Introduction
Food is closely related to human health and culture. Food-related
research has always been a current research hotspot. With the rapid
development of economy and society, people pay more attention to their
own health problems. Food quality has received widespread attention as
an important part of safeguarding human health. At the same time, the
rapid growth of the food trade has made high-quality food the basis for
success in a highly competitive market. Therefore, simple and effective
food quality assessment technology can be effectively applied in food
production, transportation, storage and other links. At present, food
quality evaluation still relies heavily on manual inspection, which is
cumbersome, laborious, and costly, resulting in subjective and
inconsistent evaluation results. In order to ensure the quality
inspection of food during production, transportation, and storage, and
to meet demands for different quality food, researchers have proposed a
variety of advanced strategies for food quality inspection, including
artificial intelligence methods [1, 2, 3], infrared spectroscopy and
biosensors [4, 5, 6, 7, 8]. However, a single biosensor can only
extract the content of specific components in food, and cannot
comprehensively evaluate food quality.
In recent years, artificial intelligence technology has been used in
different fields [9, 10, 11, 12], using artificial neural network to
accurately and quickly evaluate food quality. In addition to the
advantage in detection speed, artificial neural network can
automatically extract the intrinsic features of the target using its own
network structure. It constructs stable feature combinations through a
process of abstraction from low to high levels, which weakens the
subjectivity of manual feature selection and can save a lot of time and
workload. The generation of the predictive model is mainly dependent on
data collection and algorithm determination. In this method, food
quality data is collected from convolutional neural networks and
different types of biosensors. Instead of focusing on how feature
extraction is performed inside the neural network, we only need to use
the collected food quality data and food quality grade as the input and
output of the model, respectively. This approach is complex, but it is
also flexible. Although artificial intelligence technology can
automatically and accurately evaluate food quality through training, the
training process requires a large amount of food characteristic data.
Therefore, in order to take into account the detection speed and the
acquisition speed of food characteristic data, this paper proposes to
combine artificial neural network and biosensor to evaluate food
quality.
This article will show how to combine biosensors and deep learning to
assess food quality. First, we focus on the training of prediction
models and the required datasets. Secondly, we present representative
research work on convolutional neural networks and biosensors for
extracting food quality features. Finally, the paper summarizes and
discusses the challenges and possible solutions of the approach in the
field of food quality assessment.