基于低空遥感影像的建设用地变化检测研究

发布时间：2018-01-03 02:25

本文关键词：基于低空遥感影像的建设用地变化检测研究　出处：《南京大学》2013年硕士论文　论文类型：学位论文

【摘要】：本文根据国土资源部公益性行业科研专项课题“土地动态监测监察技术研究”(201011015-1)的需求,针对江苏省金坛市典型地区,研究基于低空遥感影像的建设用地变化检测方法,快速发现建设用地变化信息,特别是新增建设用地信息,为提高国土资源动态监测效率提供技术支撑。建设用地是工业化、城市化的核心要素之一,及时掌握建设用地变化情况对提高土地资源利用率、协调耕地资源保护与经济发展之间的矛盾等具有重要意义。无人机、无人飞艇等低空遥感平台具有机动、实时、低成本的特点,与航空、航天遥感平台相比,它能有效避免云雾、成像周期等因素的影响。在区域尺度上利用低空遥感影像快速检测建设用地变化信息,是一种新型的国土资源动态监测技术手段。通过本文研究,提出了一种集成像元变化分析和对象识别的建设用地变化检测方法。首先根据低空遥感影像的成像特点,综合利用低空遥感影像中的光谱与纹理信息,在像元级设计自适应综合差值方法检测土地利用变化信息；然后通过数学形态学等方式优化变化检测结果,在提高变化信息精度的同时,提升变化信息的完整度,实现从像元级检测到对象级识别的过渡；最后,以对象纹理分析为手段,在对象级进一步验证变化信息,并采用对象识别的方式从所有的变化对象中筛选出建设用地变化对象。本文的主要研究内容与结论如下： (1)结合光谱与纹理信息的土地利用变化检测。低空遥感影像空间分辨率高,地物的结构和纹理信息丰富；而波段较少,光谱分辨率相对不足。基于以上特点,在变化检测过程中充分挖掘影像的纹理信息。在光谱特征的基础上,选择适合的纹理特征,经逐波段差值计算后,设计自适应的阈值分割方法分别获得各光谱波段和纹理特征所表征的变化信息。以逻辑合并的方式将不同来源的变化信息综合,生成土地利用变化像元。结果表明纹理信息能够反映变化区域内部的细节信息,在提升检测精度的同时,能有效降低检测结果的破碎度。所提出的自适应综合差值方法相比于其他常用的基于直接比较的变化检测方法,其检测结果具有更高的正确率,且有效降低了漏检率。 (2)土地利用变化检测结果优化。由于低空遥感影像空间分辨率高,地物内部差异较大,加上配准误差、细部结构和噪声的影响,像元级变化检测结果较破碎,一方面生成了许多无意义的伪变化区域,另一方面变化信息表达不完整,变化区域与地物对象之间不具有明显的对应关系。依次采用形态学闭运算、孔洞填充、形态学开运算、小区域去除等四个步骤优化变化检测结果,以去除伪变化像元,提高变化检测的精度,并生成形状完整、边界简洁的变化区域,便于直接获得变化对象,实现从像元级检测到对象级识别的过渡。 (3)建设用地变化对象识别。由于受局部细微结构差异的影响,在像元级变化检测结果中仍存在部分伪变化信息,需要在对象级对变化检测结果做进一步筛选。通过区域标记将连通的变化像元追踪为单个变化对象,利用变化对象在前后时相的纹理差异滤除伪变化对象。然而由于土地利用变化检测结果中包含所有类型的变化信息,变化对象中不仅有建设用地变化对象,还存在非建设用地变化对象。采用对象识别的方式进一步筛选出建设用地变化对象。分别针对均质和非均质的建设用地对象建立相应的纹理识别标志,在单个时相中识别变化对象是否为建设用地,并输出最终的建设用地变化对象。综上所述,利用低空遥感影像进行建设用地变化检测时,纹理与光谱信息的综合应用以及变化检测结果的优化,能够有效降低漏检率；变化对象的筛选与建设用地对象的识别能够降低误检率,提高建设用地变化检测的精度。研究结果表明,该方法能够快速获得建设用地变化信息的几何范围与物理属性,有助于提高建设用地动态监测的效率。
[Abstract]:According to the Ministry of land and resources public welfare industry research special subject of "land dynamic monitoring monitoring technology research" (201011015-1) requirements, according to the typical area of Jintan city in Jiangsu Province, the construction land use change detection method of low altitude remote sensing images based on fast find the changes of construction land, especially the new construction land information, in order to improve dynamic monitoring of land resources efficiency and provide technical support.
Construction land is one of the core elements of industrialization, the city, to grasp the situation to improve the utilization of land resources construction land use change, has the contradiction between cultivated land protection and economic development coordination is the important significance. The UAV, unmanned airship remote sensing platform with mobile, real-time, low cost airlines and characteristics. Compared, space remote sensing platform, it can effectively avoid the influence of cloud imaging cycle and other factors. In the regional scale using the rapid detection of low altitude remote sensing information construction land change, is a new means of state land resources dynamic monitoring technology.
This paper put forward a detection method of construction land use change an integrated pixel change analysis and object recognition. According to the imaging characteristics of low altitude remote sensing images, the comprehensive utilization of spectrum and texture information of low altitude remote sensing image, value detection method of land use change information in the design of adaptive pixel level comprehensive optimization; and then change detection results the mathematical morphology method, improve the accuracy of information on changes at the same time, enhance the change information integrity, to achieve the transition from pixel level detection to the object level recognition; finally, the object texture analysis as a means to further verify the information in the object level, and changes from all objects of the selected object construction land use change the object recognition way. The main research contents and conclusions are as follows:
