Classification of hops image based on ResNet-ConvLSTM and research of intelligent liquor picking system

Aiming at the problem that the liquor picking in traditional brewing technology of Chinese liquor is closely dependent on manual operation, and the existing alcohol content measurement and automatic equipment have a high cost, low detection accuracy, and inaccurate classification, an intelligent liquor picking system based on hops image classification is proposed. First, an industrial camera was used to collect hops sequence images. By labeling the alcohol content and preprocessing the images, a data set of 11 categories of liquor hops images was established. Secondly, an end-to-end network model was established by combining the ResNet convolutional neural network and the ConvLSTM recurrent neural network. The algorithm performance was evaluated and verified on the established hops data set, and the model accuracy reached 96.97%. Finally, an intelligent liquor picking system was built to verify the feasibility of segmented liquor picking using hops images.