Landscape character assessment (LCA) is an effective tool for understanding people's perceptions and preferences of landscape characteristics. Other than the assessment indicators and subjects, the reliability of photos as assessment objects is equally important for the LCA result. However, the commonly used onsite photos are mainly obtained at randomly selected locations by the researchers. We can neither know whether those photos represent the researchers' own preferences, nor, to our best knowledge, can their reliability be tested scientifically. This method is also difficult to apply in large-scale geographical areas. To address these issues, we (1) propose an improved method including the protocols of photography and the sampling of photography locations, in which the fractal principle and stratified random sampling method were combined to minimize the effects of the researchers' preferences and other factors; (2) apply the method to the Guanzhong region as an example, and obtain sampling photos and their geographical coordinates, which can be used as a data package for LCA; (3) use Fractalyse to test the sampled result and receive good validity. In conclusion, this study extends the methodological chain of the LCA and supports the application of LCA in large-scale regions.

景观特征评估（LCA）是了解人们对景观特征的感知和偏好的有效工具。除了评估指标和对象，照片作为评估对象的可靠性对于LCA结果同样重要。然而，常用的现场照片主要是研究人员在随机选择的地点获得的。我们既无法知道这些照片是否代表研究人员自己的偏好，据我们所知，也无法对其可靠性进行科学检验。这种方法也很难应用于大范围的地理区域。为了解决这些问题，我们（1）提出了一种改进的方法，包括摄影协议和摄影地点采样，其中分形原理和分层随机抽样方法相结合，最大限度地减少了研究人员偏好和其他因素的影响；(2)以关中地区为例，获取采样照片及其地理坐标，作为LCA的数据包；(3) 使用Fractalyse对采样结果进行检验，得到良好的有效性。总之，本研究扩展了 LCA 的方法链，支持 LCA 在大范围区域的应用。