(1) combined with the detection of land use change on spectral and texture information. The low altitude remote sensing images of high spatial resolution, feature rich texture and structure information; while the band is less, the relative lack of spectral resolution. Based on the above characteristics, changes in the detection process to fully tap the image texture information. Based on the spectral feature, texture feature, the spectral difference calculation, an adaptive threshold segmentation method is used to obtain the change information of each spectral band and texture feature representation. Combined with logical methods will change the letter of different sources of comprehensive information, land use change to generate pixel. The results show that the texture information can reflect the change of details within the region. While improving the detection accuracy, can effectively reduce the detection results of fragmentation. The integrated adaptive difference method proposed compared to other commonly used based on the straight Compared with the change detection method, the detection result has a higher correct rate, and the leakage rate is effectively reduced.
(2) land use change detection result optimization. Because of low altitude remote sensing images of high spatial resolution, internal differences of larger features, plus the registration error, influence of detail and noise, the pixel level change detection result is broken, on the one hand, generating a lot of meaningless pseudo change regions, on the other hand, changes in expression of the incomplete information, with no obvious corresponding relationship between the object and the ground. The change area followed by morphological operations, hole filling, morphological open operation, small area to remove the four steps optimization of change detection results, to remove the pseudo pixel changes, improve the accuracy of change detection, and generate a complete shape, simple changes in regional boundaries, to facilitate direct access changes in the object, to achieve the transition from pixel level detection to the object level recognition.
(3) the object recognition of construction land use change. Due to the local subtle structural differences, there are still some false change information at the pixel level change detection results, the need for further screening at the object level on the change detection results. Through region labeling will change pixel connectivity for a single object tracking changes, changes in the object before and after use when the texture difference phase filtering false change object. However due to include all types of information change detection results of land use change, change the object not only the object of construction land use change, the existence of non construction land use change object. Object recognition further screened object construction land use change for homogeneous and respectively. The heterogeneity of construction land to establish the corresponding object texture recognition marks recognition change whether the object is construction land in a single phase, and the output end of the construction land Changing objects.
To sum up, to detect the changes of construction land use of low altitude remote sensing image, optimize the comprehensive application of texture and spectral information and change detection results, can effectively reduce the missing rate; screening and construction object identified by the object changes can reduce the error rate and improve the construction land change detection accuracy. The results show that and this method can quickly obtain geometric and physical attributes of the construction land use change, help to improve the efficiency of dynamic monitoring of construction land.

【学位授予单位】：南京大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TU984.113;P237

【参考文献】