Solar Physics Automated Feature Detection: Progress and Scientific Return

The SDO Feature Finding Team (FFT) has been implementing 16 feature finding modules for the last two and a half years. These modules have been designed to analyze the incoming stream of SDO data in near-real-time. Several modules are in regular operation now, most others are reaching that point. Our modules detect flares, filaments, dimming regions, sigmoids, emerging flux, bright points, jets, oscillations, active regions, coronal holes, and several other solar features. We are also developing a general trainable feature detection module, which can be applied to detect any phenomenon. Automated feature recognition has several advantages over the same by humans: first, and most importantly, much larger amounts of images can be analyzed by machines; second, the codes will apply consistent criteria for the detection of phenomena, much more so than humans. Of course the second point implies that the detection criteria must be carefully calibrated, otherwise the outcome will be consistent, but consistently wrong. Examples of the scientific potential unleashed our project are: i) Draw a butterfly diagram for Active Regions, ii) Find all filaments that coincide with sigmoids, and then correlate sigmoid handedness with filament chirality, iii) Correlate EUV jets with small scale flux emergence in coronal holes, iv) Draw polarity inversion line maps with regions of high shear and large magnetic field gradients overlayed, to pinpoint potential flaring regions. Then correlate with actual flare occurrence. All of these tasks will be accomplished with great ease; the power of this method is limited merely by the imagination of the researcher. In addition our modules provide space-weather alerts for flares, dimmings (proxies for eruptions), and flux emergence. In my presentation I will present an overview of the output from our feature detection codes, as well as first results of scientific analysis from the metadata.