Recognition of the storage life of mitten crab by a machine olfactory system with deep learning

Abstract In recent years, research on the qualitative analysis of aquatic product using ultrasonic or biological impedance has received increasing attention. Unfortunately, aquatic product safety incidents occur now and then, thus new kinds of analytical techniques that are fast, nondestructive, and simple to operate would be advantageous. This article addressed a machine olfactory system for recognition of the storage life of a whole living mitten crab. A type of deep learning algorithm, namely stacked denoising auto‐encoders algorithm (SdA), was applied to extract effective features of the machine olfactory system sensors' responses. The traditional feature extraction methods such as principal component analysis and linear discriminant analysis were used for the purposes of comparison. Then the support vector machine (SVM) classifier was implemented for further qualitative classification. In addition, the crabs were analyzed for total volatile basic nitrogen and total viable counts during storage process. Validation experiments showed that the highest recognition rate (96.67%) was achieved by using SdA with SVM. This study may present a promising instruction for the storage life recognition of the living crab. Practical Application Evaluation of the freshness of mitten crab is of considerable importance to guarantee mitten crab quality. However, to our best knowledge, few reports on the development of an electronic‐nose for nondestructive evaluation of the quality of whole living crab are available. A machine olfactory system for recognition of the storage life of a whole living mitten crab was present in this study. Validation experiments showed the highest recognition rate based on the system. In summary, the metal‐semiconductor sensor‐based machine olfactory system presents some advantages including fast detecting, repeatability, and nondestructive. In the future, with the improvement of sensor performance and development of pattern recognition, it will be a favorable weapon for protecting the interests of consumers in aquatic market.