Practical application of digital imaging in agricultural systems

Author: | Date: 21 Jun 2016

Take home message

A picture is worth 1000 words.


As the technology revolution and in particular the digital revolution increases data and knowledge exponentially, sensing technologies are popping up everywhere.

Satellite technologies once the realm of the military and extraordinarily expensive are now becoming progressively more available, accurate and progressively cheaper.

UAVs likewise are becoming increasingly more high-tech and able to carry much more sophisticated sensors and perform autonomous operations at the same time becoming progressively cheaper.

Ground-based sensors and particularly the communications systems between them and back to a home base are also developing exponentially.

A few basics

Types of sensors

There are a myriad of sensors providing information or imagery. These include visible, red edge, infrared, radar, thermal, lidar, EM38 to name a few. I will mainly talk about the near infrared or NIR.

  • Imagery comes back to earth as digital numbers provided by sensor (DN)
  • Taking into account sun angle, atmosphere and sensor specs, DNs are converted to reflectance values
  • The reflectance value is used to make images (true and false colour), and generate other information such as vegetation indices
  • The sensor ‘looks’ at different bandwidths
  • Blue, red and green bands which our eyes can see
  • Near infrared, short wave infrared, thermal bands which our eyes cannot see.

Vegetation indexes

Satellites and most UAVs sensors see the world in distinct wave length bands. Live vegetation reflects in the near infrared but little in the visible. The amount of reflection in the near infrared differs between species and also depends on plant health. Vegetative indexes are used to show the differences and the most commonly used Normalised Difference Vegetative Index (NDVI) compares the difference between near infrared and red to the sum of the two. There are many different vegetative indexes which can be used for specific comparisons. For example Satamap uses the MTVI2 index to try account for some soil colour and reduce saturation in high biomass areas.

Pixel size

Pixel Size ranges from

  • Vast areas such as the 20 km pixels used by DHMM which are useful on a macro scale and in extensive areas
  • Modus which is used by programs such as Pastures from Space which has a 250 m pixel but passes over every day and in the Pastures from Space application allows for a maximum chance of getting enough usable images every week to calculate average pasture growth et cetera.
  • Landsat 7 and 8 which have 30 m pixels but also has a panchromatic 15 m band. This allows pan sharpening where the 30 m pixels are combined with visual (and NIR) bands to produce higher spatial resolution images. Landsat 7 & 8 pass over every 16 days alternately, i.e. an image is available for every spot on the earth every eight days.
  • Sentinel 2A is the new European Space Agency satellite and has a 10 meter pixel. This allows enough accuracy to do most of what is required in broadacre cropping.
  • Landsat and Sentinel both provide free data meaning that the data provider such as Satamap, PA source, Geosys, PCT et cetera accesses the raw data for free and you are paying for the processing and storage of the data and subsequent conversion of that data to usable information.
  • There are a range of commercial satellites which provide information to increasingly more accurate pixel sizes such as Rapid Eye 5m through to World View 3 at 0.31 m in natural colour and 1.24 m near infrared. There is a cost to accessing this data which is reflected in the commercial offerings.
  • UAVs can operate down to 1 or 2 cm pixel size. This depends on the sensor (camera) and the height at which it is flown.

What are the advantages and disadvantages of various systems?



  • Cost-effective
  • Remote access
  • Rapidly increasing sophistication
  • Increasing number of shots available


  • Cloud- both actual and shadows
  • Not necessarily available when specific events occur such as hail, spray drift, specific growth stages for fertiliser application et cetera.
  • Expensive once you get into very small pixel sizes due to having to sequence the high resolution satellites and minimum order is often around 2000 ha. 

The future

  • Figure out what to do with Sentinel 2 ‘red edge’ bands
  • Use sensors that see through clouds
  • Planet Labs – 100+ micro satellites to capture entire Earth every day at 3-5m resolution



  • Very small pixel sizes
  • On Demand i.e. can time the flights to the events or crop stages required
  • Some are relatively cheap to buy and to operate and getting cheaper
  • Can operate in variable cloud by timing flights
  • Can have multiple sensors and interchangeable sensors
  • Are extremely efficient at analysing field trials where vast numbers of plots need to be analysed


  • Limited coverage and flight times
  • Can crash
  • Can be taken down by eagles
  • Still further to go with stitching programs and/or geo referencing
  • Legal and CASA obligations

The future

  • Fully autonomous
  • Accuracy down to levels where crop emergence can be assessed
  • New sensors and new ways to analyse the data will provide another massive amount of different information

Ground-based sensors

As I have not really discussed ground-based sensors I will give a few examples. The BOM radar would be a classic example where BOM turn the raw data into rain intensity and then other applications use this data to work out actual rainfall on specific areas.

One of the building blocks of the whole Climate Corp data system in the United States is the Doppler radar network which is very extensive and accurate allowing a very accurate analysis of rainfall and soil moisture on a sub paddock scale.

Yield data available from headers is built up into digital images from paddocks.

There are a range of tractor/vehicle mounted sensors which read nitrogen status in crops and build up a paddock image of the nitrogen status while conducting other operations such as spraying.

There are weather stations which use raw weather data and inputted crop/soil data to build up maps of plant growth and soil moisture.


  • Can be linked to provide area wide data
  • Can be linked outside the traditional communications channels
  • Weight is no limitation so any size sensor can be mounted


  • Enormous diversity of sensors giving different information and sometimes not talking to each other
  • Is usually performed with another farming operation but if performed alone would be potentially expensive and unnecessary tracking in crops.

The future

  • Sensors can be cheaply linked to provide area wide data in conjunction with other digital information.
  • Sensors for disease, insects, weeds and a whole variety of nutrients are coming.

A picture is worth a thousand words but what do we do with it?

Some of the practical uses of digital imagery are

  • Tracking crop growth within seasons and compare crop growth between seasons
  • Yield forecasting
  • Readily identifies areas of average, better than average and below average growth so that inspections and solutions are more targeted
  • Identifying resistant weeds
  • Creating variable rate maps
  • Irrigation scheduling
  • Identifying inaccurate irrigation in areas/zones
  • Emergence and or gappiness maps particularly for replanting
  • Identifying better field layouts and areas that are not profitable to farm
  • Benchmarking across localities and regions
  • Tracking previous crop performance when purchasing properties.
  • Targeted insecticide applications
  • Targeted fertiliser applications
  • Harvest timing
  • Overall tracking of crops and yield forecasts over farms spread across regions or across Australia
  • Picking trends and successful strategies across clients
  • Understand crop areas and crop status to help with input forecasting
  • Understanding crop dynamics and tracking hail/storm/drift events
  • Tracking pasture growth and utilisation
  • Crop forecasting and resource management
  • Crop growth and yield forecasting on a regional basis and an  Australia wide basis
  • Using historical data to assess trends
  • Tracking fertility especially different fertiliser regimes and the effect of previous crops and the waterlogging et cetera
  • Formulating top dressing strategies
  • Mapping out areas of high growth for fungicide and growth regulator application
  • Tracking herbicide damage and off target drift
  • Assessing trials
  • Assessing potential paddock yields for marketing strategies and insurance purposes

Some holy grails

  • To create emergence maps to allow replanting or spot replanting in a timely fashion
  • To identify various crops from their digital signature
  • To identify weeds within crops
  • Seamless application of digital imagery into usable data such as variable rate maps

Useful resources

Landsat Spectral Characteristics viewer


Precision Ag Technologies

PA Source


Contact details

Peter Birch
B&W Rural
Ph: 0428 669 157