| With the advent of the era of big data,people have urgent needs and expectations for the internal mechanism and high-value information contained in big data.As an important tool and means of big data analysis,visualization technology has become the focus of attention and research.In data visualization,the data of various industries often use geographic information technology to integrate them into a unified geographic space for analysis and expression.Therefore,geospatial data visualization is an important basis of big data analysis and visualization technology.By transforming it into common symbols such as maps,the information will be transmitted to users.In this process,label is the key for users to obtain valuable information correctly and quickly.Therefore,obtaining more readable and clear labels has become the focus and difficulty of research.In the large-scale or dense label placement of point features,the large scale of its problem and the fierce spatial competition among labels are will lead to inefficiently labeling and unreasonable placement results.Therefore,by fully mining the spatial distribution characteristics and label correlation of datasets,this paper proposes a data mining-based algorithm for automatic placement of point feature labels.Firstly,the algorithm designs the frequent patterns of point feature label by mining the special distribution characteristics of the point dataset and the correlation among labels in advance;Secondly,the spatial clustering algorithm based on frequent pattern is used to analyze the overall spatial distribution characteristics,which divides the large-scale overall dataset into several independent sub-datasets to eliminate the interference and ambiguity generated by independent datasets in the overall solution;Thirdly,the local spatial distribution characteristics of the point features and the correlation among labels are mined to design the label order based on the ascending frequent patterns,and multilevel metaheuristic algorithms are used to solve the label quality and efficiency.Finally,a bit-based spatial index of the grid is proposed to reduce the cache space and query time of index in the conflict detection.The research results of this paper include:(1)This paper systematically composes the basic principles and methods of label and automatic label placement,elaborates the principles and methods of automatic label placement such as label candidate position model,conflict detection,evaluation function,and metaheuristic algorithm,and focuses on the application of simulated annealing,ant colony algorithm and genetic algorithm in point feature placement.On this basis,the research on the point feature label placement algorithm considering the spatial distribution and label correlation is carried out.(2)Aiming at the label placement problem of dense point features,a label frequent pattern is designed by fully mining the spatial distribution characteristics and label correlation of the point dataset,quantifying the interference degree among point feature label placement.Based on the definition of the frequent pattern,an ascend order frequent pattern rule is designed to guide the label placement process,which effectively improves the quality of point label placement.The experimental results show that the evaluation function value is reduced by 2.2～18.5 under 5%～40% label density compared with other label methods.(3)For the large-scale point feature label placement problem,a spatial clustering algorithm based on frequent patterns of labels is used to analyze its whole spatial distribution characteristics,and the large-scale overall dataset is divided into several independent sub-datasets,thus reducing the complexity of problem solving;the traditional metaheuristic algorithm does not have universality for dataset of different sizes,a multi-hierarchy metaheuristic method is proposed to solve the problem,which improves the robustness and generalizability of the metaheuristic algorithm.The experimental results show that the solution efficiency is improved by 5.04%～54.80%on average under 5%～40% label density,and the label quality evaluation function is reduced by 0.1～43.8 on average compared with the original algorithm.(4)Aiming at the shortcomings of the traditional grid space filling and R-tree spatial index conflict detection algorithm,a bit-based grid space filling conflict detection algorithm is proposed,which effectively reduces the memory resource occupation and the number of addressing queries,and improves the operation efficiency of the conflict detection algorithm.Experimental results show that the average detection efficiency of the new algorithm is improved by 51.59% and 30.26%respectively compared with the traditional grid space filling and R-tree spatial index. |
