Real-time crop monitoring: a new vision for agriculture.
Two weeks ago I was at the Sensors in Food and Agriculture conference at the John Innes Centre, Norwich, where sensor developers and agronomists met to share developments from both sides of the field. I was there to talk about the multispectral imaging platform that we have developed here at Cambridge Consultants, and how it could be used for agriculture. While there I heard about all sorts of interesting ways sensing is starting to make an impact in agriculture, but there was no doubt that multispectral imaging was a key theme.
I raised my eyebrows when I saw on the program a talk entitled ‘3D multispectral imaging systems for near range plant and fruit trait assessment’ from Fotenix Ltd. They have developed a system which uses an off-the-shelf (2D) multispectral camera, in tandem with a structured light system (like those found in the Xbox Kinect) to obtain multispectral information at the same time as depth information. These two data streams can be reconstructed into a ‘3D’ multispectral image. This solution is a great insight into how we are likely to see data fusion being used with multi-sensor systems to create a more detailed and informative picture of their environment. A neat solution, but it needs a killer application where that additional information provides real additional value to show whether it’s a useful step forwards.
My favourite talk of the day (and not just because it concerned insects rather than wheat fungi…) came from FaunaPhotonics ApS. They have developed a system that detects flying insects by shining an invisible laser across a crop field and looking at the tiny reflections that return from the insects. This is much the same as how you can see insects in the headlight beam of your car. But the really impressive point is that not only can they count the insects using this method, they can also go some way towards classifying the species by measuring the wing beat frequency, the speed of the insect crossing the beam, and a basic (2-channel) colour measurement. Systems like this open up some really interesting opportunities for continuous, non-intrusive monitoring of pest populations which would otherwise be extremely difficult. You know something’s cool when you wish you’d developed it yourself.
Speaking of which, despite the multiple talks on how multispectral imaging can be useful in agriculture, Cambridge Consultants were the only ones there offering a new approach to the actual hardware demand. Our system switches the burden of complexity away from the camera by using controlled illumination to achieve the multispectral input, with a basic monochrome camera recording the images in real time. This approach not only dramatically reduces the hardware cost, but is also quick enough and data-light enough to be used at real-world speeds, for example attached to a tractor. This is all about having an appropriate tool for the job- you could achieve far better spectral resolution and range with an expensive, top of the range hyperspectral camera, but you’d have to justify collecting all that extra information, at a much slower speed. If you know what you’re looking at and what you’re looking for, you should be able to identify a limited number of spectral bands where the information actually is, and avoid wasting your time (and money) collecting the rest. For more information about real-time crop monitoring you can download our whitepaper here.