| Plant landscape design highly unifies multiple dimensions such as artistic aesthetics,social services and ecological science,which is not only a bridge between man and nature,but also a "light adder" to the quality of production and quality The continuous pursuit of more scientific and reasonable configuration schemes has become an eternal issue for designers,so scholars have put forward evaluation standards from different angles and strive to provide more scientific theoretical support.With the development of science and technology,more tools are introduced into the evaluation of plant landscapes,and deep learning,as the core of artificial intelligence and data science,perfectly integrates computer science and statistics,in "information overload" Under the challenge,the potential value of big data can be quickly mined,and new impetus can be injected into the quantification of plant landscape,and panoramic image is an outstanding medium in the combination of deep learning and plant landscape,which provides the possibility for multi-dimensional presentation of real scenes,and combines the two technologies to cleverly break the dilemma of difficult data acquisition,processing and analysis in traditional plant landscape evaluation.This paper selects the traffic node with the richest change in plant landscape configuration in urban park as the research object,summarizes the basic characteristics and plant landscape characteristics of the scene,analyzes the traditional plant landscape evaluation index and the intervention practice of deep learning and panoramic image technology,selects a total of six evaluation indicators in three aspects: plant element composition,plant sequence change and plant function service in combination with expert consultation,and establishes a comprehensive evaluation model of plant landscape visual characteristics of urban park traffic node based on AHP evaluation method.According to the analysis of the basic environmental conditions of Harbin,12 urban parks were selected according to the five aspects of word of mouth,type,land area,construction year and area of Harbin City Park,and 641 traffic nodes of each park were marked according to the number of roads and road types.The panoramic camera is used to capture panoramic images that can meet the two directions of front,back,left and right,and up,middle and bottom of each point,so that the landscape image of the site is more complete;A semantic segmentation dataset is constructed,and a deep learning network with higher segmentation accuracy is selected to segment and extract the evaluation elements of the panoramic image.Panoramic images and deep learning technology are applied to calculate the quantitative results of each evaluation index in an all-round,scientific and quantitative manner,and the results are statistically analyzed from multiple perspectives such as overall evaluation,location of each park,park category,year of construction,number of constituent roads and road type in the form of grading,and the comprehensive visual characteristics are summarized.Based on the above research results,the data,plant configuration forms and presentation effects of plant landscape evaluation are analyzed from the cases with high and low comprehensive evaluation scores,and the reasons for the formation of different plant landscape visual effects are summarized,which provides reference for related plant landscape design,and puts forward optimization suggestions for different types of plant landscape scenes in Harbin urban park traffic nodes. |
PDF Full Text Request