ABSTRACT

Color quality evaluation is key to judging map quality, which can improve data visualization and communication. However, most existing methods for evaluating map colors are tedious and subjective manual methods. In this paper, we study sequential color schemes, a widely used map color type and propose a learning-based approach for evaluating the color quality. The approach consists of two steps. First, we extract and characterize the cartographic factors for determining the quality of sequential color schemes, such as color order, color match, color harmony, color discrimination and color uniformity. Second, we present a model to predict the color quality based on AdaBoost, a type of ensemble learning algorithm with excellent classification performance and use these factors as input data. We conduct a case study based on 781 samples and train the AdaBoost-based model to predict the quality of sequential color schemes. To evaluate the model’s performance, we calculated the area under the receiver operating characteristic (ROC) curve (AUC). The AUC values are 0.983 and 0.977 on the training data and testing data, respectively. These results indicate that the proposed approach can be used to automatically evaluate the quality of sequential color schemes for maps, which helps mapmakers select good colors.

摘要

色彩质量评价是判断地图质量的关键，可以提高数据的可视化和交流。然而，大多数现有的评估地图颜色的方法是繁琐且主观的手动方法。在本文中，我们研究了顺序配色方案，一种广泛使用的地图颜色类型，并提出了一种基于学习的颜色质量评估方法。该方法包括两个步骤。首先，我们提取和表征用于确定顺序配色方案质量的制图因素，例如颜色顺序、颜色匹配、颜色和谐、颜色辨别和颜色均匀性。其次，我们提出了一个模型来预测基于 AdaBoost 的颜色质量，AdaBoost 是一种具有出色分类性能的集成学习算法，并将这些因素用作输入数据。我们基于 781 个样本进行案例研究，并训练基于 AdaBoost 的模型来预测序列配色方案的质量。为了评估模型的性能，我们计算了受试者工作特征 (ROC) 曲线 (AUC) 下的面积。训练数据和测试数据的 AUC 值分别为 0.983 和 0.977。这些结果表明，所提出的方法可用于自动评估地图顺序配色方案的质量，这有助于地图制作者选择好的颜色。