AU2021104029A4 - Dryland agriculture and yield gap analysis by machine learning algorithms using iot sensors - Google Patents

Abstract

DRYLAND AGRICULTURE AND YIELD GAP ANALYSIS BY MACHINE LEARNING ALGORITHMS USING IOT SENSORS Abstract In arid zones, agriculture is a key component of every country's economy. Another major problem that agriculture in the arid zone confronts is a lack of water, which keeps it from producing the highest potential output. Increasing yield is possible with the Internet of Things (IoT). Agricultural productivity is affected by many external influences, including biotic and abiotic challenges, which are all exacerbated by climate change, creating long-term sustainability problems. To have success with these obstacles, we will have to gather many heterogeneous data sets as well as advanced analytics to bring them together and discover the fundamental productivity problems on different sizes. We ran an ML (Machine Learning) algorithm on the data to see whether the problem persisted, and then we showed off a sophisticated IoT-powered micromanagement device that was capable of monitoring certain environmental factors constantly in various locations. We utilised an ELM (Extremely Learning Machine) to determine the moisture content of the soil surface. Following the data's pre-processing and feature extraction, it is then put into an ELM-based regression model that predicts the soil surface's humidity. This device regulates the amount of water that is in the soil as well as the irrigation system, making it possible for agriculture to be sustainable. The moisture sensors in the packaging design transmit an alert when the amount of moisture drops below the set thresholds. To avoid delays and unnecessary machine operations, users must take manual action to restore the moisture level. 1

Description

DRYLAND AGRICULTURE AND YIELD GAP ANALYSIS BY MACHINE LEARNING ALGORITHMS USING IOT SENSORS

Field of the invention:

This algorithm has been designed to handle agriculture in a dryland setting. An examination of the machine learning method is used to study the yield gap. In addition to the predictive analytics, IoT sensor-based agriculture is used in tandem with extreme learning machine analysis (ELM).

Background of the Invention:

In many areas of the globe, dry zone agriculture is considered as a poor production environment which is also seen as a high-risk and susceptible location. Nevertheless, since agricultural crop cultivation is prevalent in these areas, it has a significant effect on the agricultural ecology of many nations and serves as a classification method for "poor and less developed" countries. Growing crops in areas where the yearly rainfall exceeds 750 mm of rainfall is known as dryland agriculture. Not only are crop failures uncommon, but it is very rare for a growing season to have an extended dry period. More rainfall is needed in these regions, which are particularly vulnerable to dry spells, since there is a great deal of variability in the amount of rainfall and the phenomenon known as early withdrawal is heightened. When you have a poor output, a lack of profits, a low income, and a lack of sophisticated agricultural techniques and equipment, you have subsistence farming. A critical element influencing the properties of dryland soils is the environment in which soils develop. This affects the characteristics of dryland soils, and the connections between these qualities and present environmental circumstances are an important aspect in determining soil performance.

Water, wind, and soil erosion problems are faced in dryland regions.

Soil processes depend significantly on biological, physical, and chemical soil processing. It is important to maintain the organic matter in the soil as a repository of nutrients, since it increases the exchange of cations and reduces compaction. Improving soil texture and improving water penetration are benefits of organic matter. This soil bacteria enzyme acts as an energy source for microorganisms in the soil, and helps counteract soil pH fluctuations. ELM is a quicker learning process, better accuracy, lower training error, and weight norm, and it is thus used to help in increasing dependability.

IoT-powered agricultural farm in a dry zone has been given an eco-friendly boost. Agriculture is a large component of the financial systems of many nations. To make the crop production in the dry zone a challenge, it is necessary to take into consideration a number of relevant factors such as water scarcity, soil loss, meteorological conditions, etc. There are new methods that can be successfully used to agricultural land to improve output by maximising inputs, for example, the Internet of Things. The authors have described an IoT-powered novel micro-climate management system that can reliably monitor a wide range of environmental factors, such as temperature, precipitation, and humidity, across specific areas. Using this methodical approach, our irrigation system allows us to manage the moisture content of the soil, which enables us to control the level of water in the pot to which the plant is exposed to sustain sustainable agriculture. In the agricultural field, if the moisture levels are lower than the predetermined values, a warning signal is generated in the contact module and sent to the farmers.

took an inventory of intelligent agricultural methods that handle big data.

Conventional agricultural methods make use of the concept of intelligent agriculture, often known as smart farming, which involves the integration of computer science and information technology. Progress in automation is achieved by using simple equipment and basic machinery and this may be enhanced in the future. Because of sensors and small-scale devices, which make up the Internet of Things (IoT), crop production has been rising for quite some time. New and developing technology and methods, such as the utilisation of low-cost network storage, are anticipated to contribute to development. The survey's goal is to discover and characterise current state-of-the art intelligent agricultural techniques. The data gathered throughout this progression is sifted through and used to many successful practises in the industry's main trends.

