2.1 Application and characteristics of Internet of things
The concept of logistics was introduced to China in the 1970s. At present, there are still many obstacles to the development of our logistics industry: the construction of logistics infrastructure is lagging behind, which is mainly manifested in the lack of hard technology construction and application, especially soft technology construction [
9]. Affected by the planned economy in the past, restricted by the acceptance system and traditional concepts, our logistics theory lacks corresponding innovations and breakthroughs, let alone scientific, systematic, and technological logistics theory [
10]. The managers of many logistics companies have insufficient market awareness in terms of ideology, market awareness and market awareness [
11]. Without systematic logistics knowledge, modern logistics theory cannot be fully utilized to realize the committee's traditional concept of emphasizing material production and ignoring logistics. We have insufficient understanding of the importance of cultivating modern logistics theory research personnel [
12]. Our logistics education level is relatively low, logistics talents are still in short supply, the integrated transportation system is imperfect, there is no seamless connection between different modes of transportation, the degree of specialization of the logistics industry is low, the degree of socialization is low, and the distribution of integrated logistics elements lacks coordination. The development of our logistics industry is also hindered by the lack of strong multinational companies [
13]. Intelligent logistics is at the top of the times, and it is also the first in the development of logistics industry,and faces more problems than this [
14]. However, the company urgently needs to develop intelligent logistics, which is not enough to ensure the healthy development of intelligent logistics [
15]. Perceived logistics refers to some emerging forms of logistics, such as two-dimensional logistics barcodes. It has been used in the logistics field for some time, including barcodes, and also applicable to product numbers. However, there are only a few types of agricultural product barcode technology applications in the logistics industry, and the promotion and application of agricultural products has a long history. The logistic burden of this farm is [
16]. The perceived logistics utilization rate of agricultural products can be calculated in Eqs.
1,
2,
3, and
4 according to the following formula.
$${\text{CV}} = \int {\mathop x\nolimits^{2} } nx + \frac{{x^{{n + 1}} }}{{n + 1}} + 1$$
(1)
$$R(x) = \frac{{f^{{(n + 1)}} \left( {\omega \mu \kappa } \right)}}{{\left( {n + 1} \right)!}}\left( {x - x_{0} } \right)^{{n + 1}}$$
(2)
$$C_{{{\text{n}} + {\text{1}}}} = \{ {\text{s}}_{i} |f(c) = C_{f} 1 + n < = i + 1 < = n,1 < = j < = n + 1$$
(3)
$$T = \sum\limits_{{J + 1}}^{{V - N}} {\frac{{D_{f} }}{D}} *{\text{Info(}}D_{f} ) + \sum\limits_{N}^{V} {D^{2} }$$
(4)
CV refers to the annual penetration rate of agricultural product logistics,
R(
x) perceives the utilization rate of logistics, in formula
3C represents the overall coordination rate during the operation of the internet of things, and
T in formula
4 represents the compatibility of the internet of things. In formula
1,
n represents the penetration rate calculation factor, and
D in formula
4 is a constant for adjusting the meter compatibility.
At the macro level, logistics is a cross regional and cross industry comprehensive system. The level of its standardization is directly related to whether the internal functions, elements and modules of the logistics system can be effectively connected and coordinated development, and to a certain extent, it determines the logistics efficiency of the whole society. At the micro level, logistics standards are the key supporting factors to ensure the coordination and unification of logistics activities and the close technical connection between logistics system and other systems [
17]. Only when the logistics standardization is realized can the management efficiency of the logistics system be improved, the connection with other systems be strengthened, the economic and social benefits of the logistics system can be effectively improved, the competitiveness of the logistics industry can be enhanced, and the development of intelligent logistics can be promoted [
18]. At present, China's logistics standardization work is relatively backward, resulting in poor compatibility of logistics facilities and equipment, low degree of convergence of logistics operations, and high efficiency of the overall operation of the logistics system [
19].
