Machine-based Smart Farming Solution Developed by LTTS

By Sharada Prahladrao

Project Success Story

Customer Profile:  The client, a leading motion precision equipment and automation solution provider, smart farming solutionwanted to build an advanced, vison-based, smart farming solution to analyze the growth of plantations inside a greenhouse.  After evaluating the available market options, the client selected LTTS (L&T Technology Services Limited) for this project.  LTTS has worked on vision solutions that have transformed several of its customers’ businesses.  From vision-based packaging systems to complete plant automation, and self-driving cars, LTTS has delivered solutions across industries.  LTTS used a tomato plantation as a starting point to build this smart machine vision solution.

Challenges:  The yield in a plantation depends on the type and amount of fertilizer, CO2, humidity, and temperature.  The existing process involved a greenhouse manager to manually decide these parameters after checking each plant, flower, and stem.  This was a tedious process and lacked accuracy.  It was also found that yield can be predicted based on color, angle, direction, width and length of the tomato petal, pistil, and stem.  However, this activity cannot be performed manually and needed an automated system.  

Identifying minute discoloration and the insects that live around the plantation helps in detecting and mitigating plant diseases at an early stage.  However, the process is very labor intensive, and its accuracy is again dependent on the eye of the observer.

Solution: LTTS’ engineers conducted a thorough research and understood that the artificial intelligence (AI) solution should capture flower, leaf and other plant features for analysis.  There were three broad areas to address:

  • Analyze the plant features to optimize greenhouse environment for higher yield
  • Early detection of signs of diseases in the plantation
  • Detect and count insects based on their size for further analysis

The project execution was segmented into four major phases:  Flower detection; flower segmentation; plant segmentation; analysis of tomato flower and plant features.  While doing this, several challenges had to be navigated, such as segmenting features that are of the same color, extracting data from zoomed out images, and processing all the information in less than 2 seconds.

The team finally designed an AI-based vision system that takes continuous pictures of the plantation and processes the information within 2 seconds to churn out insightful data.

Significant Benefits:  LTTS’ solution provided:

  • Complete automation of tomato flower and plant feature data extraction
  • Automated flower feature extraction that took 80 percent lesser time (compared to the manual procedure)
  • The deep learning algorithm achieved detection and segmentation accuracy of over 95 percent

The solution uses deep learning and throws insights to optimize parameters such as humidity, CO2, temperature etc. for better yield.  It also detects leaf parameters for early detection of plant diseases;   and detects, classifies and counts the insects in a sample area of the plantation.

Engage with ARC Advisory Group