Automatic Detection of Blobs in WISPR/Parker Solar Probe Data Using a Machine Learning Approach

The WISPR instrument onboard Parker Solar Probe (PSP) has provided unprecedented observations of the solar corona, revealing fine-scale structures with exceptional spatial and temporal resolution. Among the most prominent features observed are circle or oval shaped transient density enhancements, commonly referred to as blobs. WISPR images are densely populated with these bright, quasi-circular features. We apply a machine learning (ML)–based approach for automatic blob detection, to handle variations in blob size, brightness, and image background complexity. When applied to multiple PSP encounters (E1-E24), this method reveals a clear increase in the number of detected blobs with decreasing heliocentric distance, in agreement with expectations from coronal dynamics and density dropoff. In addition, we find a significantly higher number of blobs in the aftermath of coronal mass ejections (CMEs). The structures can originate from different physical processes including tearing instabilities at the post–coronal mass ejection (CME) current sheets, interchange reconnection in the corona and magnetic reconnection between flux ropes and the ambient solar wind. This ML-based approach enables robust blob detection across varying observational conditions and provides new insights into the spatial distribution and evolution of coronal density structures in the near-Sun environment.