Something GISsy

Imagery Resolution: Detection, Recognition and Identification

GIS users often use imagery to capture data. Most users are familiar with the spatial resolution of a sensor being defined by a measure called GSD (Ground Spatial Distance) or pixel size, eg. 50cm resolution is better than 1m resolution. 50cm is considered “higher resolution” as more detail is visible.

However there are other terms used in remote sensing that users should be aware of – specifically Recognition and Identification.

Consider the low resolution image to the right. Not really useful, we could conclude there is something “Red”.

Detection = red object

 

Increasing the resolution we can now conclude that this is a red motor vehicle.

Recognition = red vehicle

We cannot tell if this is a ute or sedan so Identification is not possible yet.

Increasing the resolution further we can see that this is a red sedan.

Identification = red sedan

 

 

As resolution improves cost goes up, so it is important to determine the purpose of the imagery Identification versus Recognition. In the above examples above do we need to count cars in the carpark (Recognition) or do we need to know what type of vehicle they are (Identification)?

Guidelines for Detection, Recognition and Identification

Detection = object size/3 = GSD – eg. car 5m/3 = 1.5 – 1.6m GSD required to Detect

Recognition = object size/10 = GSD – eg. car 5m/10 = 0.4 – 0.5m GSD required to recognise

Identification = object size/21 = GSD – eg. car 5m/21 = 0.1 – 0.2m GSD required to identify