Research Article Intelligent Management of Land Resources Based on Internet of Things and GIS Technology

In order to improve the e ﬀ ect of land resource management, this paper combines the Internet of Things technology and GIS technology to build an intelligent management system for gradient resources to improve the e ﬃ ciency of land resource management. Aiming at the hybrid intelligent model of wetland resource remote sensing monitoring technology, this paper analyzes and studies the remote sensing image processing theory. Moreover, this paper studies in detail remote sensing image restoration, TM image re ﬂ ectivity simulation imaging, image enhancement technology, optimal band selection based on the characteristics of wetland resources, expert decision analysis, deep mining of image data, knowledge reasoning, and decision tree analysis to form a theoretical support system for a hybrid intelligent classi ﬁ cation model for wetland resources. The research shows that the intelligent management system of land resources based on the Internet of Things and GIS technology has a good e ﬀ ect in the collection and processing of land resource information and can e ﬀ ectively improve the management e ﬃ ciency of land resources.


Introduction
China has a vast territory and abundant resources in absolute quantity, but the per capita quantity is insufficient. Moreover, the natural resources, ecological and environmental conditions, and social and economic development vary greatly from place to place, and the stages of development and the problems existing in the development are also different. If we do not consider the characteristics and differences of regions and use the same standard to measure the development of various regions, it will be difficult for the evaluation to be objective and fair. At the same time, when the same policy is used to guide the development of various regions, the measures are often lacking in pertinence and effectiveness. In addition, with the rapid increase of the population, on the one hand, the area of arable land in China has dropped sharply, the land has been degraded, and the land reserve resources are seriously insufficient. On the other hand, extensive land use, low utilization rate, and low output rate make the situation of land resources increasingly severe and the contradiction between man and land increasingly prominent.
The research on comprehensive land carrying capacity has its own characteristics: first, the research on comprehensive land carrying capacity is a systematic perspective on regional land, grain, population, and social development. It involves many aspects and is affected by many factors, of which nature and humanity are the two most important ones. Natural factors include the quantity and quality of land resources, regional climatic conditions and water resources, etc.; human factors include population status, the level of social and economic development, the rationality of technological level resource utilization, etc. These factors together affect the level of carrying capacity of land resources makes it an intricate system. Secondly, while the research on comprehensive land carrying capacity has been improved with the development of science and technology and some natural processes, it does not mean that it can develop permanently and unrestrictedly. It is constrained by two aspects: the absolute limit of the productivity of land resources and the relative limit over a certain period of time. The former refers to the ultimate source of productivity of land resources-solar radiation energy is limited per unit area; especially, the conversion rate of solar energy cannot exceed a certain limit, thus stipulating the limit of productivity of land resources. In the foreseeable future, it is impossible for the development of science and technology to break through the limit of land productivity, such as the level of grain yield per unit. The limit of land production capacity determines the limit of land carrying capacity, which has far-reaching significance for objectively grasping the regional resource-population problem.
The "comprehensive" study of comprehensive land carrying capacity includes two meanings [1]. On the one hand, for the carrier, the land should not be limited to cultivated land but should be a generalized land including garden land, forest land, pasture land, urban settlements and industrial and mining land, water area, transportation land, and unused land. On the other hand is the question of what to carry, that is, carrying substance. Land problems are caused by human social and economic activities. The goal of land use is to coordinate human social and economic activities with the corresponding environment and to protect and improve the land resources for human survival and development. Therefore, the bearing object should be various social and economic activities of human beings, such as the city scale, economic output value, transportation scale, land pollution capacity, etc., not only the bearing population size and population consumption pressure, which forms the "comprehensive carrying capacity of land." The comprehensive carrying capacity of land refers to the maximum population and urban development scale that the land resources can bear on the basis of the foreseeable level of technological, economic and social development, the principle of sustainable development, and the premise of maintaining the benign development of the human ecological environment under the current development stage [2]. That is to say, the size of the comprehensive land carrying capacity is not only a reflection of the characteristics of the natural geographical environment but also depends on the development level of human society, economy, and technology; the effective use of land resources by humans; and the improvement of the ecological environment [3].
This paper combines the Internet of Things technology and GIS technology to build an intelligent management system for gradient resources, improve the efficiency of land resource management, and effectively promote the progress and development of intelligent land management technology.

