面向对象的煤矸石堆场SPOT-5影像识别

煤矸石是一种在成煤过程中与煤层伴生的黑灰色固体废弃物,不仅会污染环境,而且会严重损害附近居民的身体健康,目前已经成为矿区生态环境的主要影响源之一。因此,实时、准确、快速地获取煤矸石堆场的位置、形状和面积等信息,对于环境监测与管理具有重要的意义。本文以内蒙古鄂尔多斯市东胜区为试验区,将试验区内的典型地物分为：植被、水体、阴影、裸地、建筑、道路、排土排矸场、露天煤矸石堆场、堆煤场及煤渣、采煤坑和其他共11类。本文采用SPOT-5高分辨率遥感影像,面向对象提取研究区内的煤矸石堆场信息,并进行识别精度评价,精度达到89.47%。将面向对象的分类结果与最大似然分类方法的分类结果进行比较,结果表明,面向对象的提取方法可更好地应用于煤矸石堆场信息的自动提取,大幅度提高精度和效率。

Coal Gangue Yards Information Extraction Using Object-oriented Method Based on SPOT-5 Remote Sensing Images

Coal gangue is a kind of dark gray solid waste generated during mining process. Nowadays, coal gangue has become one of the biggest pollution sources to the ecological environment in mining areas. The accumulation of coal gangue not only occupies excessive land and causes serious environmental problem, but also damages the health of local people. Therefore, it is urgent to reduce the coal gangue yards in mining areas. In addition, extracting the location, shape and size information of coal gangue yards quickly and accurately is significant to environmental departments. The traditional methods of investigating coal gangue yards cost a lot of time and money. While remote sensing technologies can record the information of earth surface quickly and accurately, they have obvious superiorities in extracting coal gangue yards information. This paper takes the Dongsheng District, which locates in Ordos City of Inner Mongolia, as the study area, and utilizes SPOT-5 high resolution image as the data source. Then, this paper adopts the object-oriented method to extract coal gangue yards information from the study area. Multi-resolution segmentation and fuzzy classification algorithm are the most important steps in this method. Four appropriate segmentation scales are determined through comparisons of several tests, they are: 400, 160, 80 and 40. Next, we classify the segmented objects into different classes using the fuzzy classification algorithm that based on objects’ characteristics, such as spectrum, shape, texture and other features. The objects are further classified into eleven classes: bare area, buildings, roads, water, vegetation, shadows, dumps, coal gangue yards, coal yards, coal pits and others. The rule set used to extract different classes is built, which is aimed to provide a reference to relevant environmental departments to quickly and conveniently monitor the environment in coal mining area. In the end, we assess the accuracy of the classification results: the total accuracy is 88.78% and the user accuracy of coal gangue yards information is about 89.47%. Besides, a comparative extraction result is extracted using maximum likelihood method, whose total accuracy is 64.13% and the user accuracy of coal gangue yards information is only about 64.18%, which is much lower than the result extracted by the object-oriented method. Generally, due to the serious pollution caused by coal gangue yards, and considering the object-oriented classification method is seldom used to extract coal gangue yards information in China and abroad, this paper tries to extract coal gangue yards information using object-oriented classification method, and establish a rule set that can be applied to extract coal gangue yards information and other typical features. As we can see from the analyses, this paper has its significance in environmental protection. Relevant conclusions and analysis can be provided to the environmental protection departments as a useful reference to monitor and manage the pollution caused by coal gangue yards.