Research On Intelligent Perception-cognition Evaluation Of Urban Scene Space Based On High-resolution Remote Sensing Imagery

Urban environmental assessment is very important for us to understand the city.In recent years,picture data has provided sufficient data sources for observing the urban environment and obtaining environmental information.Humans acquire environmental information through vision mainly divided into two progressive processes.The first is that people are subjected to visual stimuli from the objective environment and produce specific sensory responses or impressions at the sensory level.This process is the subjective perception of environmental information stimuli.The second is that people start from their own needs and further analyze the objective environmental information to obtain a higher level of cognition and understanding.This process is people’s objective cognition of environmental information.At present,a large number of scholars have simulated the process of human obtaining environmental information,and carried out research on urban environment from two different dimensions of perception and cognition.These two types of research usually use machine learning technology/deep learning technology to achieve environmental information extraction and evaluation.By reviewing the existing research,it is found that:(1)Most of the urban environment research in the perception dimension is carried out on street view images,and researchers carry out urban perception evaluation based on street view images.However,they ignore the information from the top-down perspective,which is not conducive to grasping the overall perception of the environment.(2)Urban environment research in the cognitive dimension is carried out on both street view images and high-resolution remote sensing images.Research based on Street View imagery includes feature calculation and visual evaluation.However,urban environment research based on high-resolution remote sensing images is only limited to urban object recognition and urban scene classification,which cannot meet the needs of urban environmental assessment applications.Based on this,first of all,we propose a perception evaluation method for urban scene comfort based on high-resolution remote sensing images,which makes up for the shortcomings of existing research that ignores the topdown angle information,and provides a more comprehensive perspective for understanding the environment.Secondly,we propose a spatial cognitive evaluation method for urban scenes based on high-resolution remote sensing images to meet the needs of urban environment evaluation and expand the application boundary of urban environment research based on remote sensing images in the cognitive dimension.Finally,we carried out the application of scene perception evaluation and cognitive evaluation in Changsha city,and made a comprehensive comparative analysis on the results of perception evaluation and cognitive evaluation.Specifically,the following studies were carried out:(1)A comfort perception method for geographic scenes based on high-resolution remote sensing images.This method aims to study people’s direct subjective responses to objective environmental information stimuli.Specifically,this study proposes a quantitative method for subjective perception of comfort in geographic scenes with the help of high-resolution remote sensing image data.We use the Likert scale to investigate volunteers’ comfort perception results of geographic scene images,and generate a perception dataset.At the same time,we construct a comfort perception prediction model with the help of deep neural network.The experimental results show that the deep neural network model can effectively simulate human perception and generate accurate comfort perception results.(2)Spatial cognition method of geographic scene based on highresolution remote sensing image.This method focuses on fully excavating the characteristics of the objective environment.First,the method establishes a spatial cognition evaluation standard based on highresolution remote sensing image geographic scenes by selecting five indicators related to morphological structure and spatial distribution.Secondly,under the guidance of the established standards,the highresolution image geographic scene units are marked to complete the database construction of the spatial cognition of geographic scenes.Then,a spatial cognition evaluation model based on multi-task learning is constructed.In addition,in order to reduce the impact caused by clipping,this study adopts the spatial interpolation method for processing.Experimental results show that our model can effectively simulate human cognition and generate expressions consistent with the level of human spatial cognition,and broaden the application boundary of urban environment research under the cognitive dimension.(3)Case analysis of perception-cognition evaluation in the downtown area of Changsha.First,according to the comfort perception evaluation method,we study the comfort perception in the downtown area of Changsha,and generate a spatial distribution map of the perception of comfort attributes in Changsha.At the same time,according to the spatial cognition evaluation method of geographic scene,we generated the distribution map of spatial cognition score of Changsha geographic scene.Then,we conducted a comprehensive comparative analysis based on the results of perceptual evaluation and cognitive evaluation to study the relationship between objective environmental characteristics and subjective feelings.The analysis shows that objective environmental characteristics have both correlation and bias effects with people’s sensory experience.