Convolutional neural network for Iceberg Classifier

Image classification has been a popular research field in computer version, and we employ convolutional neural network to solve a practical problem that is iceberg classifier. Iceberg classifier is critical import to keep the ship safe in the ocean, because the ships can program another routine after they detect the iceberg. When radar detects an object, it cannot separate it from ship or any other solid object. The radar images contain the shape, size and brightness features. There are several ways to solve the iceberg classification problem: (1) Hough transform [1] (2) convolution neural network [2]. Convolutional neural network can obtain image features that are invisible for human beings. Convolution neural network has achieved expected performance in many image classification problems. Our experiment shows that convolution neural network can classify the iceberg from any other object and we get accuracy 96% and F1 score 96%. Support vector machine [3] achieves accuracy 79% and F1 score 78%.