Development of Model-based Gait Recognition based on Probability Distributions of Joinits' Positions and Continuous Contour Tracking

2020 Fiscal Year Final Research Report

• Project/Area Number: 17K18379
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Perceptual information processing; Social systems engineering/Safety system
• Research Institution: National Research Institute of Police Science
• Principal Investigator: Imoto Daisuke 科学警察研究所, 法科学第二部, 研究員 (10760902)
• Project Period (FY): 2017-04-01 – 2021-03-31
• Keywords: 法科学 / 歩容解析 / モデルベース手法 / 機械学習 / 輪郭動態トラッキング / 高さ制約

Outline of Final Research Achievements

In this study, we tried to solve the problems of human footage analysis caused by low frame rate and difference in clothing and viewing directions. First, we developed a novel boundary dynamics tracking algorithm that is suitable for human footage analysis based on height constraint assumption (proposed method I). The proposed method I can successfully compensate the information loss between the two silhouette images, and showed improvement in accuracy under low frame rate conditions in the applications of gait analysis and 3D human body shape reconstruction. In addition, the analysis method using feature points and dynamic shape features (Proposed Method II) and the analysis method based on 3D camera calibration (Proposed Method III) can improve the accuracy of gait analysis under the differences in clothing and viewing directions, respectively.

Free Research Field

情報学

Academic Significance and Societal Importance of the Research Achievements

本研究により、低フレームレート、服装及び撮影角度の相違に起因する歩容解析の精度向上が図られ、法科学における応用が期待される。特に、低フレームレートと撮影角度の条件緩和（提案手法I及びIII）については、アルゴリズムの適用における映像としての撮影角度条件の制約が少ない（注：比較映像同士の条件の制約はある）ため、今後の実応用が期待される。また、細胞・流体の解析に用いられてきた既存の輪郭動態のトラッキング（レベルセット法、最小二乗基準による方法）は人物映像に適さないことが明らかになったことで、提案手法Iの高さ制約の仮定の人物映像解析一般への波及が期待される。