A wireless sensor network was implemented in an effort to combat dryland farming. Fifty-one percent of the land in India is found in semi-arid and arid regions. Roughly 40% of the world's food grain supply is made up of drier grains including sorghum, pigeonpea, chickpea, and pearl millet. It is essential that both soil and water management be in place if dryland agricultural yields are to be increased. Sensors that provide reliable data on thefield conditions and proper irrigation techniques that allow capital to be applied are essential to capitalising on the investment opportunity. Monitoring the three soil-related parameters of moisture, nutrient, and temperature through Wireless Sensor Network (WSN) helps farmers use effective irrigation methods while utilising a minimal amount of water resources. They studied different cropping methods on dry terrain, including pearl millet, peanuts, and sorghum.

He is part in classifying crop types based on ALOS/PALSAR pictures with the Extreme Learning Machine. Also included in categorization maps are production and computation of agricultural disaster payouts. ELM is utilised as a supervised classifier in a newly constructed single hidden layer neural network. This study utilised multi-temporal ALOS/PALSAR images to investigate whether ELM could be used to identify crop kinds. Under contrast to the k-nearest neighbour method (k-NN), the system in the suggested model employs a standardised Extended Learning Machine (ELM) classification procedure. Through the provided data, several kinds of beans, beets, grasses, maize, potato, and winter wheat will be assessed with the suggested technique. When compared to k-NN, the ELM classification resulted in a greater average accuracy of about 79.3%. This study also shows the usefulness of regular monitoring using the ALOS-2/PALSAR-2 technology.

Delivered a smart farming initiative utilising ELM methodology built on the Internet of Things (IoT). [The better genetic algorithm] is used for enhanced smart farming to exploit the usage of highlighted high-dimensional data (GA). Around 60% of the Indian population depends on agriculture for their living. In addition to identifying plant disease as an issue, the work of a professional entomologist should be used. Research into conventional judgement studies has shown that the techniques are inadequate when attempting to determine the quantity of pesticide or fertiliser to apply. Using an excessive amount causes the crop's survival to be jeopardised, and because of this, everything else is threatened as well. the enhanced genetic algorithm (IGA) is dependent on an improved genetic algorithm (IGA) (IGA-ELM). It is also suitable to biological databases with high-dimensional characteristics, such as brain scans, and to a real-time system, and it reduces classification error by 9.52% and 5.71%, respectively, with the help of a 58.50% and 72.73% reduction in attributes. The simulations demonstrate that IGA-ELM is better able to handle complicated optimization, feature complexity, and supervised binary classification problems, while using fewer features.

Artificial Neural Network and Extreme Learning Machine were used to determine the sugarcane yield based on climatic variables. It is critical that management methods be developed for long term sugarcane production because of the exponential expansion and flexibility of the whole sugar system. In addition, the organisations must satisfy the varying business objectives and needs of different customer groups, as well as the needs of the wider community. By relying only on conventional management methods, an organisation will be unable to implement complete management strategies, and as a result, a full-structure approach is needed to achieve desired results. an acute learning machine that incorporates the 3 areas of water protection, environmental management, and cane supply management" (ELM). Therefore, the study's main aim was to overcome this gap by developing a data-driven crop production model utilising ELM. A major lesson learned was the need for innovation in all technical aspects of machine operation and sugarcane growth models. To get an integrated model, ELM will be used to predict sugarcane's ultimate growth number. The results of the suggested system model are being assessed in conjunction with two other types of models: artificial neural network (ANN) and genetic programming. Root-Means-Square-Error (RMSE) is the primary parameter used in the ELM model, along with Pearson's Coefficient and Coefficient of Determination statistics. Results indicate increased generalisation potential in addition to a steeper learning curve. The successful conclusion of the ELM's intensive supplementary research on improving a sugarcane growth estimate technique allowed it to specialise on supplementary study on this improvement.

Objective of the Invention:

1. This algorithm will be used to conduct agriculture in dryland, which has a significant impact on the agricultural ecology of many nations and contributes to the overall socioeconomic development of the country. 2. An examination of the machine learning method is used to study the yield gap. When computer vision develops, methods that concentrate on deep learning to identify regions of interest in crop pictures have gained in popularity. 3. In addition to the predictive analytics, IoT sensor-based agriculture is used in tandem with extreme learning machine analysis (ELM). The following neural network training method will reduce the time it takes to teach. ELM can provide powerful generalisation efficiency and compute faster because of its ability to simplify training. This algorithm will be used to conduct agriculture in dryland, which has a significant impact on the agricultural ecology of many nations and contributes to the overall socioeconomic development of the country. 4. An examination of the machine learning method is used to study the yield gap. When computer vision develops, methods that concentrate on deep learning to identify regions of interest in crop pictures have gained in popularity. 5. In addition to the predictive analytics, IoT sensor-based agriculture is used in tandem with extreme learning machine analysis (ELM). The following neural network training method will reduce the time it takes to teach. ELM is able to provide powerful generalisation efficiency and compute faster because of its ability to simplify training.

