Assignment Task
Introduction
The word “hunger” always incorporates the word ‘food’ in its definition. Hunger is the feeling of discomfort that is the major body’s signal that need food. Every person experience this feeling at one time but, for most people this phenomenon is a passing event because they know their next meal is taken, with no serious or permanent damage. When it is hunger or persistent lack of food, the repercussion can be really devastating.
climate change, global warming, reduction in agricultural land, disease and pesticide eliminating agricultural products are some of the reasons of the lack of adequate production of agricultural food supply. According to the World Health Organization (WHO), around 47 million children below the age of 5 are wasted globally, from which 14.3 million children are affected. So, nutritional security needs to be focused on.
Since the major components of nutrients, including carbohydrates, proteins, fats, vitamins, fibres, micro and macronutrients, are not produced in the human body, they must absorb from external dietary sources rich in them (ADD RESOURCE) The per-day recommended values of the essential micronutrients like zinc, iron, and vitamin A are 15 mg, 15 mg, and 600 micrograms, respectively. These nutrients are yet to be fortified in commonly utilized staple crops, including rice and wheat, which contains suboptimal fraction only.
Sustainable Development Goal 2 (SDG2) is about having a world without hunger by 2030. To this end, focus on improving agricultural productivity and food availability has become more and more important and one of the best ways for this advancement is using IoT.
The term “Internet of things “is called IoT. different devices in count capable of sense data placed in different locations with distinct identifications. The IoT enrichment revolutionized the future of technologies being currently used resulting in the development of a new class of smart and intelligent applications and software enabled services. IoT is being used these days in various fields, “Agriculture” is one of them. Modern methods are adopted in the field of farming. Huge volumes of data are generated from the farms, which remains unexplored.
IoT can help for forecasting of weather, identification of plant disease, control the condition of soil and status for different fertilizer application. So, IoT have profound effect on enhance crop production in order to achieve Sustainable Development Goal of zero hunger.
Literature review
We know one of the biggest reasons for world hunger is the lack of enough crops, so world’s population needs to be increased in the field of agriculture. IoT is known as one of the best ways for enhancing crop yields. There are a lot of methods to use for disease and pest control, predict climate change and reduce waste food to reach improvement in crop yield and meet zero hunger. In this literature some of the ways are mentioned and one of them is explained in more details.
The “Alerting Systems” named “Convolutional Neural Network (CNN)” identifies the images of fruit and crop to monitor and control the growth of various crops are being devised.
Huge database of plant specious are available, which use for forecasting of weather, smart methods of irrigation by using modern computational methods and use for plant identification or plant disease. In this method disease are recognised by image process, extracting identical features and then classification.
Another study identifies the use of IoT including Radio Frequency Identification (RFID), Wireless Sensor Network (WSN) and Global Positioning System (GPS). In this method agricultural products reach better quality and more quantity.
precision farming is another method used for gaining better productivity and lower costs with high returns.
A systems “Smart Asset Management System (SAMS)” can also be used to monitor soil condition and status for different fertilizer application. The experiment was performed on rice crops, a smart rice planter was used. Also, fertilizer applicator and yield monitoring harvester were used
Discussion
In this literature what is explained in more details is the different use of “deep learning tools” using IOT and how they can increase crop yield and help zero hunger.
Deep learning is a sub-field of machine learning. It is also known as deep structured learning. It is based on learning data representations. It enables us to imitate the work of the human brain in processing data and creating pattern to be used for making decisions. Applications include driverless cars to speech recognition etc. Automatic systems for emotion recognition are built for the prediction of human emotions. Improve the accuracy of classification in “Feature Selection Methods”. Still, many of these methods capture linear relationships between features. Deep learning techniques can be used to overcome earlier weaknesses by application of complex non-linear features. It was proved that higher-order non-linear relationships are better for emotion recognition. Deep learning techniques belong to a class of machine learning algorithms capable of accepting raw inputs into intermediate layers that give better results in different domains. They show improvements in the domains of biology and medicine to solve issues in these domains. Deep learning supports solving patient classification, fundamental biological processes as well as treatment of patients.
Deep learning enables many processing layers to learn representations of data with a lot of layers of abstraction. Hence these techniques have improved results in speech recognition, visual object recognition and object detection etc. Deep learning explores a complex structure in large data sets using the backpropagation algorithm to show how a machine should change its internal parameters that are used to compute the representation in each layer of representation in the previous layer. Deep Convolutional Networks (DCN) have improvements in processing of images, video, speech, and sound, whereas repeated nets have shown a light on sequencing data such as text and speech. (W. Wang, X. Wang and S. Mao, “RF sensing in the Internet of things: A general Deep learning Framework,” IEEE Communications Magazine, vol 56, no. 9, pp. 62-67, 2018.)
