Convective Cloud Detection and Tracking from Series of Infrared Images

The most significant part of prediction of precipitation is the detection and identification of convective (cumulonimbus) clouds, also the tracking of cloud movement is important for identification of location of precipitation. A very simple methodology for detecting convective clouds and then tracking its movement from a series of infrared (IR) images is proposed in this paper. IR image is segmented using k-means clustering algorithm, which has been implemented using Euclidean, Manhattan and Mahalanobis distances and the results have been compared. Cloud clusters have been identified from segmented image and subsequently the large clusters were extracted. Center of Mass (CoM) was calculated for each selected cloud cluster and its position after every 30 min was predicted and compared with the actual values. If the predicted position deviates, the proposed models automatically adjusts itself, and the next prediction becomes closer to original values of position.