Information Retrieval Of The Flow Field In The Changjing Estuary Based On The Texture Features In The TM Remote Sensing Images

The Changjiang estuary is the largest tidal estuary in China, with a changing hydrodynamic environment and complex sediment transport system influnced by runoff, tide, wind, wave, Coriolis force, terrain and other factors. Satellite remote sensing technology could provide an effective way for regularly observing the macroscopic coastal and marine hydrodynamics and sediment transport with advantages of large scale observations, fast synchronized updates and so on. Quatitative information retrieval of flow field in the Changjiang estuary has never been done based on optical remote sensing images. The present dissertation focus on quantitative information retrieval of flow field based on the texture features in the TM remote sensing images in the Changjing Estuary.Sixteen (16) Landsat TM remote sensing images from1995to2013are collected and eleven (11) tidal gauge stations are refeered to the interpretation of tidal flow field at the Landsat TM transit time. The results of interpretation by the information retrieval of flow field based on the texture features of TM remote sensing images are verified by the numerical model Mike21_FM which was built by our research group.The results can be drawn as follows:1. Gabor filter is utilized into the enhancement of texture features in the Landsat TM remote sensing images. The more information of linear texture features in the Gabor filtered images, the closer to the main direction of tidal current in the channels of Changjiang estuary. Hereinafter, the direction of linear texture features in the TM Remote Sensing images can be interpreted as the flow direction. Therefore, the Gabor filter appears an advantage to enhance information of flow direction and improve the flow direction interpetation accuracy.2. The entropy of texture features in the TM remote sensing image is calculated based on the retrieval of Gray co-occurrence matrix. The results show that the larger entropy, the more information of the texture features, which is consistent to results of retrieval data by Garbor filter. So, a possible positive correlation might occur between the entropy of texture features in the images and flow velocity and the entropy is tentatively to be utilized to the classification of flow velocity in the estuary.3. The above two methods of information retrieval of flow field based on the texture features in the TM remite sensing images are effective in the flow filed with moderate suspended sediment concentration (SSC). This implies that the suspended sediment acts as the tracer of flow for the information retrieval of texture features in the TM remote sensing images.4. The root-mean-square error (RMSE) of direction of ebb and flood tidal flow by Gabor filter and the numerical model Mike21_FM is6.2°and18.5°, respectively. The fitness of calssified velocity of ebb and flood tidal flow by entropy cluster of texture features and the numerical model Mike21_FM is84.6%and61.5%, respectively. These indicate a better retrieval of the ebb tidal flow field information other than that of the flood tidal flow.5. Gabor filter is applied to information retrieval of flow field based on the texture features in the TM images in the Changjiang estuary and the north bank of the Hangzhou Bay. Significant changes in the surface ebb tidal flow direction in the North and South Passage fore and after deep waterway regulation project. Obvious feature of convergent ebb flows and divergent flood flows around the Nanhui spit. And also, Comb-like texture features exist around bridge piers of Donghai Bridge connecting the mainland with the Yangshan Harbor in the norther part of Hangzhou Bay.Automatic extraction of flow field based on the texture features are expected in the future research and the entropy cluster of texture features for the classification of tidal flow velocity is also tended to be verified and calibarated in the future.