing scale contains a rod-platform weighing module and a surveillance camera module. The principle of the automated weighing system is to estimate population weight information using samples. Thedesign of the rod-platform weighing module was based on the perching habit of birds to attract more broilers to stand on the rod platform and thus get more weight samples. The surveillance camera module is used to detect the number of broilers on the rodusing image processing technology. A data processing method called PORW1, which includes elimination of redundant records and trim of non-redundant records, was designed to optimize the results of chicken number identification from images to improve theaccuracy of the results. An experiment was done in two small groups of broilers with approximately 100 chickens and 8.58 m2 of area for each group. A weekly weighing was conducted, and three kinds of weight information were obtained, which included manual population weight information (MPWI), manual sample-based weight information (MSWI), and automated sampling weight information (ASWI). Each weight information set comprised the group average weight and flock uniformity, which were then used to evaluate accuracy. The perching rate of chickens using the rod platform reached an average of 60 times h~', and the rate was retained with increasing age. Compared with the MPWI obtained by individual weighing, the manual sample-based measurement method provided results with errors of 0% to +5%, while our automated weighing system achieved accuracies within ±2% for average weight and ±1.5% for flock uniformity.