2.2 Application principle of agricultural product supply chain
It is generally believed that the entire Internet of Things can be divided into three basic levels: perception layer, transmission layer and receiving control layer. The perception layer mainly uses sensors to dynamically perceive the properties and changes of objects, and collects the perception status through radio frequency and other technologies. The transmission layer uses Internet technology to process the sensed data through a microprocessor to achieve long-distance transmission. The receiving control layer is the user side, which realizes the visualization of the object perception results and realizes the control of the perception objects and conditions [
20]. Agricultural product supply chain management is still a relatively new management concept and method in China. Its core is to emphasize the use of integrated ideas and concepts to guide the management behavior of each node in the supply chain, that is, to guide the operation of the entire supply chain based on consumer demand and the entire supply chain is managed as a system to improve the operating efficiency and economic benefits of the entire supply chain [
21]. Under this model, the node companies in the supply chain do not pursue their own profit maximization alone, but establish strategic partnerships, aiming at maximizing the interests of the entire supply chain, and use certain profit distribution mechanisms to make the economic efficiency of all trading partners in the supply chain has been improved. The impact of the Internet of things on the agricultural product supply chain is huge. The application of Internet of things technology can make the facilities, inventory, transportation, information and procurement involved in the agricultural product supply chain highly optimized; it can affect the production, transportation, and consumption of agricultural products. Real-time management of links can reduce supply chain costs and enable supply chain management to achieve a high degree of agility and complete integration [
22]. The Internet of things based on radio frequency/electronic product code technology has been deeply integrated into all aspects of agricultural product supply chain management, and has had a profound impact on the optimization of agricultural product supply chain [
23].
With the help of satellite communication system, the agriculture has established an agricultural information network center. On this basis, the provincial agricultural system website group and agricultural information professional website were established. The agricultural information sharing network has been established with the central government and local government to realize the agricultural information sharing with the central government and local government. Agricultural information platform integrates modern communication technology, computer network technology, information retrieval and push technology, and modern information management technology. It builds a comprehensive agricultural information service platform push, technology matching and e-commerce, which integrates hotline. SMS interaction, intelligent retrieval and informatization. With the development of the Internet of things, the internet of things technology is gradually used in the agricultural field, forming the agricultural Internet of things [
24]. The multi-scale transmission of agricultural information is realized through wireless sensor network, telecommunication network and internet. The massive agricultural information obtained is fused and processed, and agricultural monitoring, scientific management and instant service are realized through intelligent operation terminal. Using summation formula and polynomial, we can calculate the new growth probability and the new development point of enterprises brought by the Internet of things [
25]. Through the Internet of things, we can get the real popular and profitable product model of enterprises. We can use big data technology to accurately grasp the psychology of consumers and expand our own advantages. The commonly used formulas are 5,6,7.
$$F(a) = \left( {\frac{{a - 1}}{{{\text{stebucstion}}}}} \right)^{{{\text{a}}*t}} + \left( {\frac{{a + 1}}{{bxvsstn}}} \right)^{{a*t}}$$
(5)
$${\text{GH}} = \frac{{\left| {Ax_{0} + By_{0} + C_{0} + \left. D \right|} \right.}}{{\sqrt {\mathop A\nolimits^{2} + B^{2} + C^{2} } }}$$
(6)
$$E(L) = SL + \sqrt {\frac{{\sum {\left[ {S\left( a \right) - S\left( b \right)} \right]} }}{{a - b}}^{2} }$$
(7)
Among them, F(a) represents the new growth rate of agricultural products sales brought by the internet of things, the loss is the total savings of agricultural resources. GH is the profit margin of the enterprise brought about by the interconnection of all things, and E(L) represents the internet of things to save agricultural products. The amount of wasted resources. In the formula, a represents the operating coefficient of the internet of things, and b represents the growth rate operator.
