The True Value of Low Overlap Drone Imagery


We were recently asked on our Facebook page to respond to a blog post from MicaSense that attempted to compare the performance of “low overlap” and “standard overlap” multispectral imaging systems for agricultural drone applications. Per the post, their RedEdge system falls into the standard overlap group, typically requiring at least 70% overlap between adjacent images to form completed data products. Our own SlantRange systems, like our 3PX multispectral sensor, while not mentioned directly in the post, fall into the low overlap group by requiring only 20% overlap between images.

While the post does provide some valid information, the conclusions of the analysis, which state that flying at low overlap “sacrifices accuracy, resolution, and valuable outputs to save an extra five minutes in the field”, are misleading and incomplete.  There are many differences between these systems and their capabilities, some of which were covered by MicaSense, however a number of important topics should be more accurately characterized.  This is the first in a multi-part series aimed at clarifying those differences.

Image Resolution as the Relevant Point of Comparison

If one's intent is to make an honest comparison of the performance of imaging systems, one should at least start from the point of comparing parameters of relevance to the user.  MicaSense uses a common flight altitude of 120 m as their basis of comparison.  However, the post fails to recognize that when the two systems are compared at the same altitude, their imaging performances are dramatically different. 

A more relevant basis for comparison, and one commonly accepted within the industry, would be the imaging systems' spatial resolution (or, how small can objects on the ground be and still be detected by the system).  Spatial resolution is a significant driver for the value of any data products that might be derived from the system and consequently a far more relevant metric than flight altitude.

So let's look at a typical working scenario to illustrate.  Let's assume your objective is to isolate individual plants for measurement, measure the fraction of growing vegetation to bare soil, and classify some weeds from the crop population.  To make these measurements we'll further assume you need 4 cm resolution.

To achieve 4 cm resolution, the MicaSense RedEdge should be flown at about 60 m altitude. And here is the very important point that was missed in the MicaSense blog post: the SlantRange 3p has nearly twice the spatial resolution of the RedEdge and can achieve 4 cm resolution from an altitude of about 100 m. (Note that resolution also includes the effects of motion blur, optical distortion, registration artifacts, etc. but we'll assume these are equal between the systems for now for comparison.)  Why does this matter?  Because it actually has very big consequences for flight collection time.  Here are the numbers:

  MicaSense RedEdge SlantRange 3p
Required Data Product Vegetation Fraction Vegetation Fraction
Required Resolution 4 cm 4 cm
Field Size 100 acres 100 acres
Camera Horizontal Field of View 48 degrees 28 degrees

Flight Altitude Required to Achieve Required Resolution

60 m 100 m
Camera Horizontal Ground Footprint 51 m 51 m
Flight Speed 15 m/s 15 m/s
Overlap 70% 20%
Flight Time 36 minutes 14 minutes

That's a 22-minute longer flight time for the RedEdge, more than twice the time required for the 3p!  And if you're flying a multicopter this means that more than likely you'll need to land and change batteries in order to complete the 100-acre field, and that time isn't even factored into the table above.

Data Availability is the Important Speed Comparison, Not Just Flight Time

The job's not done when the aircraft lands.  For the purposes of surveying agricultural crops, the important speed metric is how fast can valuable new information be in the hands of the decision-maker. Information on quickly changing crop conditions is highly perishable; its value goes down over time.

So while low overlap has obvious benefits in minimizing the time spent collecting data, the benefits are amplified when one considers the data management side of the workflow.  That's because a 4x reduction in overlap produces a 16x reduction in data volume.  And that means huge savings for you in terms of data storage, data management, and most importantly, data processing time. These savings can help you build a profitable agricultural drone service business.  Here are the numbers on data volumes and processing times for the same scenario:

  MicaSense RedEdge SlantRange 3p
Required Data Product Vegetation Fraction Vegetation Fraction
Required Resolution 4 cm 4 cm
Field Size 100 acres 100 acres
Overlap 70% 20%
Uncompressed Data Volume 21.0 GB (5 ch) 2.4 GB (4 ch)
Cloud Upload Speed 1 MB/s Not Required
Cloud Upload Time 5.8 hours Not Required
Processed Data Availability Next Day 10 Minutes

Flying your drone at low overlap not only has the benefit of reduced flight time, it has a tremendous effect on the amount of raw data collected.  So much so that this 100-acre field can be processed on-site within a few minutes with SlantView, our software analytics platform.  And what if you don't happen to have a solid, reliable, high speed internet connection as is the case for over 90% of the world's agricultural lands?  In that case, MicaSense Atlas really isn't even an option.

Of course, one could argue that image compression can reduce that data volume and shrink those upload and processing times.  However, meaningful image compression ratios come at a price, and that price is image resolution - at which point you need to remember your image resolution requirements for your data product and adjust accordingly (lower flight altitude, longer flight time, etc.).

Vegetation Fraction.png

Incidentally, if you're wondering what "Vegetation Fraction" is as a data product, it's a valuable metric for determining how thin a field's vegetation is, it's level of maturity, and is a common proxy for biomass and an early indicator of yield potential.  And it's not available from MicaSense.  More on that, the other topics raised in the blog post, and what you can really achieve with high resolution drone imagery in our next post…

Want to learn how SlantRange can help your business?

Check out our website to learn more about our sensors and analytics, review our FAQ, and explore our online store. Or just contact us if you have any questions. We're happy to help!