In order to adapt to the development of UAV inspection technology and achieve accurate positioning of abnormal targets of photovoltaic panels, this paper applies deep learning to abnormal target location, and proposes abnormal target tracking and localization algorithm for UAV PV inspection scenarios. The advantage of deep learning lies in the ability to learn the general expression of the target. In order to obtain an efficient deep network, this paper uses a large amount of data to train the deep network, so that the features extracted by the network can achieve accurate representation of the tracking target. By testing in the abnormal target location test database, it is proved that the proposed algorithm can accurately locate the abnormal target.

No abstract

The emergence and rapid development of the Unmanned Aerial Vehicle (UAV) Photovoltaic inspection system have become an effective means of solving the operation and maintenance of photovoltaic power plants. In order to cooperate with the current UAV platform for photovoltaic panel anomaly detection, this paper proposes a photovoltaic infrared target anomaly detection system. In this paper, the Sobel operator is used to extract the photovoltaic slab area of the image, and the canny operator is used to obtain the photovoltaic small plate area to realize the complementary advantages. At the same time, the deep learning is applied to the algorithm to learn the discriminative characteristics of the image, and the brightness statistics are used. The method is supplemented by the method to achieve anomaly detection of the photovoltaic panel area. The effectiveness of the proposed algorithm is proved by a large number of experiments.

KEY WORDS: DLG and Map, quality control, the consistency of maps and DLG ABSTRACT:The product of "DLGM" is a vector dataset of DLG and map, and is software independent, which is produced by our DLGM integrated technology system. This product has been included in the fundamental Geographic Information system (GIS), and confirmed by the mapping industry with its advantages of the integration, commonality, and multi-purpose. As a new kind product of digital map, it involves a large number of theoretical issues and technical problems. Specially, quality of product is a crucial problem in one of them. In our study, the basic framework of DLGM product is given firstly. And then we talk about quality control method and mechanisms in the production process, which involved dynamic modelling based on the topological map model, data dictionary and "DLG and Map" integration template. The core task of this process is the consistency of maps and its corresponding geographic information. The rule of the process ensures the data quality in aspects of products design, manufacturing process method, tools, quality testing, application and data exchange. In recent years of practical application, a number of provincial fundamental DLGM products were produced by our DLGM integrated technology system, which has passed the existing quality standard authentication. It proved factually that the product quality control theory and technology which is adopted in our system is effective and feasible. At present, the technology and quality control method has been playing an important role in DLGM products.

Studies the regularity for changes of spatial relations based on the gradual changes of spatial location between two regional objects. Generates the conceptual neighborhood model of spatial relations like concentric circles, and provides the formal representation for them using graph theory. Based on the idea of "topylogy matters, metric refines" [1] , the model uses topological relations as basic classification and different direction and distance as the refinement of classification for spatial relations to reflect the levels of spatial relations. In addition, the changes of spatial relations between the conceptual neighborhoods are divided into six types. These changes are graded and evaluated according to the constraints degree of a variety of relations for spatial objects that in favor of formalization and computation. Finally, experiment is provided to illustrate that the conceptual neighborhood model and its formal description can describe the gradual changes of spatial relations between regional objects and compute the difference degree between spatial relations.