| The degree of green in space is one of the focuses in urban public space construction.Urban green space can help to improve many “urban diseases”,and the avenue landscape is a typical green linear urban landscape,which is an important landscape axis of the city and one of the elements of sustainable development goals.Measuring the impact of visual features of a place and its surroundings on human visual perception has long been a concern in various fields.Most of the previous research relied on low-throughput surveys and limited data sources,which were not only labor-intensive,expensive,and time consuming,but also difficult to measure human perception of urban areas on a large scale.In this paper,the image processing and machine learning methods based on big data of Tencent Street View are proposed to measure people’s visual perception of urban avenue landscape in large-scale areas.Using the data source of Tencent Street View to evaluate the visual perception of urban avenue landscape,it not only covers a wide range,but also can be used for landscape visual evaluation of any avenue available.In order to evaluate the visual perception of urban avenues,this paper firstly selects the five major cities in the Yangtze River Delta to determine the sample avenues of each city,and combines Tencent’s server to download Tencent Street View blocks to construct visual sensation indicators for urban avenues.The image processing methods such as image color analysis,image segmentation and panoramic image stitching are used to quantitatively extract the visual perception indicators.Finally,the analytic hierarchy process is used to assign the index weights to each index,so as to calculate the visual perception index values of each sample avenue,and objectively evaluate the human visual perception of the avenue landscape.The research results show that:(1)The extraction methods of different indicators have their own characteristics and needs.In terms of color elements,the brightness and saturation of Tencent Street View images are obtained by converting RGB images into HSV images.In terms of salient regions,the Saliency Computing Model(SDSP)is used to simulate areas of interest in the avenues.In terms of green viewing rate,eight Tencent Street View images covering the 360° horizontal environment were used to calculate the green rate of each sample point along the boulevard.In terms of visual entropy,image segmentation techniques are utilized to reflect image information richness and visual complexity.In terms of the sky closure index,it is necessary to first splicing a single Tencent Street View tile into a panoramic image,and then using image segmentation technology to extract the corresponding sky color block.The five indicators can better reflect the visual experience of the avenue landscape,and the visual calculation results are similar to the reality.(2)Among the major cities in the Yangtze River Delta,Shanghai’s color factor saturation value,significant regional saturation,green view index and sky closure index are all ranked first,which is the construction city explicitly proposed in the Shanghai 12 th Five-Year Plan.The goal of the avenue is inseparable.However,Wuxi’s color factor saturation,green view index and sky closure index are the lowest,but the color factor brightness,significant regional brightness and visual entropy are the first.It can be seen that the Wuxi avenue pays more attention to the overall planning and design of the street.Its richness and variability are high.Nanjing is the first batch of ecological garden city in the country,and in 2013 launched the avenue construction project,so its color factor saturation,significant regional saturation,green view index and sky closure index are second only to Shanghai,Shanghai and Nanjing is the first two cities to propose and evaluate the avenues.Therefore,the coverage of the trees is high,which is consistent with the actual situation.(3)The average visual perception index of the sample avenues of each city is Shanghai>Hangzhou>Nanjing>Suzhou>Wuxi City.The visual perception index values in the sample sections are close to each other.The visual perception index values of the Shanghai sample avenues are all greater than 0.70,and the visual perception index values of Panyu Road have the highest scores of 0.76 in the five city avenues.It is the Mengdu Street in Nanjing,which is 0.62.The frequency of visual sensitivities of each avenue is analyzed.It is found that the visual perception index values of all sample sections are mostly distributed between 0.68 and 0.74,and the lower and higher values are less.The difference in visual perception is small. |