Summary of the Invention:

The unpredictable weather of dryland farming necessitates complete control of agriculture, such as controlling the amount of rainfall. In its widest sense, dryland agriculture encompasses all types of land use in semiarid settings. Deciding how to farm is essential, but also deciding how much to farm and when to farm are vital. Precipitation patterns must be a high priority for dryland farming. Farmers who raise crops in the dry season are worried about the weather. A great deal of agricultural output is reliant on the rainfall types and locations. Additionally, meteorological elements such as hailstones and strong and dry winds may have an adverse effect on early harvests. Methodologies are especially important in choosing the most stable type of agriculture that is suitable for a certain area. It is, nevertheless, critical to understand the probability and frequency of weather. An area's agricultural ability is affected by the local environment, such as seasonal and yearly weather, rainfall, wind direction and speed, and other variables.

The degree and frequency of the atmospheric influences on the agricultural produce determines the correctness of any output that involves it. The agricultural area, which is almost entirely reliant on dryland farming, is facing significant difficulties owing to climate fluctuation. Normal rainfall may vary anywhere from less than half of the average to over two times the average in semiarid regions, and agricultural output can range anywhere from zero to three times the average. Global warming is accelerating because dryland regions are suffering from severe climatic issues. Although global surface temperatures are rising, and there is no evidence that precipitation is improving in these places, this may lead to drier conditions in dryland regions.

the Internet of Things (IoT) is a concept for comprehending the next internet (IoT). These "things" become smarter, autonomous, and omnipresent as a consequence of this concept. Things linked to the internet exchange information with one another through the Internet of Things (IoT). A number of virtual world activities are physically linked to the actual world, such as computer software and procedures. The IoT in this example provides continuous monitoring of the actual world and the remote control of smart devices through the internet. Increasing the processing capability of connected things generally enables them to become smart. This technology's combined computing capability has the potential to change the way smart goods are utilised in the IoT, and is thus crucial to the overall objective. Water scarcity is a major issue in dry-zone agriculture, particularly farming. To increase productivity, farmers will be able to better manage their available water.

In order to accurately estimate crop output, detailed field data processing is required, which is almost impossible for vast regions. You may calculate the total yield of each plant by summing the number of developed bolls and the total fibre content in each gramme of plant matter. Agriculture, on the other hand, is beginning to employ computer vision to better monitor fruit bearing plants. Crop yield prediction, maturity stage categorization, and robotic harvesting production are among the current research endeavours using the algorithms discovered in the recent studies. As computer vision evolves, methods that rely on deep learning for object detection have gained popularity. The increasing usage of Graph Processing Units (GPUs) has made these methods more prevalent (e.g., morphological characteristics). A learning machine that detects and counts the number of outputs in crop pictures. The researchers were able to observe and count the fruits that developed on a plant using a technique that used moving images. To estimate the yield from plants using fruit pictures, an algorithm is created that classifies and counts the fruits. This data may be utilised to keep the level of demand under control.

This application uses a deep learning technique to detect and count final production as well as utilising that information to estimate yield. The capacity to generalise database learning information in a poor learning environment is one of the most difficult things to do when identifying and counting objects in pictures. A database of photos from a commercial manufacturing facility, obtained throughout the day.

Detailed description of the Invention:

Water-stressed farming refers to producing crops in areas with a rain threshold of 750 mm of rainfall per year. Currently, about 53.4% of the land on Earth, or 158.65 million hectares of net agricultural land, is expected to be cultivable. Of total food grain supply, almost half (41%) is covered by the rain's rainfall pattern on agricultural land. As this data shows, food security and long-term economic growth both rely on investments in agricultural research and development that concentrate on dry-land farming. Drylands are characterised by low soil fertility, excessive evaporation, arid weather, and runoff. Dryland soils have low fertility, are unable to hold water, have a light texture, and have a tendency to crust over. Due to their unique morphological characteristics, dryland crops are able to adapt to a wide variety of ecological settings and are able to provide optimal crop yields even in the face of drought stress. the characteristics of these drought-tolerant crops' physiological structure allow them to flourish in a hot and dry climate

When water pressure is greatest, roots may rapidly and thoroughly penetrate the soil.

