Research On Integrated Subjective-objective Image Of Automobile Form

Automobile enjoys a significant standing in our daily life.China's automobile sales volume has been increasing thanks to the support from the Chinese government.Meanwhile,consumers' demand on automobile has shifted from mere material functions to a sentimental level which,serves as a vital medium linking sales and consumers and yields positive impact on the automobile industry,in that the image recognition towards the automobile stylings conveys the sentimental demand of consumers.With the development of information technology,the traditional model of store sales has transformed into a combination of online car-selection and offline consumption.To make it easier for consumers to select vehicles,an idea of constructing a retrieval platform of styling images is proposed in this thesis,positioning such images according to standard criteria,filling up the gap of car-selection platforms in the current automobile industry,and passing through the communication barrier in the sales process,so as to raise the efficiency of car selection.In order to obtain the universal method of positioning the styling images of automobile,relevant study,with A-level car as the object,is conducted on the theories,key technologies,experimental systems and data analysis of image recognition.Study contents are as follows:(1)An analysis method featuring information visualization is applied on the research results of automobile styling images during the latest decade via citespace,thus summarizing the focal points and development trend.In addition,analysis and discussion about the styling images from perspectives such as theoretical basis,quantitative methods and measuring techniques are also included.(2)Experimental study on the viewing angles of automobile is carried out.Combining the subjective results from user interview and objective ones from eye-tracking experiment,the 45° front from multiple angles of 0° front,45° front,90°side,135° back and 180° back is selected as the criterion of sample selection,hence easing the visual burden.(3)Study on the presentation forms of image samples is also conducted in this thesis.Via expert interviews and analytic hierarchy process(AHP),the evaluation and selection scheme that compares 360° rotation of the vehicles via graphic interchange format(gif)simulation with traditional single-view two-dimensional pictures is proposed in the thesis,leading to the result of the former eventually taking the place of the latter,thus optimizing the access channels of frontend data from the image recognition experiment.(4)The typical samples of A-level car are screened out via focus group,and image words via similarity detection.In addition,users' subject positioning towards typical samples is obtained by means of semantic differential(SD)for the purpose of screening according to the questionnaire survey aided with accuracy test by electroencephalographic technology.Hence the determination of image positioning of automobile stylings.(5)A standard database is established according to the basic model of the retrieval platform and the words and labels for retrieval defined in this thesis,thereby building the mapping relations between retrieval labels and the database,and eventually facilitating the establishment of an image retrieval platform designed to help consumers with car selection.Research results of this thesis may help automobile sellers understand the consumer psychology and then make recommendations in accordance with the consumer demand,thereby passing through the communication barrier and increasing the efficiency of car selection.