Methods of associating CODAR seasonde vessel detection data into unique tracks

Since 2011, Rutgers has been operating an over-the-horizon vessel detection software called `PeakPicker' at two 13 MHz CODAR SeaSonde sites located in Sea Bright, NJ and Belmar, NJ. The challenge of vessel detection using High Frequency (HF) radar is dealing with false detection peaks and then associating multiple peaks derived from different coherent integration times and background filters at a single time step into one peak representing the best estimate. Association of distinct signal peaks derived from the CODAR SeaSonde high frequency (HF) radar PeakPicker software provides a challenge. Association of peaks occurs in two distinct phases. Level 1 Association will associate multiple detection peaks derived from different coherent integration times and background filters at a single time step into one `best' peak while Level 2 Association will combine data from multiple sites representing different viewing angles. A Matlab script, plot_all_matches, finds matches target data in range and range rate (radial velocity) with AIS data providing a ground truth of our ship detection progress. Utilizing the output of the matching data from plot_all_matches, we present techniques that increase the accuracy and decrease the error of detections.