Automatic Bow Echo Shape Detection Using Combination of Image Processing and Skeletonization in X-Band Radar Image

Image acquisition from the rain scanner radar Santanu could capture the bow echo shape that manifestation of heavy precipitation and strong wind. An alternative method to automatically detect the bow echo using image radar has been proposed in this study. The method combines image processing and skeletonization. Image processing is used to identify all the objects, restructure the elements, label, and calculate their properties including the estimated length and eccentricity. Skeletonization is used to represent the basic shape of an object, combined with curve angle detection to estimate the bending factor of an object. The proposed method has been tested using 80 radar images from the observation period of 2020 until 2022 in the Bandung area. Statistical analysis is carried out to obtain the threshold of curve angle and eccentricity. From six combinations of this threshold, 41 images (51.25%) of bow echo can be detected automatically. The threshold that affects the detection result is the lower threshold of curve angle and eccentricity. However, this setting gives a more realistic result because lower curve angle and lower eccentricity value will outcome a V-shape object detection which is undesirable.

I. Introduction

Santanu (Sistem Pemantau Hujan Spasial) is a Rain Scanner Radar (RSR) system that was developed based on a marine X-band radar [1]. The purpose of the RSR is to observe rain conditions every 2 minutes with a spatial resolution of 120 meters. RSR consists of hardware and software systems. The hardware system is composed of a slotted waveguide antenna, transceiver, unit display, signal conditioning, analog-digital converter (ADC), and a computer for signal and data processing. The software system consists of signal acquisition and display routines. The result from the signal acquisition is fed into the display routine in the form of an image and matrix file. This image acquisition file is a 734x734 pixel that contains a reflectivity map (in dBZ) that represents rain information in a radius of 44 km.