Due to technological advances in recent years, the working environment in the agriculture industry has undergone significant changes [1]. The agricultural industry has entered a new age when fast and complex decision-making becomes the key to success [2]. Technological developments such as big data, IoT, artificial intelligence, neural networks, cloud computing, etc., equip farmers with tools and expertise to improve decision-making during their particular phases of production and yielding [3–6]. The existing facts such as population expansion, climate change, and workforce scarcity have led to developing intelligent farming systems from planting and watering to crop health care and harvests [7]. Environmental and remote automation and surveillance using IoT are expanding fast in agriculture to produce more competent and effective agricultural tools and services [8]. Smart farming can supply farmers with valuable intimate environmental information obtained from wireless IoT sensors in the growing fields to increase competition and profitability [9].

The recent emergence in smart farming involves several agricultural domains, as seen in Fig. 1. The smart farming system or infrastructure integrates plant disease diagnosis, fertility ratio analysis, crop monitoring, soil profile monitoring, water spreading, field surveillance, and water stress analysis [10,11]. Disease diagnosis and crop monitoring can be improved through video and image classification. Water spreading and field monitoring are employed using wireless devices such as drones and UAVs [12]. Soil profile monitoring can be done by collecting the soil profile using sensors. The water stress analysis manages the availability and distribution of the smart farming system. Finally, the fertility rate analysis can aid in effective decision-making in production planning and marketing [13]. All these services are available only if there is an effective information management system.

Data from sensors can be communicated with stakeholders in sensor-oriented precision agriculture via either the local server or the cloud according to communication network stability and internet connectivity [14]. Smart farming’s Information management has grown from basic farm recording systems into big, comprehensive farm management systems to fulfill the requirements of the many stakeholders for communication and data transfer between databases [15]. The primary responsibility of such information management is the management of the large volume of data covered by different agricultural parameters, such as information from farmers, soil physical and chemical parameters, nutrients available, crops, weather, precipitation, seasons, fertilizers, pest control, irrigation procedures, crop yield predictions, and environmental factors [16]. During the past 60 years, terabytes of unstructured data in the database have been produced from the processes involved in gathering the various agricultural data [17]. Conversely, large-scale data cleansing and preprocessing have been tedious due to unstructured data types and a single database instance [18].

To address these challenges, data optimization can be effectively used in this domain. Often including errors, inconsistencies, redundancies, or other abnormalities unnecessarily make access to essential information in a timely and complete manner difficult. The data optimization uses extensive quality by the availability of precise data, better access, organization, and cleaning, irrespective of the source, for maximizing the speed and comprehensivity of the collection, analysis, and application of relevant information [19]. Data optimization has been aimed at reducing the lifespan of information in IoT-cloud systems. Reduction in energy consumption seems to be the major advantage of data optimization [20]. Multiple access mechanisms are frequently used to accommodate this non-orthogonal system [21]. The efficient collection of data is another benefit of data optimization. For compressing collected data, compressive sensing can even be used to reduce the volume of information transmission [22]. The optimized data can be retrieved with appropriate retrieval techniques if necessary [23].

For this reason, Information-Centric IoT-based Smart Farming with Dynamic Data Optimization (ICISF-DDO) has been proposed to improve the smart farming system by efficient information management with minimal energy consumption and improved lifetime. The DDO approach with membership parameter initialization has been integrated into an information-centric smart farming system with an IoT-cloud platform.

Section-wise categorization is implemented as follows: Section 1 discussed smart farming and agricultural information management to introduce the entire research. Section 2 summarizes the relevant works on smart farming and agricultural information management by various researchers. The following section elaborates the proposed system design, and Section 4 validates the proposed system’s significance compared to the existing smart farming system. The entire research is concluded with the future scope in Section 5.

Mass transit across the agricultural infrastructure is produced in a huge volume by the IoT-based intelligent farming environment sensors. The exposure of huge data in the smart agricultural system communication network makes statistical and computational data analysis, processing, and optimization crucial. In this proposed ICISF-DDO approach, data optimization is achieved by efficient data acquisition. Two enhancements have been made, in particular in the area of data redundancy and energy usage. Optimized sensing has been used to compact the obtained data to decrease the amount of data transfer. Furthermore, the optimized data from the IoT-assisted smart farming systems can even be retrieved using appropriate recovery solutions. Therefore, DDO in the proposed ICISF-DDO contributes to a minimum delay for energy-efficient, reliable IoT data transmission. In addition, these strategies decrease overall cost needs.

