Multispectral Training

Back at Cranfield for some drone training, this time with the Multispectral DJI M3M. A few of us from UCL East (CASA, CBER, Robotics) spent a day with Ben Crisford from DPA (and Crossbill Forestry) learning the basics of Multispectral imaging in the morning and then out flying and data processing in the afternoon.

Some really good insights into reflected light in different spectral bands and in particular the four used on the M3M - Green (550nm +-16), Red (650nm +-16), Red Edge (730nm +-16), Near Infrared (860nm +-26) - and how the combination of the reflected radiance from each sensor can be used calculate the status of vegetation. For example, chlorophyll in plants reflect more Near Infrared and represent a healthy leaf.

A key part of the practical application of Multispectral imaging seems to be selecting the appropriate visual / vegetation indices for the task at hand. We mainly discussed NDVI ( NDVI=(NIR-Red)/(NIR+Red) ), NDRE and when to use in combination with RGB - a long list of indices used are at https://www.indexdatabase.de

Some sample images from one shot using the 5 different sensors during the flight are below:

RGB

Green

Red

Red Edge

Near Infra Red

After my previous challenges with getting the Matrices 4E to follow terrain while maintaining a low GSD it was really handy being able to get Ben’s feedback on approaches he uses to automate the terrain following - various notes made and was good to see that practicalities of implementing these approaches in the field later in the afternoon. Need to go and look at downloading the 1m resolution digital surface model from the UK national lidar programme.