Object-oriented Remote Sensing Images Industrial Solid Waste Information Extraction Method Research

With the rapid development of national economy, the volume of industrial solidwastes is increasing, which becomes one of the most severe issues of environmentprotection in our country. In order to detect and monitor these industrial solid wastes,the traditional field methods no longer meet the needs of environment protectingnowadays. Since remote sensing owns the advantages such as wider extension, lowercosts, space continuity and real-time monitoring, which provides an efficient way forindustrial waste detection.To detect and monitor industrial waste dump, in this paper, we employ anobject-oriented method for detecting and extracting the industrial solid wastes byusing eCognition, a professional remote sensing image classification software. Weselect phosphogypsum and manganese slag which are typical industrial solid wastesas our target, the study area of phosphogypsum is located in Fuquan in Guizhouprovince of China and the study area of manganese slag is located in Songtao inGuizhou province of China. By using the data from HJ-1CCD,SPOT-5and ASTERimage, we analyzed, detected and extracted the information from them respectively.The object-oriented method we adopted consists of the following steps:(1) Select appropriate statistic sources, and then deal with the images ofremote sense (including atmospheric correction, image mosaic andimage cutting).(2) Segment the remote sensing images for obtaining the object-basedimages by using multiresolution segmentation method.(3) Build the feature knowledge set of the object types, provide thereference sample and prior knowledge to detect the object types.(4) By using the feature knowledge set, we analyzed the heterogeneities ofdifferent objects Based on these heterogeneities, we construct theobject-oriented decision tree rule set, then the industrial solid wastes canbe easily detected from the images.(5) Extract the information of manganese slag tailing pond and phosphogypsum tailing pond respectively, and sensitive objects aroundfrom the images by using ArcGIS. Then evaluate the potential risks ofthe tailing and its possible impact to the environment initially.Through the way of object-oriented decision tree, phosphogypsum can beeffectively distinguished from HJ-1images. By using principal component analysis,we detected manganese slag from ASTER images, but some objects of river andvegetation with similar spectral characteristics mixed in. After that, we build the ruleset from SPOT-5image, which has a higher resolution, though the accuracy of theresults is not as good as the results of ASTER image, it’s good removal of vegetationand rivers by using it’s texture features. Combine the results with the results of theASTER image identification, it can improve accuracy of manganese slag recognition.At last, we verified by ground survey data.Experiments and results indicate that the object-oriented method provides aneffective method for detecting industrial solid wastes.