There are many links in the supply chain of agricultural products, including rice planting, agricultural products processing, finished products distribution, agricultural products consumption and other basic links, as well as warehousing, transportation, loading and unloading and other logistics activities, which run through the internal links of the supply chain and the upstream and downstream circulation links. According to the current situation of production, circulation and consumption of agricultural products in China, combined with field investigation, relevant data of agricultural products planting, harvesting, rice milling, processing, detection, distribution, transportation and sales are collected, and various data are recorded with organic RFID tags, which are uploaded to the system data center layer by layer, so as to realize the tracking of agricultural products supply chain nodes, including all links and references in the whole process of agricultural products. Traceability with the unit, as well as the key steps and key processes of each link of specific batches of agricultural products, to ensure the traceability management of agricultural products supply chain. According to the different factory numbers of agricultural products, we can calculate the transportation time, growth cycle and sales volume of agricultural products by formula
8 and
9.
$$V = \frac{1}{K}\sum\limits_{{j = 0}}^{n} {\sum\limits_{{i = 0}}^{n} {\left( {M_{{\left( {i,j} \right)}} - u} \right)^{2} ,\quad {\text{if}}\,M_{{\left( {i,j} \right)}} } } \ne 0$$
(8)
$$T(s) = \sum\limits_{{j = 1}}^{v} {\frac{{M_{f} }}{D}} *\log 2\left( {\frac{{M_{f} }}{D}} \right) + J^{V}$$
(9)
where
V is the transportation time and shelf life of agricultural products.
M is the minimum savings rate of agricultural products, and
T(
s) is the transportation time of agricultural products.
Where
D is the transportation time and shelf life of agricultural products, min is the minimum saving rate of agricultural products, and
t(
s) is the transportation time of agricultural products. The distribution channels of fruits, vegetables and agricultural products are complex and diverse. As far as the main body of fruits and vegetables and agricultural products are concerned, one is farmers who organize production and operation activities as a family unit, and the other is a large-scale and specialized production base. Therefore, in order to prevent certain diseases or problems of agricultural products, it is necessary to trace the source of the disease [
26]. The main tracking objects include farmers/production bases, wholesale companies at all levels, logistics supply companies and sales companies in the fruit and vegetable supply chain. The scope of corporate traceability is generally divided into internal traceability and external traceability [
27]. Internal traceability emphasizes the traceability of corporate information, such as vegetable packaging, cleaning and segmentation, operator information, internal environmental information, external traceability is mainly to trace the circulation information of fruits and vegetables in the supply chain. When there is a problem in any link of the fruit and vegetable supply chain, the company can trace the origin and processing history information of the fruit and vegetable through the traceability system to analyze the cause of the quality problem [
28]. For products that have already circulated to the next link or entered the market, the product range can be locked in time and customers can be recalled.
On the other hand, whether it is pilot projects, demonstration first, and supporting enterprises' gradual development model. It is still a system of division of labor, and the development model of all links going hand in hand, which provides ideas for the development of intelligent logistics and reduces the blindness of development. Enterprises and industries should focus on the innovation of smart logistics development concepts and actively explore. Through a large amount of relevant information, the prospects and existing problems of the agricultural product perception supply chain under the background of the internet of things are analyzed, and the problems that are conducive to the better development of the agricultural product perception supply chain are obtained. Data analysis is carried out on the investigation and research of the IoT-aware supply chain, and the data analysis adopts DCM technology. At the same time, through questionnaire surveys and model construction, relevant conclusions are drawn, through various data comparisons and analysis, through the presentation of data, to more intuitively understand the impact of the internet of things on the supply of agricultural products.
2.3 Relevant vector machines in the context of the Internet of things
In all recognition systems, machine learning mainly solves the problem of classification, inferring a complex and reasonable mapping criterion based on the basic information of the recognized object and the category of the recognized object, so as to predict the type of the object. Common machine learning algorithms include: support vector machines, neural networks, correlation vector machines, and fuzzy recognition methods, etc. These methods all have good results in specific applications. This article mainly uses multi-source data to jointly identify agricultural products in the supply chain, and the collected agricultural product data samples are small. The Relevance Vector Machine (RVM) learning method adopts the Bayesian method, introducing the prior of weights, and the weights are assigned one by one by hyperparameters, and their values are calculated through repeated iterations of data.
