The feasibility of using machine vision technology to estimate the weight of broiler chickens

This research was conducted to investigate the feasibility of estimating the weight of broiler chicks using machine vision technology. 600 Ross broiler chicks were reared under standard conditions for a 42-day period. At selected intervals (7 days), 60 birds from a total of 600 chicks were randomly selected and weighed individually using the appropriate scale. At the same times, digital images were captured individually and in groups 2, 3 and 4 of birds. The digital images were initially preprocessed and the necessary changes were made on the photos and required features were extracted from images by designing an appropriate algorithm, and these features were used to design the neural network to estimate the body weight of chicks. The correlation coefficient between the extracted features of digital images including the Major axis length, Minor axis length, Bonding box, Convex Area, Filled area, Perimeter and Area of the image with live weight of the chicks were 0/92, 0/93, 0/53, 0/99, 0/99, 0/94, and 0/99 respectively (p <0.01). A Multilayer perceptron neural network, which was trained with back propagation learning algorithm, containing 22 neurons in the input layer, 20 neurons in the mid layer and one neuron in the output layer presented the highest accuracy(99%) to estimate the weight of broiler chicks at different ages. The results of this study showed that there is a possibility of using image processing and artificial neural network as an appropriate and efficient tool to estimate the weight of broiler chicks during the breeding period.