Group behavior tracking of Daphnia magna based on motion estimation and appearance models
- 문서 유형
- 연구출판물 · 학술지 논문
- 등재
- SCIEScopus
- 저널 순위
- JCR Q2Scopus Q1
- 저널
- ECOLOGICAL INFORMATICS61 · 2021
- 피인용
- Web of Science 3Scopus 3
- FWCI (Scopus)
- 0.27
- JCR IF
- JCR 2020 3.142
- ISSN
- 1574-9541
저자
초록
Daphnia magna is a widely adopted biological indicator for studying the environmental toxicity in aquatic ecosystems. Due to the tiny size and appearance similarity of Daphnia most behavioral tracking systems can only record the behavioral movement of a single individual. In this work, a novel behavioral tracking scheme is proposed to track the group movement of Daphnia and record the accurate trajectory of each individual. Individual position, motion estimation and appearance similarity are incorporated to solve the multiple objects tracking problem. Optical flow is utilized to estimate individual motion and identify each individual in occlusions. Normalized cross correlation is employed to measure appearance similarity of Daphnia individuals. According to these feature movement trajectories are assigned to each individual by linear optimization. The main advantage of the proposed method is to segment individual Daphnia from occluded images according to the appearance and motion characters. Consequently, the success rate of tracking under occlusions could be significantly improved. The proposed method was tested on 8 video clips which captured in two different observation conditions. The correct identification ratio reached 94.74% and the average success ratio was 87.68%. The comparison tests proved that our method was superior to a well-known multiple object tracking method in dealing with occlusions. The proposed Daphnia tracking scheme was reliable and could provide accurate movement information for behavioral and environmental studies.
키워드
Multiple individual trackingOcclusion segmentationSORTBehavioral monitoringMotion estimation
식별자
- Scopus EID
- 2-s2.0-85100431804
- WoS UID
- 000632605900007



로그인
연구물 목록으로