Research On Perception Of Traditional Cityscape Characteristics Based On Multi-source Urban Big Data

As the political,cultural and economic center of China,Beijing is a world-famous historical and cultural city.It was the capital of six dynasties in history with many historical sites and traditional buildings.However,with the rapid development of Beijing’s modern city construction,the issue of urban style convergence has gradually become prominent.Traditional buildings have been replaced by modern buildings,and the traditional style of Beijing has been gradually fade out in the process of urban development.In response to the crisis of urban style convergence,the fourth chapter of the 2017 Beijing Urban Master Plan(2016-2035)puts forward the planning objective of"strengthening the protection of historic and cultural cities,strengthening the urban characteristics of capital style,ancient capital style and times style".In order to achieve this planning goal,the primary problem is to realize the perception of urban feature on a large scale,so as to master the spatial distribution of urban traditional landscape.In the era of big data,the urban computing research based on urban big data has gradually replaced traditional methods and become an emerging research mode of urban perception.Therefore,the use of multi-source urban big data to explore the sensing mode of urban traditional landscape is one of the hotspot topics need to be studied.This study quantitatively explores the multi-level perception and spatial distribution of the traditional style of Beijing city based on the methodology of deep learning and spatial analysis theory,using three datasets including points of interest(POI),street view images and weibo check-in within Fifth Ring road of Beijing.Firstly,spatial distribution pattern and aggregation characteristics of POI data of scenic spots in traditional style have been analyzed based on spatial analysis theory,so as to realize traditional style perception based on POI entity.Sencondly,based on street view panoramic image,a deep learning model for intelligent recognition of traditional style buildings and visual indicators of traditional style buildings are proposed,which can perceive the location of street attractions and the spatial distribution characteristics of traditional style buildings at street level from the perspective of pedestrian vision.Then,weibo check-in with text information,spatial thermal map and natural language sentiment tendency analysis model are used to quantify the public’s perception of Beij ing traditional style.Finally,aiming at the comprehensive analysis of the results of multi-source data,a research framework of multi-spatial collaborative perception of urban traditional landscape based on feature threshold is proposed,which solves the spatial distribution analysis problem of multi-source urban large data collaborative perception of urban traditional landscape.The main research contents and conclusions are as follows:(1)From the perspective of POI entity,this paper explores the basic processes and key technologies for perceiving spatial aggregation and spatial distribution patterns of Beijing traditional style based on POI data which represent traditional landscapes.First of all,POI scenic spots are divided into official style and folk style according to the traditional architectural style system.Then,the proximity degree and spatial pattern characteristics between POI are described quantitatively by using the nearest distance index method.Kemal density estimation was used to explore the spatial morphology of traditional style POI cluster within Fifth Ring road of Beijing.The results show that POI of traditional scenic spots has significant spatial aggregation characteristics,and the spatial distribution characteristics of large-scale aggregation of the second ring and small aggregation distribution outside the second ring.The specific conclusions are as follows.①POI of traditional style scenic spots and its subclasses(official style and folk style)both show significant spatial distribution characteristics.Among them,the agglomeration characteristics and the overall distribution characteristics of the official style scenic spots POI are close to the traditional style POI,which indicates that the official style scenic spots POI occupies a dominant position in the traditional style scenic spots POI.②traditional style scenic spots POI within Fifth Ring road of Beijing generally take the spatial distribution characteristics of large-scale aggregation centered on the Palace Museum in the Second Ring Road,small aggregation centered on large aggregation in the Second Ring Road and small aggregation scattered outside the Second Ring Road.The spatial distribution characteristics of POI of official style and folk style are quite different.From the overall spatial form and spatial expansion trend,the distribution pattern of official style POI is similar to that of traditional style scenic POI.The POI of folk-style scenic spots is quite different from the general characteristics.(2)For the lack of labeled data and directly available deep learning model for intelligent recognition of traditional style buildings in large scale,this paper focuses on:①Classification system of Chinese traditional architectural forms,and puts forward that traditional style buildings can be divided into two categories:official buildings and folk buildings.Relying on the images of Tencent Street View and Baidu Street View,the data set of traditional architecture and the data set of traditional architecture labeling are constructed.