This section deals with Dynamic Data Optimization algorithm. Since the data acquisition and processing in IoT-based smart farming systems involve dynamic variables whose values change in time, this research aims at designing the DDO algorithm as follows:

Algorithm 1 defines the various computation steps and flow of the proposed DDO algorithm in the research model ICISF-DDO. As seen in the above algorithm, for all the nodes set NS in the smart farming environment, the sub-optimized data is computed for the optimal node selected using Sij′ based on the data flow volume DijFV(p→q), energy Eij(p→q), and cost cij(p→q). Then, membership parameter induction has been performed by computing the initial membership parameter Fij(p,q) and final membership parameter Fij′(p,q). The centroid for the state p and q in the data transit have been then computed as Gijq and Gijq. By then, sub-optimized data, ODi is aggregated for the final optimized data allocation, OD. Each step of the proposed algorithm is explained as follows:

In the large space-smart farming system, it is assumed that multiple nodes set in a single, smart farming infrastructure, for example, as assumed in Fig. 2, a large-scale cherry farming environment with each tree is individually observed with multiple sensors including, fruit sensor, leaf sensor, root sensor, and soil sensor along with an environmental sensor. Let us assume in general and there is M number of such nodes set (indexed with i=1,2,3,…,M) with N number of nodes in each (indexed with j=1,2,3,…,N). For the statistical design model, the cloud resource storage variable is initialized as follows: (1)R0=∑M⁡{IMN∑i=1M∑j=1NI0}

Eq. (1) depicts the cloud resource R0 for the entire smart farming system, which is initialized with the incidence ratio IMN of the Nth node in the Mth nodes set over an arithmetic summation of the I0 initial incidence ratios of each N node in the M nodes set. If the cloud resource is not enough or equal to the required sensor data storage, the data processing through cloud retrievel fails by missing data. After initializing the maximum enough of cloud resources for the entire smart farming system, the cloud resource for the individual nodes in each node-set is computed as follows: (2)Rij=argargmaxsij∂(sij)+∅

In Eq. (2), the variable Rij is the individual cloud resource value for each jth node in the ith node-set. The sij represents the state vector of the jth node in the ith node-set. This resource variable finds the arg argmax of the overall state momentum ∂(sij) for each state vector sij. The argargmax sometimes called the argument of the maximum computes the elements of the state vectors of state momentum at which the momentum values are maximized, which helps in finding the maximum cloud resource available for each N node. The variable ∅ is the constant cloud resource value reserved for the garbage collection. After reserving the cloud resource, the steps for optimal data allocation have been calculated. For this, the optimization criteria are then fixed for each node in every node-set, as follows: (3)Ypqi=Ki,where0≤Kij≤Rij

The optimization criterion for the state transition p→q as denoted in Eq. (3) uses the variable Ypqi in value storage and the Ki is the constant randomly chosen from the range of 0≤Kij≤Rij. The p and q are used to represent the states of the non-repeatable data selection criteria of nodes set i. Based on the selected optimization criteria, sensor node selection at the selected location is performed by evaluating the data flow volume, sensor energy, and sensor cost as follows: (4)DijFV(p→q)=Di0FV−∑p=1P∑q=1QAij(p,q) (5)Eij(p→q)=Ei0+R0Aij(p,q) (6)cij(p→q)=ci0−∑p=1P∑q=1QAij(p,q)∗Cj(Aij)

As seen in Eqs. (4)–(6), the node selection criteria such as data flow volume DijFV(p→q), sensor energy Eij(p→q), and sensor cost cij(p→q) are computed for the state transition p→q. In Eq. (4), the variable Di0FV denotes the initial data flow volume for the node-set i, and variable Aij(p,q) describes the server space or the ranging area of the node j in the node-set i for the states p and q, which are included in the state group of maximum P and Q, respectively. In Eq. (5), the variable Ei0 is the initial sensor energy for the node-set i, which is measured with ∑j=1NEj0. The variable Ej0 is the sensor energy of the individual node j in the node-set i. In Eq. (6), the sensor cost is computed with the initial cost ci0 of the node-set i, with the sensor coverage area Cj(Aij) in the state space Aij. The following gives the pictorial view of these three basic criteria of optimal node selection.

