Extracting the Ocean Surface Feature of Non-linear Internal Solitary Waves
in MODIS Satellite Images
Chih-Chung Kao
Fortune Institute of
Technology,
Kaohsiung, Taiwan
kcc7879@
cneter.fotech.edu.tw
Liang-Hwei Lee
National
Kao-Hsiung University
of Applied Sciences,
Kaohsiung, Taiwan
lhlee@cc.mail.kuas.
edu.tw
Chih-Chiang Tai
I-Shou University,
Kaohsiung, Taiwan
g9422012@
isu.edu.tw
Yung-Chung Wei
I-Shou University,
Kaohsiung, Taiwan
wei3504D@
isu.edu.tw
Abstract
The objective of this study is to integrate an effective approach to extract the linear features of oceanic non-linear internal waves in MODIS images based on the technique of Multiscale Retinex and linear feature extraction. To evaluate the performance of the integrated method, the extracted linear features were vectorized and overlapped with the original image in the Geographic Information System to investigate the localization accuracy of the extracted features, which were rarely seen before. The results indicated that the MSR method provided an enhanced image with improved color contrast and brightness, which resulted in a better quality of the extracted linear features. Finally, we compared the performance of two methods of linear feature extraction, the Canny edge detector and the Wavelet Transform Modulus Maxima edge detector. It is shown that the Canny edge detector is superior to the WTMM edge detector in terms of visualization quality and localization accuracy.
1. Introduction
Oceanic internal waves are waves traveling within the interior of the ocean, which may be generated by several sources including tidal currents, ocean frontal boundaries, and special atmospheric conditions [1]. When the internal waves flow over underwater obstacles, such as seamounts, shelf breaks, and troughs, they tend to become particle-like with large amplitudes within several kilometers, which are called internal solitary waves [7] or Non-Linear Internal Waves (NLIWs). One of the very early interests in studying these waves was their unexpectedly large shear stress
as they impacted on offshore oil-drilling facilities [2]. Thereafter, the effects of internal solitary waves on fishing activities, the safety of oil drilling platforms, and the navy’s antisubmarine warfare (ASW) have been investigated intensively.
Remote sensing imageries are capable of detecting the sea surface features (Figure 1) caused by NLIWs, and showing the spatial dynamics of the internal waves in a wide range of ocean area. Among the available satellite imageries, MODIS images of Terra and Aqua satellites provide a large spatial coverage with a high repetitive rate (i.e. twice in one day) in the visible band. Hence, MODIS images have been applied to investigate the dynamic of NLIWs recently. Our specific research interest is focused on the internal waves in the northern South China Sea with a goal to study the temporal and spatial dynamics of the NLIWs from a generation phase to a propagation phase and then to a dissipation phase. Integrating the extracted linear features of NLIWs from historical MODIS images using a geographic information system (GIS), the dynamics of the internal waves can be better understood.
Therefore, the purpose of this study is to integrate a proper approach to extract the linear features of NLIWs in MODIS images based on the techniques of image enhancement and linear feature extraction (LFE). In this study, the Multiscale Retinex (MSR) method will be applied to enhance the MODIS image and two edge detection methods will be used to extract the linear features of NLIWs in the enhanced image. To evaluate the performance of the integrated method, the extracted linear features will be vectorized and overlapped with the original image in the Geographic Information System (GIS) to investigate the localization error between them and the true features’ boundary.
