Data records and outputs

Selecting a compatible format, such as GeoTIFF*, ensures seamless integration with analysis tools. Prioritising shareable formats facilitates collaborative sharing of findings. Adhering to standardised formats, following industry norms like those set by the Open Geospatial Consortium, promotes consistency and compatibility. Accurate error measurement and assessment of data accuracy are essential, with methodologies detailed in comprehensive review papers.

Finally, implementing robust data storage practices, like redundant systems and cloud backups, safeguards against potential loss, ensure the accessibility and reliability of UAV-derived data. *GeoTIFF is a public domain metadata standard which allows georeferencing information to be embedded within a TIFF file.

It is also important to collect appropriate metadata on the imagery collected at the time of capture where possible. Datasets automatically have their own metadata that will be recorded, and this can be accessed later via suitably scripted code. However, for weed detection, some of this data is not captured automatically and it is very useful for later processing and for keeping consistent records to collate this data at the outset. Such data should include where relevant:

• Biological information, such as date, plant name, growth stage, soil and surface characteristics and GPS coordinates.
• Data capture details, such as sensor make, lens, and lens focal length, sensor height, angle, ground speed. Other data may include data capture conditions, such as weather, seasons and light availability
• Dataset generation details, such as label type, description and values, the number of labels and crops generated (including crop settings) and the number of images and instances of target plants in the images per class.

This data for each dataset could be collated in three small tables, similar to the following:

Biological Data

Capture Details

Dataset generation details