Related Work
Literature [4] establishes a real estate database management system, which integrates the decentralized management of land registers and other information, which greatly improves work efficiency; in the application research of basic farmland management information system, literature [5] uses MapGIS secondary. The development library SDK builds a permanent basic farmland database and uses C++ language to design and implement a basic farmland demarcation system. Based on the MapGISK9 platform, Yang Weibin of Xiamen University adopts a combination of C/S and B/S architecture to solve the possible defects of the basic farmland management information system. Literature [6] developed the basic farmland management information system of Ninghua County based on ArcGIS Engine components and C# as the development language. Literature [7] used Oracle as the database management platform and ArcSDE as the spatial database search engine, participated in the design and establishment of the basic farmland management information system in Sichuan Province, and used ArcGIS Server server and WebService technology to enable the public to conduct related business queries on the browser side. These attempts in basic farmland information management have provided effective advanced experience for the system construction of farmland protection [8]. However, a large part of the region has not established a comprehensive and effective basic farmland information system, and we still need to explore and work together [9]. The use of UML language and CASE tools is the trend of GIS software development [10]. Literature [11] studies and summarizes the development process of cadastral information system by adopting the object-oriented GIS development method and combining with UML. Literature [12] describes the land change process realized by drawing lines with the mouse using UML and realizes the land change subsystem in the land management information system by means of secondary development. These studies reflect the application of UML technology in land information management software development from different angles and meet the needs of software system development in GIS projects. On the basis of the above practice, further research is carried out in combination with the characteristics of permanent basic farmland data and applications. The theoretical technology of geographic information system has gone through the process of desktop geographic information system, client/server mode in local area network, and WebGIS [13]. At present, the land information system is still in the stage of continuous development and generally shows the following development trends: (1) The integration with remote sensing system (RS), global positioning system (GPS), and other technologies needs to be strengthened. Surveying and mapping technologies such as RS and GPS can provide strong technical support for the real-time collection of large-area land use information and achieve high precision and automation of ground data collection [14]. (2) Modern land management information systems not only care about the distribution area, and it is a research direction of land management information system [15]. (3) With the help of artificial intelligence theory and expert system technology enables computers to use expert knowledge to simulate human brain thinking for reasoning, which can greatly improve work efficiency, and its applications in cartography, spatial decision support, and intelligent data processing need to be further deepened [16]. (4) With the advent of the era of big data and the rapid development of network technology, the use of the Internet to publish spatial data and provide users with the functions of spatial data browsing, query, and making thematic maps and analysis has become an inevitable trend in the development of land information systems [17]. Literature [18] proposes to build a "land and resources cloud" 2 Journal of Sensors technology system and create an "Internet + land and resources service" model, which is to make full use of advanced concepts and technologies such as cloud computing and big data to achieve unified deployment of services and comprehensive data sharing.

