Research On The Extraction Of Water Information From High - Resolution Remote Sensing Image Based On Object - Oriented

Beautiful Colorful Yunnan in recent years, abnormal weather, drought, coupledwith the protection of water resources in Yunnan Province in the study focused on theDianchi Lake Basin, the lack of research on other small bodies of water, is notconducive to research and planning of water resources protection, resulting indroughts water shortage period. Water for industrial and agricultural production,tourism and the people living in Yunnan bring a lot of trouble. So the study areawater remote sensing image information extraction, the understanding of waterresources in the region will change, and then take the appropriate measuresconducive to the effective use of water resources, planning and protection. Therefore,remote sensing study area water body information extraction, and the mostappropriate way to find a region is necessary, but also a lot of great significance inthe water body information extraction methods have been formed.Feature information extracted from remote sensing images and diverse researchmethods have advantages and disadvantages, this article will traditional remotesensing information extraction of water object-oriented technology and remotesensing information extraction techniques were compared.First, the use of remote sensing image ENVI4.8software for extracting waterbody information based on traditional methods, because the experimental data forhigh-resolution QuickBird remote sensing images, multi-spectral bands onlynear-infrared, so the method can only be used with a single-band threshold,vegetation index, water index. Experimental study, the experimental resultsobtained.Secondly, object-oriented information extraction using water eCognition8.7software, eCognition8.7is an object-oriented technology based software. Focuses onthe object-oriented multi-scale image analysis segmentation algorithms and processes,in-depth analysis of the band weights, selected on the basis and principles of color,shape, firmness, smoothness, and other parameters of the scale division. Use of the method of the object and the object inside the standard deviation of the averagedifferential neighborhood quality of segmentation function selects optimalsegmentation based on scale. Analysis of the membership function and the knowledgerule base object feature selection policy and other related issues.Statistical analysis ofthe value of each band image through the data, select the image segmentation band,as well as the heavy weight of each band, and then divided according to the differentsegmentation scale, compared to find the best segmentation scale, according to thefuzzy rule base target feature information information extraction, and the accuracy ofthe results extracted from the membership perspective of the evaluation.Again, object-oriented and traditional water body information extractionaccuracy of information extraction methods comparison can be drawn from theobject-oriented technology in the areas of remote sensing image informationextraction accuracy of high-resolution than traditional extraction methods.Finally, experimental verification and found that in an object-orientedmulti-scale segmentation, different segmentation scale in a prominent part of theinformation, it would also lose another part of the information, which is inevitable.Massive data processing segmentation and feature parameters to calculate theresulting information also makes it difficult to extract. For problems inobject-oriented information extraction technology based on the proposedhigh-resolution remote sensing image library pond pixel extraction method andobject-oriented combination of tested, through efficiency and accuracy of evaluation,this method results not only to ensure the accuracy and improve the classificationefficiency.