Giant Benthic HD Image Feature Extraction and Size Estimation Based on Canny Algorithm

摘要

Deep-sea benthic image features are difficult to extract because of its large amounts of information, auxiliary light imaging, and complex environmental background. To solve these problems, current study presents an approach to get texture information of sponge image captured by Jiao Long DSV. First, linear grayscale transformation is used to remove the seamount background and enhance contrast based on the image histogram analysis. The noise introduced by the suspended particles impurities is suppressed by median filter. Subsequently, compared with Prewitt and LoG algorithm, Canny operator is sure to get better edge extraction. Sponge texture information is most complete and noise is further reduced. Finally, mathematical morphology processing is carried out to perfect the texture by connecting intermittent textures, and the size estimation of the sponge based on hypothetical laser ruler is reliable and applicable.