Video: Automating Tasks in Your Greenhouse Using Artificial Intelligence

[fa icon="calendar"] November 23, 2018 12:13:36 PM EST / by Jason Behrmann

 Expo FIHOQ 2018 Drummondville cover Nar

Artificial intelligence (AI) is growing in importance within agriculture. In this informational video, being a narrated presentation, we explain in simple terms how AI technology works and its current use in automating tasks in greenhouse farming. A transcript of the video appears below.

This presentation first appeared at the 2018 Expo-FIHOQ conference and trade show in Drummondville, Québec, Canada. 


Motorleaf AI overview greenhouse auto ENG


Video transcript


Hello there, my name is Jason Behrmann and I manage marketing and communications at Motorleaf, a Montreal agriculture technology company that makes artificial intelligence services to automate tasks in greenhouses.

AI is the latest technology that is changing virtually all industries, including agriculture. Today I share with you a simple explanation for AI and how this technology can make greenhouse farming more efficient and competitive.

What is AI?

So what is artificial intelligence exactly?

First off, no, it is not robots. Some robots have some AI capacities to help guide a robot’s basic decision-making capacities.

But to be clear: AI and robotics are two separate categories of technology and AI services today often do not involve robotics.

AI is just a new tool in computation/computer science. AI is essentially sophisticated software.

This software can learn specific, repetitive tasks. It learns a task by analysing many previous examples of the task--this is called “training” the software.

Past examples are in the form of “big data”, which is a lot of information in a digital format.

By analysing big data, the AI software can find patterns in the information. Once it understands the patterns, the AI software will be able to complete the repetitive task alone and make decisions for that task with a high degree of accuracy.

Also, since it is computer software, it can analyze a lot of information quickly and do analyses too complex for a human being--just like how we use a calculator to do complex math.

What is interesting is that the software does actually “learn” and does get smarter over time when given more and more data.

A simple example: machine vision

Here is a simple example, something we call “machine vision”, where AI software learns how to identify objects in images.

We can show AI software hundreds of thousands of photos. One group contains images of cats, the other group has images of anything else. We train the AI software by telling it this is a cat, this is also a cat, this is not a cat. After training, we can show the software a new image and ask it: is this a cat? And it will be able to identify this image--and any other image--as a cat with high accuracy.

So why use AI?

AI enables automation.

AI is a great tool to automate common, specific and repetitive tasks, including many complicated tasks.

Think about all the repetitive tasks you do each week in your greenhouse. All may be automated by AI; however, AI cannot understand tasks that require creative thinking, or make decisions based on unspecific information, or make decisions in unfamiliar contexts.

This is why AI will not steal all our jobs right now. You should think of AI tools as being your digital assistant that makes you more efficient; the AI assistant will do boring, time-consuming, repetitive tasks so you can focus on more important work.

Let’s look at some amazing examples in greenhouse farming.

Examples of AI in greenhouse farming

So let’s go back to the example of cats and machine vision. We can use machine vision to recognize images of practically anything. Using thousands of images of crops with diseases, a software learned the patterns of spots and scars on plants.

Now using a free app for your smartphone, you can take a picture of a diseased plant in your greenhouse and the software will be able to diagnose, or identify, the disease and then recommend a treatment, like a pesticide, to eliminate the crop disease.  

Here is another example for predicting your future harvests of vegetables like tomatoes and peppers.

Problems with harvest forecasts for tomatoes and peppers

One repeated task in greenhouse farming is harvest forecasting--knowing how much produce you will have each week. 

Today, your team of growers walk around the greenhouse counting under-ripe tomatoes and peppers and then consider current growing conditions like temperature and amount of sunlight. From these factors, your greenhouse team makes a rough estimate on how many tomatoes that week will be ripe for harvest.

This takes about 1.5 days of work for each week for each worker making the harvest estimate. That’s a lot of time. And the results are not that great.

Often errors in predicting weekly tomato harvests are more than 20% over or under the actual yield. This is an example from our client in California. This high error rate makes planning common business operations difficult.

With such variability:

It is hard to know how much labour and packaging materials you need.

It is hard to know if you will not have enough produce to meet contractual agreements with your buyer.

It is hard to plan promotions and deals for your produce.

Is there a better way to predict future harvest yields? Yes.

Motorleaf automated harvest forecasts by AI

Remember, AI is really good at automating repetitive tasks. Harvest forecasting is a repetitive task, so it’s perfect for automation.

At Motorleaf, this is what we do: we make AI tools that predict the future yield of tomatoes and peppers in commercial hydroponic greenhouses. And our AI service is better at doing the task than human being. 

