| Fish pond culture is a basis for aquaculture. Taishan City, Guangdong, which is a county-level city in the economically developed Pearl River Delta region, has many fish ponds, and well developed aquaculture industry, which is a vital composition of its agriculture industry. At the moment, an urgent demand for modern management of it is spatial and temporal distribution of fishpond area. Traditionally, the area of fish pond is mainly through statistical reporting channels. However, due to the fact that the lack of actual measurement and many other reasons, the statistics of local fish pond area information is always inaccurate. It proposes a serious obstacle to the aquaculture macro-control of the government. Because of the advantage of area measurement, the application of remote sensing on monitored fish pond efficiently as well as information extraction from the fish pond reveals practical significance, and it could be an important direction for fishery remote sensing. This paper tries to estimate the fish ponds area quickly and accurately through remote sensing method.The research of this report set Taishan City as study area, employs The Environment for Visualizing Image (ENVI) ZOOM software platform, Landsat-5 TM data source and DEM data source, adopts a mixed methodology of object-oriented and creating knowledge rules to estimate the area of fish ponds in study area. First of all, using object-oriented multi-scale image segmentation theory to divide the image by multi-scale multi-spectrum segmentation techniques and constructing the objective land use classes; And then, creating various knowledge rules which involves dividing the use classes by spectral information for land and water, weeding out the non-water surface features, excluding linear water such as rivers and ditches by spatial information, excluding planar water such as lakes and shallows by spectrum information for setting the threshold in the infrared and near infrared bands, excluding the remaining lakes and shallows by texture and elevation information, excluding small bodies of water through area index. To this, the distribution of chosen fishpond can be acquired; Finally, analyzing the distribution of the ponds and estimating its square meters,Meanwhile, in order to verify the feasibility of this method, use random sampling to test the accuracy of the extraction. In the study area, randomly selected 300 test points, record their feature type in the corresponding by using visual identification methods. Compared with the result of ISODATA unsupervised and SVM supervised classification, it indicated that this object-oriented method has the the highest accuracy in extraction. It effectively decreased the "salt and pepper" effect incurred in the traditional classification based on pixel. The extraction result of this method is of high accuracy 81.44%, which shows an improvement to accuracy to the unsupervised classification and supervised classification methods of 40.22% and 38.62%, respectively. Consequently, the knowledge rules based on multi-scale object-oriented segmentation method is effective way to improve the accuracy of ponds information extraction.According to this study method, the fish ponds area in Taishan City was 98529300 square meters in 2006. The methodology is examined covering a range of landscapes in Taishan, and use of visual interpretation to analogy the effects of ponds extraction. The results show that it performs better in Wide coastal plain of the DaTong River fishpond with a high accuracy which most closely with the visual interpretation results. And it is obviously higher than the accuracy obtained in coastal zone and hills. Overall, the area estimation of fish ponds in TaiShan by this method can be said have fairly reliable and be more suitable for usages in coastal plain watershed ponds, which arranged ordered.The main innovation points of this report includes as follows:Currently, using remote sensing methods for extraction of fish ponds is little, immature, therefore, this paper find a remote sensing method which comprehensively considers the spectral, shape, texture and area characteristics. It creates appropriate rules to achieve rapid and accurate area statistics of fish ponds. And it identifies the fish ponds distribution then analyzes the spatial distribution of fish ponds in Taishan City.
|
PDF Full Text Request