EISCAM — Digital image acquisition and processing for sea ice parameters from ships

详细信息    查看全文

• 作者：B. Weissling ; S. Ackley ; P. Wagner ; H. Xie
• 关键词：Sea ice ; ASPeCt ; SIMBA ; Antarctica ; Ice camera
• 刊名：Cold Regions Science and Technology
• 出版年：2009
• 出版时间：June 2009
• 年：2009
• 卷：57
• 期：1
• 页码：49-60
• 全文大小：1092 K

文摘

Human observations and records of sea ice conditions from ships of opportunity in polar regions have been formalized under ASPeCt (Antarctic Sea Ice Processes and Climate) protocols as part of the Scientific Committee on Antarctic Research (SCAR) initiative, Global Change in Antarctica. Human observations, however, are subject to both bias and level of expertise. In support of ASPeCt observations, a ship-based, ice condition imagery acquisition, processing and analysis system is developed and presented in this paper. Digital images captured from recorded oblique-view video of sea ice during the inbound and outbound transits of the icebreaker NB Palmer, during the 2007 SIMBA (Sea Ice Mass Balance in Antarctic) cruise, were used to demonstrate the method. The resultant orthorectified and spatially registered images can thus be used to derive ice concentration, ice types, floe sizes, and area of deformed ice, parameters recorded by an ASPeCt observer based on experience. An assessment of total ice concentration, utilizing a digital number threshold method, for a 30 image sequence spanning 10 min, was compared to the corresponding ASPeCt observation and found good agreement. A k-means unsupervised classification technique was employed to evaluate the spectral separability of pixels associated with differing ice types. A significant source of visual perspective bias was discovered in associated with uneven spatial distribution of open-water leads. Convolution spatial filtering was employed to assess area of deformed ice during night-time conditions, when ship's floodlights illuminate the ice field. Transformation of individual image elements, such as floes, to vector-based polygons enabled accurate measurement of dimension, thus providing a frequency distribution of floe sizes, currently not possible with existing ASPeCt protocols. Limitations of this acquisition and analysis technique, future improvements, and coordination with the ASPeCt program were discussed.