Increased AI adoption will be key to farming---improves management of soil, crops, diseases, and weeds

Soil management is an integral part of agricultural activities. A sound knowledge of various soil types and conditions will enhance crop yield and conserve soil resources. It is the use of operations, practices and treatments to improve soil performance. Urban soils may contain pollutants which can be investigated with a traditional soil survey approach [11]. The application of compost and manure improve soil porosity and aggregation. A better aggregation indicates the addition of organic materials that play an important role in preventing soil crust formation. It is possible to adopt alternative tillage systems to prevent soil physical degradation. The application of organic materials is essential to improve soil quality [12]. Production of vegetables and other edible crops is often significantly affected by several soil-borne pathogens that require control through soil management [13]. Sensitivity to soil degradation is implicit in the assessment of the sustainability of land management practices, with recognition of the fact that soils vary in their ability to resist change and recover [14].

A summary in AI soil management techniques is shown in Table I. Management-oriented modeling (MOM) minimizes nitrate leaching as it consists of a set of generated plausible management alternatives, a simulator that evaluates each alternative, and an evaluator that determines which alternative meets the user-weighted multiple criteria. MOM uses “hillclimbing” as a strategic search method that uses “best-first| as a tactical search method to find the shortest path from start nodes to goals [15]. Knowledge of engineering for constructing the Soil Risk Characterization Decision Support System (SRCDSS) involves three stages: knowledge acquisition, conceptual design and system implementation [16]. An artificial neural network (ANN) model predicts soil texture (sand, clay and silt contents) based on attributes obtained from existing coarse resolution soil maps combined with hydrographic parameters derived from a digital elevation model (DEM) [21]. The dynamics of soil moisture are characterized and estimated by a remote sensing device embedded in a higher-order neural network (HONN) [22].

IV. CROP MANAGEMENT

The crop management techniques are summarized in Table II. Crop management starts with sowing, and continues with monitoring growth, harvesting, and crop storage and distribution. It is summarized as the activities that improve the growth and yield of agricultural products. In-depth understanding of class of crops according to their timing and thriving soil type will certainly increase crop yield. Precision crop management (PCM) is an agricultural management system designed to target crop and soil inputs according to field requirements to optimize profitability and protect the environment. PCM has been hampered by lack of timely, distributed information on crop and soil conditions [26]. Farmers must combine various crop management strategies to cope with water deficit resulting from soil, weather or limited irrigation. Flexible crop management systems based on decision rules should be preferred. Timing, intensity, and predictability of drought are important features for choosing among cropping alternatives [27].

Proper understanding of weather patterns helps in the decision-making process that will result in high and quality crop yield [28]. PROLOG utilizes weather data, machinery capacities, labor availability, and information on permissible and prioritized operators, tractors, and implements for evaluating the operational behavior of a farm system. It also estimates crop production, gross revenue, and net profit for individual fields and for the whole farm [30]. Crop prediction methodology is used to predict the suitable crop by sensing various soil parameters and parameter related to the atmosphere. Parameters like soil type, PH, nitrogen, phosphate, potassium, organic carbon, calcium, magnesium, sulfur, manganese, copper, iron, depth, temperature, rainfall, humidity [31]. Demeter is a computer-controlled speed-rowing machine, equipped with a pair of video cameras and a global positioning sensor for navigation. It is capable of planning harvesting operations for an entire field, and then executing its plan by cutting crop rows, turning to cut successive rows, repositioning itself in the field, and detecting unexpected obstacles [32]. The use of AI in harvesting cucumber comprises of the individual hardware and software components of the robot including the autonomous vehicle, the manipulator, the end-effector, the two computer vision systems for detection and 3D imaging of the fruit and the environment and, finally, a control scheme that generates collision-free motions for the manipulator during harvesting [33]. Field-specific rainfall data and weather variables can be used for each location. Adjusting ANN parameters affects the accuracy of rice yield predictions. Smaller data sets required fewer hidden nodes and lower learning rates in model optimization [38].

V. DISEASE MANAGEMENT

To have an optimal yield in agricultural harvest, disease control is necessary. Plant and animal diseases are a major limiting factor regarding the increase of yield. Several factors play role in the incubation of these diseases which attack plants and animals, which include genetic, soil type, rain, dry weather, wind, temperature, etc. Due to these factors and the unsteady nature of some diseases causative influence, managing the effects is a big challenge, especially in large scale farming. Table III lists the AI applications in disease management available in the literature. To effectively control diseases and minimize losses, a farmer should adopt an integrated disease control and management model that includes physical, chemical and biological measure [39]. To achieve these is time consuming and not at all that cost effective [40], hence the need for application of AI approach for disease control and management. Explanation block (EB) gives a clear view of the logic followed by the kernel of the expert system [42]. A novel approach of rule promotion based on fuzzy logic is used in the system for drawing intelligent inferences for crop disease management. A text-to-speech (TTS) converter is used for providing capability of text-to-talking user interface. It provides highly-effective interactive user interface on web for live interactions [45]. A rule based and forward chaining inference engine has been used for the development of the system that helps in detecting the diseases and provide treatment suggestion in [46].

VI. WEED MANAGEMENT

Weed consistently reduces the farmers’ expected profit and yield [51]. A report confirms a 50% reduction in yield for dried beans and corn crops if weed infestations are not controlled [51]. There is about 48% loss in wheat yield due to weed competition [52, 53]. These losses may at times rise up to 60% [54]. A study on the impact of weed on Soybean showed about 8%-55% reduction in yield [55]. A study on yield losses in sesame crops accounts them to about 50%-75% [56]. The fluctuation in yield losses may be attributed to the length of exposure of the crops to the weeds [57, 58] and spatial heterogeneity of weeds [59]. Beyond these, weed has both positive and negative effects to the ecosystem. According to the relative Weed Science Society of America (WSSA) report, weed effects include flooding during hurricane, some species of weeds can pave their way during rampant fire, some cause irreparable liver damage if consumed, and they muscle out plants or crops by competing for water, nutrients and sunlight. Some weeds are poisonous and cause allergic reactions or even may threat public health. Table IV lists a summary of the AI in weed managements uses.

An intensive management with herbicides has been deployed over the past decades to reduce its effect on crops. However, even with this management pattern, it was predicted that crop losses due to weed in western Canada field crops are estimated to exceed $500 million annually [60], hence the need for a more expert weed management technique to compensate for this loss emerges [51]. A system can utilize an unmanned aerial vehicle (UAV) -imagery to divide image, compute and convert to binary the vegetation indexes, detect crop rows, optimize parameters and learn a classification model. Since crops are usually organized in rows, the use of a crop row detection algorithm helps to separate properly weed and crop pixels, which is a common handicap given the spectral similitude of both [64]. Weed control in sugar-beet, maize, winter wheat, and winter barley, can be done by applying online weed detection using digital image analysis taken by an UAV (drone), computer‐based decision making and global positioning system (GPS)‐controlled patch spraying [67]. The drone in [68] travelled at a speed of 1.2km/h, with 58.10ms and 37.44ms execution time to find the tomato and weed locations to the spray controller respectively