Remote Sensing Image Processing Based on Internet of Things and GIS
After the GCP ground control point selection is completed, the pixel coordinates (x, y) should be converted into image coordinates that are related to the reference coordinates (X, Y) through a mathematical model. The selected coordinate change function (mathematical correction model) is a polynomial correction model, and commonly used image correction functions include polynomial, Legendre polynomial, and collinear correction. Because the polynomial principle is intuitive, the calculation is simple, and it has better accuracy; the polynomial correction method is generally chosen. In this study, a quadratic polynomial function is used to perform precise geometric correction. The formula is as follows: Polynomial coefficients (conversion parameters) are obtained by least squares regression. According to the root mean square error formula (RMS error ), the difference between the estimated coordinates and the original coordinates is obtained.
That is, by calculating the root mean square error of each control point, the ground control points with larger errors can be checked, and the accumulated overall root mean square error can be obtained, so as to obtain the accuracy of coordinate change.
The absorptivity and reflectivity of an object vary with the temperature of the object and the wavelength of incident radiant energy, so we define the monochromatic reflectance and monochromatic absorptivity of an object as [19]: In the formula, aðλÞ is the monochromatic absorptivity of light with wavelength λ, ρðλÞ is the monochromatic reflectance of light with wavelength λ, E a ðλÞ is the energy absorbed by light with wavelength λ incident on the surface of an opaque object, E r ðλÞ is the energy reflected by the light with wavelength λ incident on the surface of an opaque object, and E is the total incident energy.
We use the TM data to make the principle of simulated reflectivity image, and the energy carried by the color light of different bands and the weightW i of the total energy are shown in the table. From this, we can obtain the total reflected energy: In the formula,Eρis the total reflected energy,K i is the weight of thei-th band in the total energy,TM l is the remote sensing data of thei-th band,W 7 is the weight of the seventh band in the total energy, andTM 7 is the remote sensing data of the seventh band.
From Wien's displacement theorem λ max = 2897/T, it can be known that with the increase of temperature, the peak value of radiant emission shifts to the short-wave direction. When the surface temperature is -60°C, the wavelength peak of the surface radiation is 13:6 μm, and when the surface temperature is 60°C, the wavelength peak of the surface radiation is 8:7 μm. The surface temperature is generally between -60°C and 60°C, and the radiated wavelength peak falls within the wavelength range of TM6. Therefore, the relative value of the energy lost by the earth due to long-wave radiation can be represented by TM6.
The surface of the earth with the size of each pixel receives the light energy of the sun as E 0 , then: The energy change of a pixel on the earth's surface (expressed as a temperature difference change) is E Δ , and the scattered energy is E e , then: That is, Since the relative value of the energy lost by the earth due to long-wave radiation can be represented by TM 6 , formula (7) can be written as follows: We set: In a very small time, the temperature difference change of the surface can be ignored; that is, E Δ = 0, and formula (9) can be transformed into: Since E Δ is not strictly equal to 0, we call P sin as the monochromatic reflectance. Through the different energy values of different bands, the monochromatic reflectance data of each band in the TM band can be obtained, and the synthesized color map is called the simulated reflectance map.
The first step in band combination is to select a band by numerical evaluation according to the amount of information contained in each band of a remote sensing image. Through analysis, it can be determined which parts or which bands (i.e., band subsets) contain the amount of information. The standard deviation of each band reflects the total dispersion of the gray value and the average value of each pixel in the image and to a certain extent reflects the amount of information in each band. The larger the value, the greater the amount of information contained. The amount of ground object information contained in each band of a TM image is generally measured by the radiation quantization level covered by the band image, that is, the range of brightness values or brightness difference (maximum brightness value-minimum brightness value).
The standard deviation formula is An image is A ij , ði = 1, 2, 3⋯⋯;j = 1, 2, 3⋯⋯Þ, its image matrix size is M × N, and A o represents the average gray value of the whole image. The brightness difference reflects the degree of change in the gray value, and its magnitude is equal to the maximum brightness value minus the minimum brightness value, and the formula is f rangeði, jÞ = f max ði, jÞ − f min ði, jÞ. The mean vector represents the average reflection intensity of the objects in the image.
The correlation coefficient between the bands reflects the degree of information overlap between the two bands. If the correlation coefficient between the two bands is large, it means that their information overlap is high. Therefore, two channels with high correlation can be combined into one or one of the channels can be taken as one. The formula for the correlation coefficient between the bands is as follows: In the formula, S ij = COVði, jÞ = ð1/NÞ∑ n k=1 ðXik − XiÞ ðY jk Y j Þ.
Among them,i, jis the number of bands 1, 2, 3, 4, 5, and 7, andS ij is the covariance between thei-th band and thej-th band, xiis the spectral gray mean value of thei-th band, xis the gray value of thek-th pixel in thei-th band, andY jk is the gray-scale value of thek-th pixel in thej-th band.
Y j is the spectral gray mean value of the j-th band; k = 1, 2, 3⋯⋯,n is the number of pixels in the experimental sample area. The larger the R, the greater the information overlap between the bands. The calculation of the band correlation coefficient matrix is relatively simple, and the general image processing software has this function. The analysis of the correlation coefficient becomes the basis of the band selection, and the calculation of the correlation coefficient is also necessary for the optimal index method.
