The Development of a Weight Prediction System for Pigs Using Raspberry Pi

• Published in: Agriculture (Basel) Vol. 13; no. 10; p. 2027
• Main Authors: Cho, Wan Hyun, Kim, Sang Kyoon
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 01.10.2023
• Subjects: Age; Agricultural economics; Agriculture; Algorithms; Animal sciences; Animals; Artificial intelligence; Automation; Behavior; body and shape feature extraction; Cattle; computer vision techniques; Data capture; Data entry; Economic impact; Farmers; Farms; Feeds; Hogs; Livestock; Livestock farming; Livestock farms; livestock measurement device; Methods; Neural networks; Performance evaluation; pig segmentation; prediction of pig weight; Predictions; RGB-D sensor data; Swine; Vision systems; Weight; Zoology
• ISSN: 2077-0472; 2077-0472
• DOI: 10.3390/agriculture13102027

Generally, measuring the weight of livestock is difficult; it is time consuming, inconvenient, and stressful for both livestock farms and livestock to be measured. Therefore, these problems must be resolved to boost convenience and reduce economic costs. In this study, we develop a portable prediction system that can automatically predict the weights of pigs, which are commonly used for consumption among livestock, using Raspberry Pi. The proposed system consists of three parts: pig image data capture, pig weight prediction, and the visualization of the predicted results. First, the pig image data are captured using a three-dimensional depth camera. Second, the pig weight is predicted by segmenting the livestock from the input image using the Raspberry Pi module and extracting features from the segmented image. Third, a 10.1-inch monitor is used to visually show the predicted results. To evaluate the performance of the constructed prediction device, the device is learned using the 3D sensor dataset collected from specific breeding farms, and the efficiency of the system is evaluated using separate verification data. The evaluation results show that the proposed device achieves approximately 10.702 for RMSE, 8.348 for MAPE, and 0.146 for MASE predictive power.