②Making full use of the general shallow features extracted from large data COCO datasets by deep convolution neural network and using transfer learning technology,the object detection and recognition model Mask R-CNN is extended to the recognition of traditional style buildings in urban street scenery.Based on this,a traditional style building intelligent identification model(TBMask R-CNN)is put forward.This model,of which MAP value on the validation set reached to 80%using annotation data set of the traditional style architecture built,can be used for intelligent identification work of traditional style architecture.(3)From the perspective of pedestrian vision,TBMask R-CNN model using street view panoramic data set to perceive the location of street attractions and the traditional urban landscape at street level.The TBVI and STBVI are proposed to reflect the degree of pedestrians’ perception to the surrounding traditional architectural environment.The results show that the spatial distribution of traditional style perception based on street view image presents the characteristics of aggregation in the second ring and dispersion outside the second ring,and the distribution presents a positive spatial correlation.The specific conclusions are as follows.①The spatial distribution of the traditional landscape perception based on street view images is characterized by the aggregation,and perception in north area is larger than those in south area in Second Ring.Traditional landscape perception in Third Ring,Fourth Ring and Fifth Ring are dispersed.②Moran’s I value of TBVI in the location of street scenic spots indicates that both official style and folk style buildings show positive spatial correlation,with a larger value for folk style.③Results of STBVI show a consistent pattern of decreasing traditional buildings from the center to the periphery from Second Ring to Fifth Ring,as well as the official style and folk style buildings in the subcategory.Among them,value of Second Ring is the highest,followed by those of Fifth Ring,Third Ring and Fourth Ring.(4)Sentiment perception and spatial feature analysis of traditional urban features based on weibo check-in data.Using weibo check-in data,thermal maps are used to identify the areas of concern and the heat of concern in traditional style areas.Relying on the text sentimental tendency evaluation model based on deep learning,semantic analysis and evaluation of sentimental tendency are conducted on the Weibo texts,so as to realize the sentimental perception in the traditional style area.The research found that the hotspots of check-in were concentrated in the Second Ring Road and mainly positive emotions.① Check-in thermograms show that the geospatial distribution and agglomeration scale of POI check-in activities of traditional scenic spots in Beijing’s Fifth Ring Road are mainly concentrated in Tian’anmen and the Forbidden City in the Second Ring Road,areas with Taoran Pavilion and Temple of Heaven as small centers,national parks in the Fourth Ring Road and Summer Palace and Yuanmingyuan in the Fifth Ring Road.The second ring is the gathering area of official style and folk style buildings.The spatial distribution of POI signing hotspots of official style is similar to that of the whole.The signing hotspots of folk style are mainly in the areas of Jingli Hutong Siheyuan,Maoer Hutong,Black Sesame Hutong,Dongyuan Theatre Building,Glass Factory Antique Street and so on.②The results of emotional perception research show that the public’s emotional perception of traditional style area represented by traditional style POI is mainly positive emotion,and the positive emotion of traditional style area of official style is higher than that of traditional style area of folk style.On the one hand,Beijing is an ancient capital of the Six Dynasties,which has preserved a large number of official style buildings;on the other hand,the artistic value of official style buildings is higher than that of folk style buildings.(5)Aiming at the comprehensive analysis of the perception results of multi-source data,a framework of multi-spatial collaborative perception of urban traditional landscape based on feature threshold is proposed.The results show that POI-Street view vision-Weibo emotional perception in Beijing’s Five Rings is dominated by high-high-high and low-low-low modes with high spatial distribution consistency.①The high-high-high model and the low-high-high model are the main models in the second ring,which indicates that the public has a high emotional evaluation of the traditional style POI,and the positive emotions are the main ones.②Official style POI-Street view vision-Weibo emotional perception is dominated by low-low-high mode and high-low-high mode,which indicates that whether the number of POI is high or low,the traditional style of Street View is significant or not,the public’s emotional evaluation is high,and positive emotion is dominant.The main emotional perception modes of folk style are low-high-high mode and low-high-low mode,which indicates that the public’s emotional evaluation of folk style is lower than that of official style.