RVM solves the weight of the correlation vector by maximizing the posterior probability (MAP). For a given training sample set:
$$\{ x_{i} ,t_{i} \} \{ i = 1,2,...,a\} ,x_{i} \in R^{d} ,t_{i} \in R$$
(10)
$$x_{i} = \{ x_{{i1}} ,x_{{i2}} ,..,x_{{ib}} \}$$
(11)
\(x_{{ij}}\) represents the
j-th feature of the
i-th sample, a is the number of samples for training, and
b is the number of sample features), the model function of RVM is:
$$y(x,u) = \sum\limits_{{i = 1}}^{n} {u_{i} k(x,x_{i} ) + u_{0} }$$
(12)
where
\(u_{i}\) is the correlation weight,
\(k(x,x_{i} )\) is the kernel function, assuming that the target has a noise
\(\vartheta _{i}\) that obeys the expectation of 0 and the variance
\(\sigma ^{2}\) is Gaussian distribution.
$$t_{i} = y(x_{i} ,u) + \vartheta _{i}$$
(13)
Therefore, for a given sample
\(x_{i}\), the probability of belonging to
\(t_{i}\) is:
$$p(t_{i} |x_{i} ) = n(t_{i} |y(x_{i} ,u),\sigma ^{2} )$$
(14)
Then the likelihood function of the training data is:
$$p(t|u,\sigma ^{2} ) = (2\pi \sigma ^{2} )^{{ - \frac{n}{2}}} \exp \left\{ { - \frac{{||t - ou||^{2} }}{{2\sigma ^{2} }}} \right\}$$
(15)
among them:
$$t = (t_{1} ,t_{2} ,...,t_{n} )^{T} ,u = (u_{1} ,u_{2} ,...,u_{n} )^{T}$$
(16)
\(o\) is an
N*(
N + 1)-dimensional high-dimensional structural matrix composed of multiple kernel functions, and its expression is:
$$o = [o(x_{1} ),o(x_{2} ),...,o(x_{n} )]^{T}$$
(17)
Each element corresponds to:
$$o(x_{i} ) = [1,k(x_{i} ,x_{1} ),k(x_{i} ,x_{2} ),...,k(x_{i} ,x_{n} )]^{T}$$
(18)
The weight
u in the Yees method is estimated according to the maximum likelihood function, but it is prone to over-learning. In order to avoid this problem, a conditional probability distribution function is set to constrain u, so the prior probability distribution function of w is as follows:
$$p(u|a) = \prod\limits_{{i = 0}}^{n} {n(u_{i} |0,a_{i}^{{ - 1}} )}$$
(19)
a is a hyperparameter vector, which controls the deviation degree of
u. The likelihood distribution of the output can be obtained by integrating the weights, namely:
$$p(t|a,\sigma ^{2} ) = \int {p(t|u,\sigma ^{2} )} p(u|a){\text{d}}w$$
(20)
According to Bayes' criterion, the posterior probability expression of w is:
$$p(t|a,\sigma ^{2} ) = \frac{{p(t|u,\sigma ^{2} )p(u|a)}}{{p(t|a,\sigma ^{2} )}}$$
(21)
In order to maximize it, the derivative of Eq. (
20) is obtained:
$$a_{i}^{{{\text{update}}}} = \frac{{1 - a_{i} \sum {_{{ii}} } }}{{v_{i}^{2} }}$$
(22)
$$(\sigma ^{2} )^{{{\text{update}}}} = \frac{{||t - ov||^{2} }}{{n - \sum {_{{ii}} (1 - a_{i} \sum {_{{ii}} } )} }}$$
(23)
According to formula (
20) and formula (
21), iteratively update
a and
\(\sigma ^{2}\) until the preset convergence condition (parameter change range is small or reaches the set number of iterations) is satisfied.