As science and technology progress, environmental designers and planers not only build landscape preference model by the variety of landscape elements and cognition factors but also use advanced biofeedback instruments to explore the function of the landscape and the observers’ physiology reaction. This study selected Huisun Experimental Forestry as case study, and classified visual landscape zones by considering environmental perception, environmental preference and related theories. Using eye-tracking and questionnaires to detect observers’ visual focus and measure preference score. Visual focuses of observation time, area and perimeter of each landscape element were analyzed via Mapinfo and Visual Basic software. The result suggested the regression model of area had the highest predictive ability, followed by perimeter model and observation time model was the least. The longer observation time of vegetation zones was, the higher preference score was. Whenever the nonvegetation objects merged in immediate and intermediate zones, there was a negative effect. In the perminter model, the more integrated and non-seperated shape of landscape zones, the higher preference score was. Water zone, perimeter of intermediate nonvegetation, observation time of immediate nonvegetation, observation time of sky, and observation time of intermediate vegetation can be used as the landscape preference predicted factors. This study suggests the application of eye-tracking technology is worthy for advanced research instrument which are associated with landscape preference.