Flying high to locate crop failure in Sugarcane

Saksham Bhutani

Saksham Bhutani , Marketing Head at Indshine

February 12, 2020

blog-image

Within the last few years, sugarcane has become very important in Brazil’s economy and in the world scenario. Brazil is the largest producer of sugarcane in the world, sugarcane occupies 9.8 Million hectares that represents about 12% of all cropland in Brazil.

Brazil produces 25% of the world’s sugar & ethanol, so any damages and pests in Sugarcane can be crucial to the economy. Some of Brazil's main pests of sugarcane and their respective losses are:

  1. Sugarcane drill, from the genus Diatraea – with only 1% infestation, the loss can go up to 35 kilos of sugar and 30 liters of alcohol per hectare;
  2. Root and leaf cicadas of the genus Mahanarva – can cause productivity reductions ranging from 25% to 60% in sugarcane, and 11% in the plant cane;
  3. Sugarcane pecking, of the genus Sphenophorus – in average values, the losses incurred are around 20 to 23 tons per year per hectare cultivated;

There may be several causes associated with sugarcane crop failure some of them being: faulty planting methods, seed quality, climate, genetics, and soil quality. To cope-up with these issues, precision agriculture has become a necessity.

Here's a case story of Velbrax Agro, one of the leading companies working in the precision Agri-Tech market in the Sao Paulo region of Brazil who has helped a renowned agro company to locate crop failures in sugarcane farms through drones.

Concept

To indicate planting quality in sugarcane crops, an important parameter that is monitored is "gaps in crop plantation". Gaps are mostly caused by problems during planting operations (e.g., failures on stalks deposition, pests, dry weather, erosion, etc.).

sugarcane crop orthomosaic on Indshine

The main problem of sugarcane cultivators is how to increase the output per unit of input and thereby reduce these gaps. Earlier, the gap was figured out from visual analysis, and then those gaps were refilled with the sugarcane seeds to allow them to grow. But one needs to know how many tons of seeds will fill all the empty spaces. Before the use of drones, this process was visual and less efficient.

Quick facts about Sugarcane Crop

  1. In general, January to March is the period of planting and December to March is the period of harvesting.
  2. Sugarcane crop is harvested three times before a complete harvest.
  3. Mechanized farming has picked-up pace, leading to an increased gap in sugarcane.
  4. 0.5 to 1m is the standard spacing for sugar cane plantation
  5. If the spacing is more than 1.2m - 1.5m, it is assumed that there is a crop failure.

Problem Statement

A renowned agro-based company had been majorly facing the following issues with sugarcane production:

  1. Due to the inaccessibility of sugarcane fields, crop monitoring throughout the season becomes difficult
  2. Inefficient use of available resources, such as water, agrochemicals, vinasse, etc.
  3. High crop failures because of multiple reasons, including a growing pest problem in the Sao Paulo region.

This aforementioned agro company had partnered with Velbrax Agro for the following area of concern, for the aerial inspection of gaps in sugarcane field at the State of Sao Paulo.

project statement for Velbrax Agro

Approach

Carlos Orion, the co-founder at Velbrax Agro, figured a unique solution to the problem statements by proposing the use of drones. He used a fixed-wing drone for the detection and quantification of planting failure in sugar cane fields, near the City of Barrinha, in the State of Sao Paulo.

aerial survey procedure

A fixed-wing drone mounted with an RGB sensor Sony A6000 (24MP) was used to capture raw images. Orthomosaics were generated using Agisoft Photoscan (photogrammetry software) by stitching raw images.

Picterra algorithms demarcated gaps in the sugarcane fields by creating spacing lines. Before moving ahead, let's understand a bit about Picterra. Picterra is an end-to-end, code-free geospatial cloud-based platform for training deep learning-based detectors quickly and securely. Picterra help build and deploy unique actionable and ready to use deep-learning models without a single line of code and with only a few human-made annotations. It automates the analysis of satellite and aerial imagery, enabling users to identify objects and patterns (road cracks, damaged roofs, etc.) at scale, anywhere on Earth. Concept can be easily understood through the process below:

  1. Outline all objects inside your training area.

Use the Polygon or Circle tool to outline all the objects or patterns contained within your first training area that you want to detect. Don't forget about objects partially inside the area.

Picterra Training Area - 1

  1. Draw more training areas and outline all objects inside them.

Draw a few more training areas over the regions where examples of your object are present. Remember to outline all the objects inside those training areas. You can train your detector with multiple images if you place training areas on each of them.

Picterra Training Area - 2

  1. Draw testing area(s)

Now, it's time to decide where your detector will be tested. Use the 'Testing' area tool to draw a few areas where your detector will output a preview of your results after training.

Picterra Testing Area

  1. Train your detector and assess the detection

Finally, click on the 'Train Detector' button to visualize how it performs over your testing areas. If you feel like these training results can be improved, click on Improve Detector and add a few more training areas where you see false detection or detections that are missed.

Picterra Train Detector

Here's link to Picterra results. Final Reports were prepared based on the spacing lines / gaps to figure out:

  1. How much is the percentage failure?
  2. How much is the number of occurrences?
  3. How much is the fault length range?

The project link shows the orthomosaic with spacing lines on it:

Results

Sharing the results of area-2 for glimpse:

  • Fault Report highlighting bar chart between "distribution of the number and length of failures" ; "number of occurrences" and "fault length ranges"

fault report for area 2

  • Parallelism Report highlighting distribution curve between "% of spacing" and "spacing interval".

Parallelism report showing histogram of spacing intervals

Summary

In a total area of 250 hectares, drones were able to save R$ 18,44,000 (Brazilian Real), which could have been sort of impossible by visual analysis.

cost analysis of crop failure Market price as per CONSECANA report is R$ 79.57 per ton

Carlos Orion (Co-founder at Velbrax Agro) explains - Indshine is the go-to option of showing their works to their existing as well as prospective clients. Indshine is a new concept of mapping share. Through Indshine, you can also choose how many and how people have access.

There's close to 5.6 million hectares sowing of sugarcane crops in the São Paulo region, hence a lot more potential to save. Technologies such as VRT, remote sensing, drones, GPS and GIS technology can add a lot of value in increasing productivity, decreasing labor overheads and enabling optimal use of resources.

References:

  1. http://www.igidr.ac.in/pdf/publication/WP-2013-011.pdf
  2. http://www.scielo.br/scielo.php?script=sci_arttext&pid=S1413-70542016000300347
  3. https://www.researchgate.net/publication/318215770_Identification_of_gaps_in_sugarcane_plantations_using_UAV_images

Contributors : Shashank Tewari, Deepali Joshi


We use cookies to serve you better. Privacy Policy


Sign up is now closed

To know more refer here

Indshine Logo

Indshine is a technology company that provides enterprise drone software and solutions for infrastructure, mining, forestry and other industries.

Get in-touch


© 2019 Indshine Inc. |Terms of Service - Privacy Policy