Research On Photovoltaic Panel Extraction Method Based On Region-Line Primitive Association Analysis And Template Matching

In object-based image analysis(OBIA)technology,the accuracy of object classification is mainly determined by image segmentation,sample or rule set and classifier.The quality of segmentation has a great influence on the subsequent feature analysis and extraction.On the other hand,template matching technology identifies specific objects from images,and can prevent shape defects caused by image segmentation effectively.However,the template creation and editing are time-consuming and difficult.In this study,the Region-line primitive association framework(RLPAF)is used as the technical support,and the technology of OBIA and template matching is fused and applied to the automatic extraction of the photovoltaic panel on the high resolution remote sensing images.The main works of this paper are described as following:(1)The template is automatically generated based on RLPAF.Firstly,a hard-boundary-constrained segmentation method and the phase marshaling method are used to obtain the region-line primitives.After that,we extract multi class features of primitives respectively.Then,through the analysis of the photovoltaic panel spectral and shape features,we set the rules based on the OBIA technology to extract the suspected photovoltaic panel region primitives.Finally,the standard templates set of photovoltaic panels that satisfying the best-fitting template requirements are established based on RLPAF.(2)Template matching based on gray level.In order to improve the efficiency of matching algorithm,the mask is set up based on the spectral characteristics of the segmentation region primitives and the search area has been reduced.At the same time,a rough-precise searching strategy is adopted to reduce computation of each pixel-point.The rough area of the photovoltaic panel is quickly positioned.Finally,the structural similarity is used as a similarity measure index to achieve the precise template matching and extraction of photovoltaic panel.(3)The application of the proposed on high resolution remote sensing is carried out.The size of photovoltaic panel is small and densely distributed.It is hard to get the exact location of a single target.In this paper,unmanned aerial vehicle images with different regions and environmental backgrounds are selected as experimental data sources for experimental verification.Finally,comprehensive assessments of the performance and comparative analysis of different methods are completed.Experiments show that compared with conventional OBIA technology and template matching technology,the proposed algorithm can get a higher precision in the recognition of photovoltaic panels based on the high resolution remote sensing image.At the same time,the users do not need to select template or train sample manually.Thus,the proposed algorithm has good universality and robustness,which embodies the technical features and advantages of the algorithm.