基於深度學習的街道景觀品質偏好模型構建及關鍵影響因素分析

Construction of a Deep Learning-Based Model for Street Landscape Quality Preferences and Analysis of Key Influencing Factors

隨著城市經濟快速發展和城市化進程加快，政府在制定城市街道建設政策時通常優先考慮經濟效益，忽略了街道品質的重要性。然而，街道品質對居民的日常互動和行為習慣有顯著影響。傳統街道改進與設計方法需要大量時間和人力進行實地調查和評估，並通過訪談和問卷獲取數據，這些方法不僅耗時，而且樣本量有限、成本高、效率低。隨著技術進步，街道景觀品質評估領域出現了新發展。深度學習(Deep Learning, DL)等先進技術的應用，不僅推動了街道景觀品質評估技術的發展，還為探討街道景觀品質及其影響因素提供了新視角和工具。
本研究旨在探討影響街道景觀品質的關鍵因素，並利用DL及相關技術構建街道景觀品質偏好模型，以克服傳統方法在效率和準確性上的局限。首先，通過文獻回顧法確定了一系列影響街道景觀品質的相關因素，並從中篩選出能被DL等技術識別的5個構面及19個因素。接著，運用德爾菲法篩選出5個構面及16個關鍵因素，最後使用層級分析法計算這些關鍵因素的權重，並進行重要性排序。通過這些步驟，構建了街道景觀品質偏好模型。希望能為城市規劃師、設計師和政策制定者提高在街道景觀品質調查和硏究方面的效率和準確性，為未來的街道設計和政策制定提供科學依據。

With the rapid development of urban economies and the acceleration of urbanization, governments often prioritize economic benefits over the quality of street environments when formulating urban street construction policies. However, the quality of streets significantly impacts residents' daily interactions and behavior patterns.
Traditional methods for improving and designing streets require extensive time and labor for on-site surveys and evaluations, and data collection through interviews and questionnaires. These methods are time-consuming, have limited sample sizes, high costs, and low efficiency. With technological advancements, new developments have emerged in the field of street landscape quality assessment. The application of advanced technologies, such as Deep Learning (DL), has not only driven the development of street landscape quality assessment techniques but also provided new perspectives and tools for exploring street landscape quality and its influencing factors.
This study aims to investigate the key factors influencing street landscape quality and to develop a street landscape quality preference model using DL and related technologies to overcome the limitations of traditional methods in terms of efficiency and accuracy. First, a literature review method was used to identify a series of factors related to street landscape quality, from which five dimensions and nineteen factors that can be recognized by DL and other technologies were selected. Next, the Delphi method was employed to filter out five dimensions and sixteen key factors. Finally, the Analytic Hierarchy Process (AHP) was used to calculate the weights of these key factors and rank their importance. Through these steps, a street landscape quality preference model was constructed. The goal is to improve the efficiency and accuracy of street landscape quality surveys and research for urban planners, designers, and policymakers, providing scientific support for future street design and policy-making.