Ship detection based on SVM using color and texture features

Abstract

Nowadays, many applications related to maritime security and ship monitoring require a correct detection of ships. In the field of ship detection, different types of images are used depending on the application. Regarding high-resolution images, the variable characteristics of the sea environment often complicate a precise detection. These characteristics make the extraction of general properties from individual pixels difficult. To overcome this issue, a block division that divides the image into small blocks of pixels which represent small ship or non-ship regions is proposed. In contrast with a pixel approach, this block division characterizes better the properties of the regions and is more computationally efficient. For the classification of blocks, a supervised learning algorithm Support Vector Machine (SVM) is trained using color and texture features extracted from the blocks. On one hand, color features describe the chromatic characteristics of these regions. On the other hand, texture features provide information about the spatial distribution of pixels. Once the classification is performed, ship detection is improved using a reconstruction algorithm, which corrects most wrong classified blocks and extracts the detected ships. The combination of color and texture features achieves the highest precision, up to 96.98%, in the classification between ship blocks and non-ship blocks, and up to 98.14% in the final ship detection.