Here is a comparison. Within months, we were able to reduce the error in predicting their tomato harvest by 50%. After one year, we reduced the error by over 70%. How is that possible?

It actually learns and gets better

Remember: AI technology does get smarter over time when trained with more data. That means it can become more accurate and able to make harvest forecasts further into the future.

First: getting smarter

This graph is for the same client in California. They have used our technology for over year. You can see that with each harvest cycle, and more data from growing conditions and their weekly harvest yield, the error in predicting harvest yield got less and less while the manual harvest forecasts--here in pink--remained volatile. Our average error in predicting weekly harvest is now about 8% for this client, which is a first in history.

Second: further into the future

And now trained with more data, we can see further into the future.

Here you can see in pink, the actual harvest yield for this same client in California. And the green line is the yield our AI technology predicted 2-weeks in advance.

This accuracy for 2-week predictions is also a world first.

AI automated harvest forecasts are more accurate

So why is AI so much more accurate than humans?

Manual harvest forecasts done today by humans is inaccurate because we only take into consideration a few factors to estimate the ripening of fruit.

We know that many factors influence the growth of plants: temperature, light, humidity, amount of CO2, water temperature, nutrients and more. A human being cannot analyse all these factors at the same time. A human being also cannot calculate how one growth factor will influence another factor.

So for example, we know that high temperature will increase plant growth. But what if you have high temperature, little water and some CO2. How do all those conditions affect plant growth. That is too complex and will take too long to calculate.

Where does the data come from?

Remember that AI needs to be trained with a lot of data and information in order to become smart.

For some AI services and for some AI companies, you will need to collect data in your greenhouses using sensors and cameras for months to years before you have enough data for training.

In other cases, you probably have enough historical data already.

I cannot explain all the data we use in detail because this is a company secret. But at Motorleaf, we use your past history of harvest yields and information from your growing equipment, such as your irrigation system and heating system. Since you already have this data avaialble, we can develop our AI services within weeks.

It is important to note that since we use data from your greenhouse, the AI technology we build for you will only work in your greenhouse because a greenhouse in another area of the world has different growing conditions.

Benefits of AI harvest forecasts

AI will bring many benefits to farming.

In terms of automated harvest forecasts:

They are more accurate.

You can know how much produce you will have so you can then better plan your business operations, better plan labour and packaging needs.

You can foresee when you will have a surplus and better plan promotions to sell it fast.

You will know when you produce too little and can plan how to buy the missing produce so that you meet contractual agreements with your buyers and wholesalers.

You can fully automate harvest forecasting. Forget this boring work! This will free up a day and a half of work for your agronomists; they now can focus their time on developing new plant varieties and ensuring you produce the best tomatoes and peppers possible.

Future uses of AI in greenhouse farming

We can build upon our AI technology to automate other important tasks in greenhouse farming. Here are a few examples of EMERGING services under development at Motorleaf. So, here is a look at the future.

First, Automated disease scouting

Environmental stress like high temperatures and dryness make plants vulnerable to disease. We have observations for tomatoes that show our AI technology can identify areas in a greenhouse that are under stress and are thus at risk for insects and parasites. This opens the possibility to automate disease scouting in a greenhouse; it also enables greenhouses to intervene early to stop an infestation using fewer pesticides.

We are currently looking for more greenhouses to test this technology.

Like with harvest forecasting, we can analyse growing conditions to know when is the best time harvest cannabis so it contains the highest concentration of THC and CBD.

We are also analysing how growing conditions affect the quality of tomatoes and peppers. For example, We aim to be able to predict the size of fruits and conditions that cause them to become imperfect or have blemishes.

Their are so many possibilities; and we are just at the beginning of using this technology in greenhouse farming.

Concluding slide

And there you have it.

Artificial intelligence is just sophisticated software you can use to automate common, specific, repetitive tasks in your greenhouse.

With AI, we can now make complex calculations and analyses of growing conditions. This enables us to predict the future and reduce uncertainties about your crops.

Some technologies will work anywhere, like crop disease diagnosis using machine vision. Other AI services are custom-made and will only work in your greenhouse.

We can collect data from your greenhouse using sensors and cameras; or use historical records you already have.

So don’t wait! Join the AI revolution in agriculture today.

Do you want to learn more about how AI can automate operations in your greenhouse? Tell us a bit about your farm in this form and we will contact you in a jiffy.

Submit Form

Topics: ai, greenhouse, tomato, agricultural technology, automation, yield prediction

Jason Behrmann

Written by Jason Behrmann

Subscribe to Email Updates

Recent Posts