Improved peak detection at high particle image densities using single stage CNNs

Abstract An important step in the application of Lagrangian Particle Tracking (LPT) or in general for image based single particle identification techniques is the detection of particle image peaks on the measurement images and their sub-pixel accurate position estimation.In case of volumetric measurements, the peak detection constitutes the first step in the process of recovering 3D particle positions, which is usually performed by triangulation procedures. For two-component 2D measurements the peak detection results directly serve as input to the tracking algorithm.Depending on the quality of the image, the shape and size of the particle images, and the amount of particle image overlap, it can be difficult to find all, or even only the majority, of the projected particle peaks in a measurement image. Advanced strategies for 3D particle position reconstruction, such as Iterative Particle Reconstruction (IPR) are designed to work with incomplete 2D peak detection abilities but even they can greatly benefit from a more complete peak detection as ambiguities and position errors are reduced. We introduce a convolutional neural network (CNN) based peak detection scheme that significantly outperforms current conventional approaches, both on synthetic and experimental data, and enables peak detection with a vastly higher completeness even at high particle image densities.