Apparatus and method for classification of backscattered full-waveform signals
The proposed method aims at classifying backscattered signals (e.g. airborne LiDAR full-waveform data) through a two-steps procedure, that allows to exploit both the raw signal, the spatial position and the geometric relationships among near points. Thanks to the employment of a classifier and an algorithm of segmentation, an accurate and completely automatic classification is obtained.
The patented method in the first stage, raw waveform data are given as input to a classifier that outputs the probability that the analyzed input belongs to a certain class. In the second step of the proposed procedure, waveform data are mapped into a two-dimensional image, where every pixel stores the probability distribution vector, provided by the classifier employed in the first stage of the procedure, and the height of the data falling in the pixel. Then a segmentation algorithm is used to partition the image, assigning a label to every pixel in the image such that pixels with the same label share common properties. Deep learning methods can be used to perform this task. The method allows to get an automatic classification that in comparison to the actual technologies improves the accuracy and it allows to identify with precision, also objects with a small surface (as for instance the cables of a power line).
- Remote sensing.
- Completely automatic classification;
- Performances superior to the state of the art;
- No requirement of user-defined parameters that can influence the result of the classification.