| China has carried out three national land census,a census of geographical conditions,and sustaining monitoring,formed a large scale,multi-tempo,nationwide database of natural resources,using the two technical paths of "micro-numerical analysis and macro-multi-scale expression",scientifically supporting the construction of ecological civilization in China.The geographic conditions thematic is a macro-multi-scale expression of the mixing of the spatial cover of natural resources with the content of important topographic elements,it is an indispensable and basic space support for China's land and space planning and natural resource development,utilization,protection and restoration.The core of realizing micro large-scale data to macro multi-scale expression is the automatic generalization of maps.Domestic and foreign experts and scholars,research institutions,for decades unremittingly committed to this research,has made great progress in the topographic map single or multi-element generalization,quality evaluation,as well as human-machine interactive topographic mapping drawdown system and other research.Compared with the above-mentioned topographic map,geographic conditions thematic map's type is more diverse,involving the semantic rich,thematic space cover with very different spatial structures;the data mass is even larger,mostly to county,city,province and even the national unit of coverage;the scale span is more leaping,from 1:1million data to 1:1 million scale expression.As a result,it is more difficult,complex and challenging.The geographic conditions thematic patches are an important part of the geographic conditions thematic map,and its generalization quality directly affects the generalization effect of the geographic conditions thematic map.Existing research on thematic map synthesis have been presented various generalization operator such as dissolving,amalgamation,agglomeration,and simplifying,however,in the actual geographic conditions thematic patch data,the individual pattern of the patches is diverse,the group space pattern is complex and changeable,how to select the corresponding generalization operator according to the pattern and pattern of different local patterns has been widely concerned by experts and scholars.In addition,compared with the generalization topographic map(?)due to geographical differences,the geo-thematic maps are complex and diverse,its automatic generalization should consider not only the constraints on the spatial scales of geometrics,topology,direction and measurement involved in the topographic map,but also the constraints of.its unique spatial distribution structure,semantic and statistical balance of the area of the earth,etc.,the existing methods are still incomplete and need to be optimized.To this end,in view of the limitations of the spatial pattern understanding and automatic generalization of the geographic conditions thematic patches,first,the machine learning algorithm of spatial pattern understanding is studied,the typical patches smudges are identified automatically,and then the automated generalization processing methods to maintain the pattern of different spatial patterns are studied separately,four major innovations have been achieved:(1)A method of automatic patches recognition algorithm based on neural network decision tree.The identification of geo-features in traditional comprehensive algorithms relies on human intervention,this paper makes use of the efficiency of the decision tree and the adaptive characteristics of neural networks,studies the understanding method of the spatial pattern of geographic conditions thematic patches based on neural networks,and reduces the artificial intervention of the identification of typical patches(regular patches,aggregate patches,long-narrow patches and small patches),and to help improve the efficiency,accuracy and automation level of the classification of land objects,lay the foundation for subsequent generalization of different types of patches.(2)A method for merging patches to maintain contour characteristics of structured geographic objects.In summary,the present merging methods adequately maintain the natural appearance of areas and effectively keep the sum of area variation of each land-cover class as low as possible.However,these methods often proceed on a global basis and consider each land-cover class as a unit to account for the characteristics of "no gap" and "no overlap" in such land-cover data,less attention is given to the characteristics of earth objects with inherent regularity in spatial distribution,such as buildings,pits,etc.Therefore,during merging,the boundaries of the area features with unique spatial structures are altered,and the spatial structure characteristic is partially or fully lost.On the basis of the previous studies,on the premise of maintaining the law of spatial distribution,proposing a merging method to maintain the contour characteristics of structured objects,on the basis of structured object recognition such as linear pattern,grid pattern,adjacent area,through mosaic and boundary contour reduction technology,the structured features of the patches are maintained on the basis of maintaining the balance of the area of the ground class.(3)A method of long-narrow land patches dissolving algorithm that considering structural features.The extraction of partition lines for long and narrow patches(LN patches)is an important,yet difficult problem in the generalization of thematic data.When current methods are used to process polygons with irregular shapes or complex branch convergence zones,the extracted line structural features tend to be inaccurate and topologically erroneous,in this article we proposed an improved partition lines extraction algorithm of constrained Delaunay triangulation for these issues.On the basis of using the third type of triangle to identify the branch aggregation areas of A,B and C,the shape jitter and topological inconsistency on the skeleton line are adjusted by considering the directional consistency and distance elements.The split line results extracted by the method in this paper take into account the geometric features of natural smoothness,the structural features conforming to visual cognition and the topological features.(4)A method of small-area patches amalgamation algorithm considering both local optima and overall area balance.Dissolving small patches is a common operation in patches generalization,the key to this operation is to uphold semantic constraints and local spatial patterns proximity of patches,at the same time,to maintain the overall area balance between different land uses after dissolving.In view of the existing research can not be very good at the same time to solve the above two problems,this paper proposes a dissolving method that considering for both local optima and overall area balance.Through the pre-allocation of small patches and the iterative adjustment algorithm of overall area balance,the amalgamation results take into account the local spatial geometry,semantics and overall area balance of the area.Finally,the WJZ-? geographic conditions thematic map intelligent generalization system was developed,inline this paper put forward the spatial pattern understanding and keeping method,solve technical problems such as operator,algorithm based on logical reasoning adaptive selection and parameter adaptive adjustment,through the establishment of knowledge and algorithms library,for the g geographic conditions thematic map automatic generalization provides a scientific tool.The system has been fully applied to the multi-scale expression of geographic conditions thematic data,supporting more than 20 provinces nationwide to carry out the province-wide multi-scale geographical situation map,atlas production,generalization automation has increased dramatically,generalization efficiency increased by at least 6 times,the results are all through quality inspection. |
PDF Full Text Request