Applied Research In Remote Sensing Image Classification Of Industrial Solid Waste Used By Object-oriented Approach

Industrial solid waste pollution problem has become one of the tenmajor environmental problems internationally recognized. With the accelerated process ofindustrialization, emissions and store up the quantity of industrial solid waste has alsoincreased by a large margin. Although the development of environmentalprotection technology have made the industrial solid waste can be recycled, but the waypeople deal with solid waste treatment is more concentrated stacking，which not onlyinterfere with the landscape, but also a lot of harmful substances contained in industrialsolid waste will go through precipitation and infiltration into the soil, causing groundwaterpollution, killing the microorganisms in the soil, destroying the ecological balance andforming of industrial solid secondary pollution waste. Therefore, the use of scientific andeffective methods to monitor and manage them, reducing pollution and effects ofindustrial solid waste on the environment, maintaining the benign loop of ecology and theenvironment is very necessary. However, the traditional means of monitoring the groundcan not meet the needs of today’s environmental protection, and remote sensing withlarge-scale, low-cost, continuous and real-time monitoring and other advantages formonitoring of industrial solid waste provides an effective means.Solid industrial classification of remote sensing image as an effective wayto industrial solid monitoring by remote sensing, and its classification precision directlyaffects the accuracy of monitoring results. However, traditional remote sensing imageclassification are mostly based on pixel-level classification is not considering a number offeature information of the image, which led to the classification accuracy is not high. Forthis, this paper presents the remote sensing image classification method which utilizesthe shape, spectrum and texture information of the image and mixes graph theory withsupport vector machine (SVM).Industrial solid waste remote sensing image classification method in this paperincludes the following steps: Firstly, we do quad-tree segmentation of remote sensing images based on Remotesensing image preprocessing; then, we need obtain the spectral, shape andtexture weight component by calculating spectral similarity, the matching degree betweenpixels and the texture similarity and use R-cut cut sets guidelines of graph theory forfurther multiple features image segmentation; at last, we do the SVM classification onthe segmentation result, get the final classification result.In this paper, we take into account the spectral, shape and texture features of remotesensing image, use the remote sensing image classification method which put the graphtheory and SVM combined, used overall accuracy and kappa coefficient of accuracyassessment, and compared with the traditional classification method of Mahalanobisdistance, maximum likelihood, spectral angle mapping, neural networks and support vectormachine. Experimental results show that the proposed method effectively improves theclassification accuracy. Therefore, this method can be effectively used for remote sensingimage classification of industrial solid waste.