Animated Territorial Data Extractor (ATDE): A Computer-Vision Method for Extracting Territorial Data from Animated Historical Maps

TL;DR

This work tackles the challenge of extracting quantitative territorial data from animated historical maps, where data are embedded in raster video frames rather than structured GIS formats. It introduces Animated Territorial Data Extractor (ATDE), a computer-vision pipeline that counts territory-colored pixels per frame using HSV-based color segmentation, RGB channel restrictions, and Direct Neighbor Filtering, with temporal alignment to produce year-aligned time series. The authors demonstrate ATDE on ten Chinese dynasties (200 BCE–1912 CE) with a shared basemap, applying scale normalization via invariant water features and relative normalization by the global maximum to enable cross-dynasty comparisons. Although not a substitute for authoritative datasets, ATDE provides a practical tool for education, rapid data exploration, and comparative analysis of territorial dynamics, and it is designed to be easy to apply to any suitable animated map video.

Abstract

We present Animated Territorial Data Extractor (ATDE), a computer vision tool that extracts quantitative territorial data from animated historical map videos. ATDE employs HSV-based color segmentation, RGB channel filtering, and Direct-Neighbor Filtering to identify and count pixels representing territorial control. Combined with preprocessing for temporal alignment and cross-video scaling, the pipeline converts animated videos into structured time-series data. We demonstrate the tool on ten Chinese dynasties (200 BCE - 1912 CE), producing year-by-year pixel counts that align with expected historical patterns. While not a substitute for authoritative historical datasets, ATDE is well-suited for educational demonstrations, preliminary data exploration, and comparative analysis of territorial dynamics. The tool requires no pre-existing shapefiles and can be applied to any animated map video given seed colors and basic configuration. Code and examples are available on GitHub.

Figures (7)

• Figure 1: Frame clock transitions. The x-axis indicates the frames throughout the preprocessed video. The y-axis depicts the sum of the absolute differences in RGB values across adjacent frames. The black trace represents this sum of absolute differences throughout the preprocessed video.
• Figure 2: Demonstration of HSV color space properties. Top: varying saturation produces different shades of the same hue. Bottom: varying hue produces distinct colors. This separation enables ATDE to capture multiple shades of dynasty colors while maintaining specificity.
• Figure 3: Color Segmentation Checks. (a) A frame from the original video, where the goal is to extract the number of pixels representing the dynasty’s territory (light blue). (b) The corresponding validation frame produced by color detection; however, because the ocean shares a similar hue, pixels outside the territory (ocean) are incorrectly counted. (c) To address this, we introduce RGB channel restrictions. Here, we exclude pixels with a green value below 150, which effectively removes most of the ocean, though some border noise remains. (d) Applying the Direct Neighbor Filtering (DNF) function with 8-neighbor connectivity eliminates this residual noise, leaving only the targeted territory.
• Figure 4: The Effect of Direct Neighbor Filtering on a $5 \times 5$ matrix
• Figure 5: Territorial Extent of the Song Dynasty (960–1279 CE). Extracted pixel counts show rapid initial expansion, a major territorial loss around 1127, and gradual decline through the 13th century. The trajectory aligns with known historical periodization of the Northern Song (960–1127) and Southern Song (1127–1279) periods.