景觀屬性對觀景行為之影響

The Effect of Landscape Properties on Viewing Behaviour

黃詩涵(Angela Shi-Han Huang)

指導教授 ： 林晏州

國立臺灣大學/生物資源暨農學院/園藝暨景觀學系/碩士(2016年)

https://doi.org/10.6342/NTU201610400

摘要

人們對景觀的理解與記憶與其觀看之處息息相關，因此瞭解人類的觀景行為變得重要，瞳位追蹤技術為最客觀且直接量測人眼觀看位置的儀器，因此本研究以瞳位追蹤技術探討人類景觀中的觀景行為。藉由文獻回顧，發現近來與景觀有關之瞳位追蹤研究仍多為探索性之初步研究，大致分為由上至下及由下至上兩種理論，由上至下多欲以認知模式進行連結探討，由下至上發展出醒目圖來預測凝視。然而，到目前為止醒目圖還無法準確預測凝視停留之位置。有鑑於此，本研究欲以由上至下之偏好以及由下至上之景觀屬性兩種方面著手以探討人類之觀景行為。本研究之主要目的在探討景觀偏好、景觀屬性及觀景行為之間的關係。首先透過專家法選出山景、水景、開闊景觀及樹林景觀之四種景觀的照片各24張，其中各4張作為測試照片，再透過ImageJ軟體將餘下80張自然景觀照片轉換為CIE L*a*b*之色彩空間，得出L*、a*、b*之平均値及標準差，並計算出彩度值；Benoit 1.31軟體計算出碎形維度值。此過程得出8個景觀屬性變項，分別為色彩組成7項、複雜度1項。各變項分別計算出相對值及差異值以便進行組間的比較。瞳位追蹤實驗將4張同類型的景觀照片組合成2×2的組圖，20組試驗隨機排序播放10秒，中間夾有2秒之校正十字，瞳位追蹤實驗後再重新播放15秒並給予1至10分的偏好評分，一張試驗內評分不可重複，研究共收集45份有效樣本。瞳位追蹤指標包含總凝視次數、總凝視時間、平均凝視時間、觀看次數及瞳孔直徑，瞳位追蹤指標及景觀偏好同景觀屬性計算出相對值及差異值。結果顯示在景觀屬性對景觀偏好之影響方面，紅綠變化大、黃藍變化大、彩度高及複雜度低的景觀之景觀偏好會高。在景觀屬性對觀景行為之影響方面，紅綠變化大及彩度高之景觀會吸引視覺注意力、黃藍變化大的景觀促使視覺探索性。在景觀偏好與觀景行為之關係方面，發現偏好高之景觀會促使人觀看。研究同時將凝視移動建立成凝視移轉機率矩陣，發現在第一次凝視時人們會偏向觀看左上方之景觀，接下來跳視多為水平移動、次為垂直移動，並有繼續由左上方、右上方、右下方到左下方之順序觀看，但當景觀之偏好高、彩度高時會凝視會停留更多次。

關鍵字

瞳位追蹤 ； 景觀偏好 ； 色彩組成 ； 複雜度 ； 凝視移轉機率

並列摘要

What we comprehend and recollect in a landscape is closely related to where we look, therefore attention is required for the understanding of human viewing behaviour. Eye tracking technology is an objective and first-hand observation of human visual perception; thus we employ eye tracking technology as a means of understanding visual perception in the landscape. Studies have shown that eye tracking studies on landscape were mostly exploratory, and can roughly be categorized into top-down and bottom-up approaches. Top-down approaches usually employ the cognitive method of landscape perception, bottom-up approaches can be seen using computer generated salience maps to predict fixations, yet so far they have not matched humans’ ability to predict fixation locations. As a result, we employ both top-down and bottom-up approaches in examining viewing behavior. The objectives of our research is aimed at finding the relationship between landscape properties, landscape preference, and viewing behaviour. First, we employed expert approach to select 24 photographs out of four landscape categories, including mountain landscape, aquatic landscape, open landscape and forest landscape, 4 photos in each category were used as a filler. The remaining 80 photographs were converted into CIE L*a*b* color space, and were measured the mean and standard deviation of L*, a*, b* respectively using ImageJ, then calculated the chroma value. We also computed the fractal dimension value of each photographs using Benoit 1.31. This process resulted in 8 landscape property variables, 7 color composition variables and 1 complexity variable. All variables were converted into relative values and difference values in order to make comparisons between groups. All photographs were arranged in a 2 by 2 across display, resulting in a total of 20 trials and 4 fillers. In the eye tracking experiment, every trial was presented for 10 seconds, with a 2-second calibration display in between. Trails were displayed again for 15 seconds while landscape preference was tested. Participants were asked to rate each landscape photograph a preference rating from 1 to 10, without giving the same score in one trial. Our study collected 45 valid samples. Eye tracking metrics includes total fixation duration, total fixation count, average fixation duration, gaze count and pupil diameter. Landscape preference scores and eye tracking metrics were both being calculated relative and difference values accordingly.
With regard to landscape properties and landscape preference relationship, high red-green variation, high yellow-blue variation and high chroma lead to high landscape preference. With regard to landscape properties and viewing behaviour, high red-green variation and high chroma lead to visual attention, high yellow-blue variation lead to visual exploration. In respect of landscape preference and viewing behaviour, high landscape preference is associated with strong visual activity. Our study also established a fixation transition matrix; results suggested that people tend to direct their first fixation to the upper left landscape, followed by horizontal saccades and vertical saccades. Gaze sequence tend to initiate in the upper left, followed by the upper right, the lower right to the lower left in a counterclockwise manner, but will fixate more in a landscape when the preference or chroma is higher.

並列關鍵字

eye tracking ； landscape preference ； color composition ； complexity ； fixation transition matrix

參考文獻

3. 李素馨、陳育文，（2006），駕車者對廣告招牌與植栽形式之視覺認知與偏好研究，戶外遊憩研究，19(3)，45-67。

4. 李素馨、雷祖強、吳仕傑、侯錦雄、曾國雄，（2015），以眼球追蹤技術檢驗眺望藏匿理論中的景觀偏好，戶外遊憩研究，28(2)，1-18。

8. 華鈺菁、林晏州，（1998），堤防與河灘地美化型式對視覺景觀偏好之影響，中國園藝，44(2)，144-159。

11. 廖婉婷、鄭佳昆、林晏州，（2013），不同地點標籤對自然度感受及偏好影響之研究，戶外遊憩研究，26(4)，31-56。

12. 謝孟倫、林晏州，（2011a），景觀色彩對自然景觀偏好之影響，戶外遊憩研究，24(2)，27-50。