Fig. 4 pictorizes the statistical derivations of the three basic optimal node selection criteria. The data flow volume must have been the minimum at the transition from p to q with the total server space for all the states computed as ∑p=1P∑q=1QAij(p,q), depicted in Fig. 4a. In Fig. 4b, the sensor energy criterion should be minimum for the state transition from p to q with the minimal ratio of energy coefficient to the server space. As seen in Fig. 4c, the minimal criteria for node selection in terms of sensor cost is computed based on the total coverage volume of all the sensors, ∑p=1P∑q=1QAij(p,q)∗Cj(Aij). Under these above criteria, the optimal node is selected and represented as follows: (7)Sij(x)=Cj(Aij)β∗Ypq→(0,Cj(Aij))+DijFV(p→q)+Eij(p→q)=Sij′(x)

In Eq. (7), the sensor node positioned at x, that represented as Sij(x) has been chosen as the optimal sensor node for the optimized data allocation that further transmits to the cloud data resource reserved for the smart farming system. The optimal sensor node Sij′(x) selection satisfies the condition (0,Cj(Aij))+DijFV(p→q)+Eij(p→q) from Sij(x). Since the proposed DDO algorithm induces the membership parameter, it is to be initialized as: (8)Fij(p,q)=eDijFV(p→q)∂(sij)∗P∗Q

Eq. (8) defines the initial membership parameter for the state transition from p to q, and represented by Fij(p,q). The membership parameters help identify the association among the states and quantifies it. In this equation, the eDijFV(p→q) measures the exponential value of the data flow volume for the state transition from p to q. The parameter ∂(sij), which is the state momentum has been considered as the state value. The initial membership parameter depends on the total area of node placement in the node-set and is denoted by P∗Q. (9)(p,q)=eDijFV(p→q)∂(sij)+(P+Q)

In Eq. (9), the final membership parameter for the dynamic data optimization is denoted by Fij′(p,q). In this case, the final membership parameter depends only on the total number of optimal sensor nodes selected in the node-set, denoted by P+Q. To compute the optimized data, the centroid of the state p and state q have been found and computed using the following equations. (10)Gijp=∑p=1PAij(p,q)∑p=1P∑q=1QAij(p,q) (11)Gijq=∑q=1QAij(p,q)∑p=1P∑q=1QAij(p,q)

Eqs. (10) and (11) find the centroids, where Gijp is the centroid for the state p, and Gijq is the centroid for the state q in the ith node-set. In Eq. (10), the numerator ∑p=1PAij(p,q) finds the server space for every p states with the area Aij(p,q), whereas the denominator ∑p=1P∑q=1QAij(p,q) computes the total server space. Similarly, for Eq. (11), the computation expression ∑q=1QAij(p,q) has been used for finding the server space for every q states with the area Aij(p,q). The centroids are the datapoint used for the optimal data allocation in cloud resources. For every node-set i, the optimal data allocation has been performed and termed as sub-optimal data allocation. This sub-optimal data allocation can be expressed as follows: (12)ODi=Gijp−GijqFij(p,q)+Fij′(p,q) (13)OD=∑i=1MODi

As seen in Eq. (12), the sub-optimized data can be computed by finding the argument minimum argargGijp−GijqFij(p,q)+Fij′(p,q), of the relation among centroids and the membership parameters. The d0 is the initial data in the optimal cloud resource, and the final optimized data allocation OD can be obtained by aggregating all the sub-optimized data, as seen in Eq. (13).

In summary, this elaborated section included the proposed ICISF-DDO framework and the layered architecture having three significant layers such as front-end layer, the back-end layer, and the gateway layer. After designing these, the DDO algorithm is presented, and the steps in the proposed algorithm are detailed. The entire research model is designed to ensure the IoT-assisted smart farming system with intelligent information handling for reliable performance. The following section includes the validation details in terms of many significant evaluation parameters. It compares the results from the simulation with existing smart farming models.