Using AI to Increase Accuracy of Soil Nutrient Predictions

An innovative soil nutrient testing system named Alternative Analytical Technology (AAT) in collaboration with the Indian Institute of Technology Madras (IIT Madras) to provide soil nutrient analysis service to farmers. MCRC aims to increase the accuracy of technology further by increasing the system datasets and improving program performance.

Project Overview

Project Description

Shri AMM Murugappa Chettiar Research Centre (MCRC) developed an innovative soil nutrient testing system named Alternative Analytical Technology (AAT) in collaboration with the Indian Institute of Technology Madras (IIT Madras) to provide soil nutrient analysis service to farmers. This technology is based on subjecting a small sample of soil to generate a chromatogram image using the simple chromatographic method. A comprehensive soil nutrient analysis is generated by the image processing method. This technology analyses pH, organic carbon, humus, macro-nutrients as well as secondary and micro-nutrients of soil (i.e. 16 parameters) within 6-8 hours in a single test.

The existing system database was developed using the 20,000 soil samples through a set of machine-learning and image-processing programs. Currently, this data set is being used for generating soil nutrient analysis reports for the new samples. The present accuracy of the system, to predict nutrients in the soil is about 85%. MCRC aims to increase this accuracy further by increasing the system datasets and improving program performance. MCRC has additional datasets for 10,000 soil samples that can be used for training the model, and this dataset can be even further increased in future for better results.

Primary responsibilities include:

  • Developing an improved database by amending the data of 10,000 soil samples and evaluating the accuracy of the improved database by using suitable methods like machine learning/ Artificial intelligence/ deep learning.
  • Improve the performance of Alternative Analytical Technology (AAT)
  • To enhance the prediction of nutrient level on the soil sample with the developed image by comparing with the new enhanced database
  • To support the development of an improved user interface

Impact

Soil nutrient testing through the conventional analytical method is time-consuming, expensive and needs well-equipped laboratories leading to practical difficulties such as inaccessibility and affordability. This rapid soil testing method provides alerts on excess or deficiency of nutrients as well as gives recommendations on optimum agri-input levels. The AAT technology consumes less energy and is 20 times cheaper than conventional techniques.

Over a decade, around 50,000 soil samples from farmers of more than 12 districts of Tamil Nadu and Puducherry were tested, and fertilizer and crop recommendations suitable for that particular field were provided. MCRC has been working towards providing sustainable solutions and interventions and inculcating environment-friendly practices in rural sections of Tamil Nadu state of India for over the past two decades.

Eligibility Criteria

Skills / Experience:

  • Required:
    • Back End Programming
  • Desired:
    • Skills and experience working with systems or models based on Machine Learning, Artificial Intelligence and Automation
    • Skillset for image processing techniques

Software needs:

  • N/A

Discipline:

  • Computer sciences
  • Electronics Engineering

Fellow location (all projects can be completed remotely):

  •  India

Time zone compatibility (when the Fellow should be available for meetings):

  • IST (UTC+05:30)

Leave a Reply

Join a global community of changemakers.

Become A Member