Performance Comparison of Shape Fitting and Moments-Based Techniques for Detection of Convective Cloud Location

Mesoscale convective systems (MCSs) represent an aggregation of storms extending hundreds of kilometers. MCSs are one of the significant factors for precipitation over tropical and subtropical regions of the Earth. They also cause severe weather conditions such as floods. Therefore, the identification and prediction of such events can be useful for informing people about their occurrences. MCSs are made up of cold convective clouds representing infrared brightness temperature below 220֠K. One of the important steps to get the location of these convective clouds is to get the center of the detected clouds. A cloud is a deformable object that changes its shape, which is irregular, leading to a non-convex polygon. So, getting its location is a challenging task. In the proposed work, two techniques for getting cloud center location are implemented: approximation by regular shapes fitting and moments-based technique. Infrared Brightness Temperature (BT) data from the Kalpana-1 satellite is used for the study. In the first technique, different shapes, such as rectangles, ellipses, and inner circles are fitted. In the second technique, spatial moments of the contours are used to get the center of the clouds. The accuracy obtained by an inner circle fitting, ellipse, and rectangle is 80, 91 and 92% respectively. The accuracy obtained by spatial moments of the contours is 94%. Performance comparison of the results showed that spatial moments of the contours outperformed shape-fitting techniques for the detection of cloud location.