In terms of transpiration, the leaf decreases the rate at which water is lost, and specific cells on the leaf are there to assist with this.

Desiccation of the roots is prevented with roots having advanced cell walls.

In wider rooting regions, root multiplication is higher.

The morphological characteristics of dryland crops are dynamic, meaning they respond rapidly to a broad range of environmental conditions, which in turn allows them to perform effectively even during periods of water shortage. Soil and water are both powerful tools and significant limitations in the arid regions. Poor dryland agricultural output is attributable to suboptimalfield management, instead of poor soil potential.

Farmer soil and environmental conditions conditions may be tracked and calculated, and appropriate irrigation plan adjustments can be made based on those findings. Climatic and soil characteristics must be measured and statistically evaluated. Using irrigation systems that take into account changing field conditions saves a huge quantity of water, while also ensuring the soil is replenished to the proper moisture level. Environmental factors such as wind speed, temperature, rainfall, and sunlight radiation should be monitored using the proper sensors. Sensors in the field detect data pertaining to the field and relay it to a node, which processes and relays it to the base station. The sensors, ADC, CPU, power supply, and transceiver are all connected to a node. The nodes of the Internet of Things are built to link to one another and then transmit information to a base station. When connected to IoT sensor nodes located in the field, the data from those sensors is transmitted to the base station, which dynamically monitors the field conditions.

soil water sensor: Field usage of water: A soil water content sensor may be used to determine how much water is utilised in a field. To evaluate the water retention potential of the soil, the dielectric constant of the soil is different from the capacitance, thereby causing a measurable difference.

A soil moisture sensor measures how hard the plant roots work to keep water. The soil moisture and soil temperature sensor determine the rate of soil desiccation.

A soil temperature sensor measures the temperature of the soil at a certain depth. This information is used to track seedling growth in the soil.

This sensor measures soil electrical conductivity Salinity is measured by using the millisiemens per metre (mS/m) conductivity sensor to monitor soil conductivity. Electrical conductivity is used to discover soil characteristics such as soil composition, electrolyte levels, and organic matter (EC). Temperature, soil water quality, porosity, salt content, and cation exchange potential all have an impact on the EC of the soil.

Acidity or alkalinity of soil in the agricultural field is determined by hydrogen ion concentrations in a pH sensor. pH sensors use the voltage generated in between electrode and reference electrode to detect the concentration of hydrogen ions.

Even though different models have been developed to provide more accuracy and reliability in yield gap forecasting, more complex contemporary robust methods are required to predict yield gap accurately. Therefore, the research's aim is to determine whether an ELM method might help increase yield gap forecast precision and accuracy. A dedicated ELM model is being created to calculate the yield gap on a periodic basis. Five accessible characteristics (e.g., soil salinity, soil moisture quality, electrical conductivity of soil, temperature of soil, and soil water content) are utilised as input components to compute yield analysis related to physical variables influencing yield. ELM learning was developed as a feed-forward, single-layer neural network (SLFN) learning method. Using the ELM, the arbitrary input weight parameters are calculated and then the output weight parameters are defined numerically. An interconnected structure of linked addition blocks and radial basis function (RBF) intermediate nodes serves as the mathematical representation of the SLFN function.

Claims (3)

DRYLAND AGRICULTURE AND YIELD GAP ANALYSIS BY MACHINE LEARNING ALGORITHMS USING IOT SENSORS Claims: In development is a technique for conducting dryland agriculture, as well as yield gap analysis. This invention claims the following:

1. This algorithm will be used to conduct agriculture in dryland, which has a significant impact on the agricultural ecology of many nations and contributes to the overall socioeconomic development of the country. i. From the first assertion, we know that IoT can be utilised to achieve water balance and control moisture and nutrient levels. ii. From the first statement, it can be seen that IoT applications are designed to alert users if human contact issues occur.

2. This starts with the fact that the data collected from IoT sensors distributed across the area is transmitted to the base station. Soil moisture sensors, pH sensors, soil conductivity sensors, and temperature sensors are included as in-field equipment.

3. An examination of the machine learning method is used to study the yield gap. When computer vision develops, methods that concentrate on deep learning to identify regions of interest in crop pictures have gained in popularity. i. In order to predict yield, a new algorithm is created based on plant pictures taken during harvest season. ii. In addition to the predictive analytics, IoT sensor-based agriculture is used in tandem with extreme learning machine analysis (ELM). The following neural network training method will reduce the time it takes to teach. ELM is able to provide powerful generalisation efficiency and compute faster because of its ability to simplify training. iii. This supports claim 3, since the ELM has the capability to meet weight guidelines and has the least training error.

Elm uses training and assessment of datasets, as well as fresh data received, to enhance model accuracy and performance.