Landmark Visual Salience Model Based On Eye Tracking Experiment

With the progress of science and technology,a variety of navigation electronic products have been developed rapidly,which can provide the route guidance information for travelers under different travel modes to help people with a shorter time,distance or less cognitive load to reach the destination.At present,many existing route planning systems are providing ‘turn by turn – distance' route instructions.However,the landmark knowledge which is consistent with the cognitive habits of way finding has not been fully utilized.In recent years,with the development and improvement of the landmark theory and extraction methods,especially based on the salience model as the most classical method to extracte landmark automatically,which from different aspects including visual salience,structural salience and semantic salience to measure the possibility of spatial objects as landmarks.Among them,the visual salience is the primary factor to measure a spatial object as landmark,but there are still some problems.Firstly,the indicators to be considered are not comprehensive and can not fully reflect the landmark characteristics.Secondly,the weight of the indicators given has a strong subjectivity.Finally,it's lack of systematic experimental test.On the other hand,with the development of eye tracker to high accuracy,intelligent,portable,it is gradually used to test the landmark theory and achieve significant research results.At present,the quantitative evaluation model using eye movement experiment has made important progress in related fields.These research results provide important ideas and methods for the construction of landmark visual significance assessment model based on eye tracking experiment.Therefore,based on eye movement cognitive experiment,this paper tries to explore the quantitative relationship between visual attention(eye movement ind icators)and landmark visual salience.Base on it,the statistical regression method is used to analyze the weight of visual factors and construct a landmark visual salience model.The main work of this paper includes:(1)The research and development of landmark visual salience model and the application of eye movement experiment are deeply discussed.The purpose and significance of this study are analyzed.The main research ideas and framework are proposed based on the perspective of cognitive experiment.(2)The theory and method of landmark visual salience model are analyzed.The landmark salience theory,the influent factors of the landmark salience,and the method for evaluating visual salience based on eye movement experiment.O n the basis of the above research and analysis,the paper put forward the research ideas and research steps.(3)Some visual attributes in the visual salience model are described and measured,including the color,shape,size and visibility of building.Then,the difference of each indicator among buildings is calculated by outlier detection method and the calculation steps are given in detail.At last,the formula of the overall difference degree is given by each influencing factor.(4)The saliency of the street building image(taking into account the influence of comprehensive factors)based on eye tracking are evaluated by some subjects,a mathematical relationship model between eye movement indicators and visual salience are constructed.In order to objectively reflect people's visual attention to spatial objects in real scenes,First of all,the visual significance assessment experiment of Street building landmark is designed and implementated.It involves the selection of subjects,the production of experimental materials and the arrangement of experimental tasks.Then,the eye tracking experimental data is collected and preprocessed.Finally,statistical analysis method is used to select the eye movement ind icators related to the visual saliency of landmark,this paper attempts to establish the mathematical relationship between the landmark visual salience and eye movement indicators,and it is explained and verified effectively.(5)The landmark visual salience model composed of compound visual factors is constructed.First of all,a virtual environment controlled by factors for conducting the landmark cognitive experiment is builded.Secondly,based on the research of the first part(3),the difference degree of each factor is calculated.At the same time,the data of eye movement ind icators of this experiment are collected and the building saliency value is calculated according to the mathematical relation constructed in part(4).Finally,the influence weight of the factors is obtained by multiple linear regression analysis and the final visual salience model is constructed.From multidimensional perspectives,the model is strictly analyzed and tested.