Article 4. Performance of target detection.
There are very small statistical data about performance of target detection.
Anyway GPR gives some information about underground world; it is better than nothing.
But often it is not simple to make interpretation to identify objects and structures.
GPR image gives only the picture of electromagnetic waves distribution. Waves have reflections, scatterings, attenuations.
GPR image does not contain other soil features (for example, humidity). We can guess about some of them looking at changes
in wave forms. We need experience to do so because not all information are presented in GPR data.
In the real world types and positions of buried targets are hidden from the operator.
Sometimes there are hints on the surface to explore some spots more carefully.
GPR survey can be a precursor of geological or road works. GPR results can be compared with those obtained by borehole data.
The GPR operator may define levels of confidence in classifying targets. Depths of targets are depended on soil properties.
It is similiar to radar operator task to determine whether a blip on the radar screen is an object, clutter or noise.
Probability of Detection (PD) and the Probability of False Alarm (PFA) are key parameters in evaluation of GPR performance.
Performance of target detection is depended not only on device features and specifications.
Certainly, they are very important for obtaining of good source data.
Also in many cases experience of operator has big value.
Modern state of GPR technology still demands high level of operator skills.
In common case GPR data do not contain full information about underground soils so you need some extra information for interpretation.
For some jobs the requirements to the operator can be lower.
For example, GPR surveys of concrete walls give limited types of reflections. Though the GPR image can have a lot of reflections
(in the case the number of included objects is big), you have preliminary knowledge about possible objects and their positions.
Advanced software tools for data processing help operator to extract information about objects and structures and visualize them.