Development of Portable Device for Non-Destructive Testing of Livestock Meat Quality based on Machine Vision

A portable non-destructive testing device for meat quality is designed, including the design of the lower computer and the design of the upper computer. The lower computer design involves a single chip control unit, a light source unit (LED light source), an image acquisition unit, data processing and communication (CCD binocular camera). The upper computer uses Anaconda Prompt software and TensorFlow framework to establish a visualization program, which is connected to the lower computer through serial communication to display the gas sensor data collection results, the meat quality judgment results, store the collected images and other data. A training data set is established, and the meat quality identification model is trained based on the ResNet network framework, and the model identification accuracy rate reaches 96%.