90+% precise plant distancing reports and gap detection in few hours with DJI Multispectral Drone and Proofminder AI Model

Understanding the plant spacing and number of missed plants is an important factor in successful crop and seed production. Plant population has a direct impact on the yield, quality, and health of the plants, as well as the overall size of the harvest. With the rise of technology such as artificial intelligence, computer vision, accessibility of drones and their mainstreamed usage in agriculture, it is now possible to implement trending precision seeding techniques to maximize yield and quality.

 

Impact of Plant Spacing

 

Growers strive for homogeneous emergence, as this is key to maximum yield. The quality of sowing basically determines the success of homogeneous germination, which is a critical factor for the later life of the crop.

 

Unfortunately, for the time being, not everyone can afford to purchase the most modern, most precise and technologically innovative sowing machines of all time, thus ensuring the accuracy of sowing. The genetic background of seed hybrids can provide a solution to uneven plant spacing.

 

The purpose of precision agriculture is to provide the optimal conditions for each plant, that is, to create harmony between soil conditions and existing technologies to utilize our area as efficiently as possible.

 

Figure 1: Drone usage for plant stand count and plant distance measurement

 

Current agricultural practices to measure plant population and distance

 

Manual plant distance measurement is a common method of calculating the number of plants and the spacing between them. This process assumes that the grower measures a distance manually on chosen field spots and makes assumptions. Modern agronomists already use digital scouting tools to estimate the plant number and distancing, but the biggest issue here again, is still sampling approach and “guesstimations” but not solid data you can rely on to make confident decisions. In addition, none of the methods above provides information about issues or problem areas on the field, seed quality, plant performance or additional insights.

 

Differential seeding just makes this whole process more complex and impossible to handle with manual measures. In recent years, this technique has become increasingly popular in crop and seed production. Also known as precision seeding, it is becoming a more and more common practice. It involves adjusting the rate of seed dispersal based on the soil type and other factors, such as the rate of emergence and the size of the seed. This allows farmers to optimize the number of seeds they use, resulting in higher yields and improved crop quality.

 

The differential seeding method core practices and benefits:

• Sowing zones are formulated based on the soil conditions. Conditions are either based on measurements (sampling or calculated from the electric resistance of the soil) or derived from past year's vegetation health (satellite images and vegetation indexing);

• For zones where the soil has higher capacity (water, nutrients) growers put more seeds, resulting less distance between plants;
• Figuring out the right zoning is a multiple year process, which leads to more precise results and help to obtain the best agronomic decisions and outcome;
• Stand counting is key for zoning and precision seeding process/machine quality control.

 

Having the exact data for comparing variability in planter unit performance or various seeds and improving the yield is extremely important.

 

Figure 2: Proofminder AI-powered model for sowing quality and sowing zones analysis

 

To address the challenges of seed and crop growers, we created our new AI model to measure the plant distance and plant gap analysis at scale, which generates 90%+ precise and actionable reports in just few hours for fields of any size.

 

Proofminder approach

 

Proofminder extracts data from high-precise aerial images to provide growers with actionable reports on the level of plants of leaf across the season. Why it is different:

• By analyzing data on a micro level, the platform sees and count very single plant and able to detect issues on the leaves;
• We scan the whole field, so each cm2 of it is available to review on the screen or as a file to share and use for spraying drone or any Ag machine;
• Platform creates orthomosaic automatically, highlights problem zones and show its GPS-coordinates;
• No specific knowledge of equipment needed, we provide support and partnering with drone service providers to cover the whole innovation process;
• Quick innovation cycle and report building. It takes few days for us to build a use case for a new plant type or a few hours to generate an actionable report for existing use case;
• The data can be used for other calculations on the same platform such as wildlife damage, disease recognition, yield estimation, identifying gopher holes, and many more.

In this article we will describe our latest project for plant distancing and gap analysis with our new AI model for sugar beet, but the same approach could be applied for field crops, vegetables or trees.

 

Step 1. High-precise image collection

 

Here is the field of a large sugar beet producer in Hungary who running various R&D projects across the season and has to measure and analyze tons of things. The challenge is to understand how different seed coatings perform on a given field. The method is to sow 6 rows of each coating and there are 50+ kinds of them. The goal is to see how they perform during the season, especially during and right after germination where the coating has a big role in preventing the seed from diseases and insects.

Using our AI model, we can evaluate the sowing quality and identify potential problems at the early stages.

 

Figure 3: Image gathering with DJI Phantom 4 for sugar beet plant distancing analysis

 

The producer has multiple test fields to run the experiment in different conditions. Each of them is 3-8 ha in size so counting and documenting the experiment is a huge manual and time-consuming work. It also must be super precise as the differences are sometimes small or minimal.

 

As the field sizes are not that big, we used smaller drones to capture the data. DJI Phantom and DJI Mavic [add links] series drones with RGB cameras are well capable drones for these kinds of data-capturing missions.

 

In these kinds of missions, we cannot compromise on image quality. Fortunately, even smaller DJI drones provide excellent image quality, and we can capture the needed 0.4-0.5 cm/px. Obviously, the best is to fly in sunny conditions and light wind.

 

Step 2. Data processing with AI-powered model for precise plant distance measuring

 

After uploading images to the Proofminder platform, it took a couple of hours to get the report with:

• Exact number of plants
• Exact distance between plants
• Number of missing plants per coating
• A visual overview of the field on a micro level
• GPS coordinates of problem zones and missing plants
• Additional insights and possibility to upload or share the report or file.

 

Figure 4: Proofminder report: Precise plant stand count

 

Figure 4: Proofminder report: Sowing lines

 

Figure 4: Proofminder report: Plant distancing in cm and missed plants identification

 

Step 3. Outcomes

 

The seed producer was able to assess the protective power and refine the coating formula. Which is especially important as one of the main components in their current recipe will be banned in the EU soon and they only have a few years to come up with an equivalent or better greener solution.

 

Proofminder’s mission is to create more sustainable and innovative agroindustry and enable growers to achieve production goals with confidence. To try out our new algorithm for precise plant spacing measures, learn more about the platform capabilities or talk about your production challenge – submit a form on our website – proofminder.com, your first AI report is on us!