Collecting external data
The training of the predictive model requires the acquisition of a large
amount of data, so the speed of data collection needs to be guaranteed.
If the external quality characteristics of food are extracted in an
artificial way, it will undoubtedly consume a lot of time. As a branch
of deep learning, computer vision [13, 14, 15, 16, 17, 18, 19] has
been widely used in food classification and feature extraction. For
every food features are figured out for which the quality is either
directly or inversely varied.
Computer vision technology is a branch of artificial intelligence that
aims to eventually replace the human visual decision-making process with
automated programs. Computer vision technology is a mechanism that
artificially simulates the human thinking process. Through continuous
learning, it can make accurate, fast and complex judgments. At present,
the more mature solution in computer vision technology is CNN
(convolutional neural network). The difference from ANN is that CNN
introduces the operation of convolution. CNN are mainly composed of
these types of layers: input layer, convolutional layer, pooling layer
and fully connected layer. The core part is the convolution layer and
the pooling layer. The function of the convolutional layer is to extract
the information in the input image, which is called image features.
These features are reflected by each pixel in the image in a combined or
independent way, such as texture features and color features of the
image. The function of the convolution layer is to perform convolution
operations. The convolution operation is to perform the
cross-correlation operation from left to right and from top to bottom
through the convolution check of the matrix of each channel, and slide
from the upper left corner to the lower right corner step by step. The
sliding step size is a hyper parameter. It means that the corresponding
positions are multiplied and then added, and finally the values of the
three channels are also added together to obtain a value. The role of
the pooling layer is to select the features extracted from the
convolutional layer. The basic structure of the convolutional layer and
the pooling layer is shown in Figure 1(B)
The convolutional neural network can not only improve the detection
speed of acquiring the external features of food, but also realize
automatic detection based on this technology. Therefore, convolutional
neural network becomes a new solution for food external feature
extraction