Because the larger the standard deviation of the image data, the larger the amount of information it contains, and the smaller the correlation coefficient between the bands, which indicates that the independence of the image data in each band is higher and the information redundancy is smaller. Therefore, the concept of the optimal index (OIF) proposed by Chavez in the United States can also be used.
Among them,S i is the standard deviation of thei-th band, andS ij is the correlation coefficient of theiandjbands. The larger the OIF, the greater the information content of the corresponding combined image. The OIF values are arranged from large to small, and the band corresponding to the largest OIF value is the best band combination. This method is currently the most commonly used band selection method. Moreover, the calculation method is simple, easy to operate, and closer to the principle of band selection.
The joint entropy represents the information amount of the combination of multiple bands, and the combined band with the largest joint entropy is the best band in the sense of "the largest amount of information." However, the joint entropy method requires a considerable amount of calculation and is generally not easy to operate. When computing, we need to write a program to implement it.
The entropy of an 8bit image X is In the formula, p i is the probability that the gray value of the image pixel is i. Similarly, the joint entropy of 2-3 images is as follows: In the formula,P i1 , P i2 , P i3 represents that the gray value of the pixel in the imageX 1 isi1; the gray value of the pixel with the same name in the imageX 2 isi2, and the gray value of the pixel with the same name in the imageX 3 is the joint 4 Journal of Sensors probability ofi3. Generally speaking, the largerHðxÞ andHðx 1 , x 2 , x 3 Þ, the richer the information contained in the image (or group of images). Using entropy and joint entropy, bands and band combinations with a large amount of information can be obtained. The learning process of BP algorithm consists of two aspects: forward propagation and back propagation. In the forward propagation process, the input pattern is processed layer by layer from the input layer through the hidden layer and finally transmitted to the output layer. The output state of each layer of neurons only affects the input state of the next layer of neurons. If the expected output cannot be obtained in the output layer, backpropagation is performed, the error signal is returned along the original path, and the weights and thresholds between neurons in each layer are corrected to make the error meet the requirements. For the BP algorithm, the commonly used network error is the mean square error function, which is defined as follows: Among them, E p is the error of the p-th network vector, t pj is the expected value of the j-th output neuron, and o pj is the actual value of the j-th output neuron.
The central idea of the BP algorithm is to adjust the weights to minimize the total error of the network; that is, the mean square error E p between the expected value and the output value of the neural network tends to be the smallest. The negative gradient of the error function of the weights points in the direction in which the error function decreases the fastest. If the algorithm moves along this vector in the weight space, it will eventually reach a minimum, at which point the gradient is zero. The weight adjustment formula is mathematically described as follows: In the formula, △ P W ji represents the change of the weight between the source neuron i in the L − 1 layer and the destination neuron j in the L layer. This change in the weights causes the gradient to change in the direction of reducing the error in the weight space.
The goal of the entire BP algorithm is to determine how to adjust each weight to make the network converge. Formula (3) describes the relationship that each weight w ij will change in the direction of the negative gradient where the local error decreases most rapidly. Converting it into an expression suitable for computer implementation, the differential equation required by the BP algorithm can be obtained: In the formula, η is the learning rate, δ pj is the error signal of the neuron j in the L layer, and 0 pj is the output of the neuron i in the L − 1 layer.
The error signal can be expressed as follows: For output neurons: For hidden layer neurons: In the formula, O pj represents the output of neuron j in layer L, O pi represents the output of neuron i in layer L − 1, and δ pk represents the error signal of neuron k in layer L + 1.
According to the above description, the whole BP learning algorithm can be described as follows: (1) The algorithm randomly gives the input layer unit to the hidden layer unit, the hidden layer unit to the output layer unit, the threshold value of the hidden layer unit, and the threshold value of the output layer unit. Generally, a small random number within (-1, +1) is assigned (2) The algorithm performs the following operations on ðx k , y k Þðk = 1, 2,⋯mÞ for the sample pattern (2-1) The algorithm sends the value of x k to the input layer unit and calculates the new activation value of the hidden layer unit of the network.
(2-2) The algorithm calculates the activation value of the output layer unit.
(2-3) The algorithm calculates the output layer unit and the generalization error.
(2-4) The algorithm calculates the training error of the hidden layer unit.
(2-5) The algorithm adjusts the connection weight from the hidden layer to the output layer.
(2-6) The algorithm adjusts the connection weight from the input layer to the hidden layer.
(2-7) The algorithm adjusts the threshold from the output layer to the hidden layer.
(3) The algorithm judges whether the error meets the requirements. If the requirements are satisfied, the algorithm executes step 5, otherwise, executes step 4 (4) The algorithm judges whether the maximum number of iterations is exceeded. If it exceeds, it means that the training fails, and step 5 is performed; otherwise, it returns to step 2 (5) The algorithm ends The shape and size of a feature is one of the important signs to identify a feature. The description of the shape includes regular geometric shapes, such as rectangle, square, rhombus, diagonal, circle, ellipse, and pentagon, as well as some irregular geometric figures. The shape soil of the mea-surement object is the perimeter P and area A of the baseinch object, and its shape index K is defined as follows: The shapes of various types of objects are summarized into the following types.
3.1. Round. For areas of equal area, a circle has the shortest perimeter, and an object that has the shape of a circle will have the largest area/perimeter ratio. For any circular object, it is not difficult to derive the following relationship: The shape index of a square is smaller than that of a circle.