It shows different inputs, such as, soil, water, and energy, which play an important role in crop production. Also, timely use of quality seeds and fertilizers results in higher production. Pesticides, fungicides, and equipment used help farmers to get better results as a precaution to avoid different diseases spread in crops.
All these inputs are applied to the crops, livestock and fisheries and so crops are grown and harvested. Farmers also need storage facilities and timely transport them for usage as food or manufacturing other products from these crops and finally they are sent to retail to consumers.
One of the Deep Learning Techniques is Auto-encoder Neural Network. It is an algorithm. It is composed of 3 parts, an input layer, one or more hidden layers, and an output layer. The input layer and output layer have same number of nodes while hidden layer has lesser nodes than input layer. The training process comprises 3 stages pre-training, unrolling and fine-tuning. In pre-training each neighbouring set of two layers is modelled as a Restricted Boltzmann Machine (RBM). Now the deep learning network is unrolled to get the input with forwarding release. Back-release technique use for results to be fine-tuned. This technique is widely used in activity recognition and health sensing applications.
Another technique is the Convolutional Neural Network. It is done by depicting how living things develop their visual perception and is used successfully in computer vision. Its first architecture was LeNet-5. The convolution and subsampling operations are applied to input data in respective layers. Two groups of computations the output layer is processed via a neural network resulting in improved classification results. Microsoft improved the results via residual learning network Res Net.
A third technique is Long Short-Term Memory (LSTM). It was developed to handle data with long-term dependencies. Uses 3 gates to control the data flow. The input gate decides the amount of new data to be entered, forget gate controls if retain or forget this amount and output gate decides if this value can be used in output calculation for the unit. The gates are an effective way to optimize training of LSTM. It is used in machine translation, speech recognition, and time prediction. (S. Hoch Reiter and J. Schmid Huber, “Long Short-Term Memory,” Neural Computation”.)
Findings
Below shows different IoT sensors and their uses in the measurements of field inputs. Detailed sensor names are shown. A modified framework has been proposed, which shows the relationship between various field inputs and their connectivity with IoT sensors to gather data for further processing which is shown in
Our proposed framework covers various inputs including soil condition, pH value of soil, water condition, energy, seeds, animals, fertilizers, pesticides, fungicides, and equipment used. The inputs are connected via Wi-Fi or RFID or Bluetooth. Various IoT sensors are used for monitoring the field’s inputs.
The soil moisture sensor is used to monitor the contents of moisture in soil and data can be used for automated irrigation planning and climate research.
Evaporation are key players in soil moisture maintenance. pH Value sensors are used to monitor the land suitable for each crop is planning for cultivation. Location Sensors are used monitor the agriculture equipment as well as animals. Land characteristics are also monitored.
Optical Sensors are used for better use of fertilizers, pesticides, and fungicides. They can be placed on vehicles, drones or even satellites for effective use. In addition, optical Sensors using optical technology to measure soil properties.
Animals can be equipped with collar tags to give their health information. Also, their temperature and nutritional insights can be gathered for individual animals as well as herd.
Electrochemical sensors are also used to gather pH and soil levels for the field. And Mechanical sensors can be deployed to measure soil compactness and its resistance.
Dielectric sensors also measure soil moisture. Airflow sensors are used to measure soil permeability. It is also used to detect soil compaction, structure, soil type and moisture level.
N-Sensor is used to monitor the nitrogen level in crops. It guides us to maintain the fertilizer requirements of a specific crop.
Image sensors are used to monitor seeds, animals, pests as well as diseases in crops. Acoustic sensors are used to detect insects or pests attack at the initial stage for the protection of crops. They are also used to monitor animals as well as soil conditions. All types of data from sensors are automatically transmitted via Wi-Fi, RFID or Bluetooth to the data server for further processing.
Deep learning algorithms are applied to real-time data to enable better decision making for crops. The field can be divided into different zones and predict possible performance maps may enable farmers to focus on those areas with low yield prediction and finding the responsible factors can empower farmers to reduce the reasons for low yield prediction in real-time to improve the yield
Conclusion
This literature suggests suitable sensors deployment for better crop yield and a framework which can be adapted by farmers to increase crop production. It is conducted to compare the application of different deep learning tools for the crop data gathering using different IoT sensors to make decision making easier. Deep learning techniques accept raw inputs into intermediate layers then give better results in different domains. So, using IoT and Deep Learning together enable farmers to meet the increasing demand for food for the world. Farmers are enabled to have more suitable land for every crop. Early detection of a disease and attack of insects or pests can help mitigate the disease and protect crops. It is possible to have better use of fertilizers, pesticides, and fungicides.