Polarimetric information to enhance synthetic aperture radar automatic target recognition capabilities

This chapter has investigated the potential benefits of the use of polarimetry when dealing with ATR. CTD is particularly suitable to represent man-made targets and is physically interpretable. Moreover, it was shown how a particular decomposition belonging to this family of polarimetric decompositions can improve the performance of ATR frameworks. In particular, it was shown how the roll-invariant Krogager decomposition could be easy and effective to integrate in a feature-based ATR framework exploiting invariant image moments. The results presented demonstrate that the use of polarimetric information helps to enhance target-recognition capabilities as well as reduce data collection requirements for classifier training purposes.

However, a number of research questions still exist in this area, such as the selection of the best polarimetric decomposition for ATR purposes, or the possibility to design a custom ATR-oriented polarimetric decomposition with the aim to extract the most relevant information for the target-recognition task. Another topic to investigate is the necessity to differentiate approaches based on the level of target classification required. For example, it could be useful to assess which type of approach/decomposition would be more suited for different classification levels. As often only two polarimetric channels are available, it would be interesting to understand how much of the polarimetric information that supports the ATR task can still be extracted. A different perspective would be to investigate the latest advances in artificial intelligence with CNNs applied to the ATR task, and in this sense, recently some progress has been made [40, 41]. However, these techniques have not reached a sufficient level of maturity in the field and more research effort would be required in the future. The above are likely to be explored in the future by the research community, broadening the interest on this relevant application domain and strengthening the use case for polarimetric SAR sensors.