Rectangle.
If the long side of the rectangle is k times the short side, it can have the following relationship: 3.4. Linear Objects. Linear objects have small shape index values, and such objects include roads, rivers, and airports.

Irregular
Objects. The more complex the shape of this type of object, the smaller its shape index. For some objects that share obvious shapes, it is the most effective method to use shape information to identify them. The shape index K can be used as an important indicator to measure the shape of an object. A complex or long object always has a smaller shape index value. Conversely, a simple or round object has a larger shape index value. The area of an object can also be used as another important shape indicator to identify some different types of objects. For example, the shape index of a lake and a reservoir may be similar, but the area of a lake is much larger than that of a reservoir, and they can be completely distinguished based on the area index.
After the knowledge is expressed in an appropriate way, it is necessary to study the reasoning mechanism for using the knowledge. Reasoning is the thinking process of drawing a new judgment from a known judgment according to a given principle. In the remote sensing image knowledge inference system, rule inference and rule expression use production rules to describe the solution to the problem. This kind of rule-based inference is also called production inference. At the same time, roughly speaking, production reasoning is the reasoning engine selects the rules from the rule base and applies the rules. When these selected rules are applied, they perform the actions of the concluding part of the rules and thus lead to the addition, deletion, modification, etc. of the   Figure 1(a) is a schematic diagram of the changes of rules and facts in the working process of the inference engine, and the direction of the arrow indicates the direction of data flow. Figure 1(b) is the rule discovery.
In the object-oriented knowledge representation, the operation of the object is contained in the object; that is, the reasoning is distributed in each object. Rules represent the way and strategy of reasoning and are used to represent the experience of experts in the specific application of knowledge. During inference, when the fact expression is consistent with the conditional expression of the rule in the knowledge base, inference can be generated according to the rule, and the credibility value given by the expert can be obtained.

Intelligent Management of Land Resources Based on Internet of Things and GIS Technology
The design of the system database mainly adopts the topdown design scheme and the object-oriented programming idea. The top layer is the workspace, which includes three parts: data source set, map set, and layout set. Map set and layout set are mainly used for map output of spatial data, and they are composed of map elements and layout elements, respectively. Moreover, the data source set can derive subclass data source, the data source is composed of the data set, and the data set is a subclass of the data set, which  includes the vector data set attribute data record set. Its structural frame is shown in Figure 2. The data can be logically divided into spatial data, image data, attribute data, public configuration data, and document data. Spatial databases are divided into two categories. One is basic geographic data, administrative area data, and image data, which are provided to other applications as shared data. The other category is each thematic application elements as shown in Figure 3.
The implementation strategy that goes hand in hand with prototype trial, database building, and software customization and transformation is specified. The implementation scheme   Figure 4.
The design of the database E-R model is shown in Figure 5, in which Figure 5(a) is the ER model of the land registration business, and Figure 5(b) is the basic geographic ER model.
The construction land resource evaluation study is based on the workflow shown in Figure 6.
This system uses the current land survey data, planning data, and agricultural special data as the analysis data; uses C# as the development front-end; and is designed and developed based on ArcGIS Engine. The system structure is shown in Figure 7(a), and the framework generally includes five parts: map display, query module, analysis module, and land evaluation, as shown in Figure 7(b).
On the basis of the above research, the intelligent management system of land resources based on the Internet of Things and GIS technology proposed in this paper is evaluated, and the effect of intelligent processing of land resources and the effect of land resource management are counted, and the results shown in Tables 1 and 2 are obtained.
From the above research, it can be seen that the intelligent land resource management system based on the Internet of Things and GIS technology has a good effect in the collection and processing of land resource information and can effectively improve the management efficiency of land resources.

Conclusion
According to the resource and environmental carrying capacity, existing development density, and development potential of different regions, this paper considers the future population distribution, economic layout, land use, and urbanization pattern and divides the land space into four main functional areas: optimized development, key development, restricted development, and prohibited development. Moreover, this paper clarifies the main function, guides the development direction, regulates the development order, controls the development intensity, adjusts the development policy, and gradually forms a new pattern of spatial development in which population, economy, resources, and environment are coordinated. The main functional area is the product of implementing the scientific concept of development, and the proposal of the main functional area plan is also a new breakthrough in promoting the coordinated development of the region, and it is the regional embodiment of the sustainable development strategy. This paper combines the Internet of Things technology and GIS technology to build an intelligent management system for gradient resources. The research shows that the intelligent management system of land resources based on the Internet of Things and GIS technology has a good effect in the collection and processing of land resource information and can effectively improve the management efficiency of land resources.

Data Availability
The labeled dataset used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest
The authors declare no competing interests.