A Hybrid Approach: Image Processing Techniques and Deep Learning Method for Cow Detection and Tracking System
- Resource Type
- Conference
- Authors
- Mar, Cho Cho; Zin, Thi Thi; Kobayashi, Ikuo; Horii, Yoichiro
- Source
- 2022 IEEE 4th Global Conference on Life Sciences and Technologies (LifeTech) Life Sciences and Technologies (LifeTech), 2022 IEEE 4th Global Conference on. :566-567 Mar, 2022
- Subject
- Bioengineering
Computing and Processing
General Topics for Engineers
Robotics and Control Systems
Signal Processing and Analysis
Deep learning
Image segmentation
Image color analysis
Switches
Cows
Feature extraction
Robustness
cow detection
cow tracking
color features
CNN features
location feature
- Language
Cow detection and tracking system plays an important role in cattle farming and diary community to reduce expenses and workload. This research presents how the conventional image processing techniques can be combined with deep learning concepts to establish cow detection and tracking system. Specifically, we first employ a Hybrid Task Cascade (HTC) instance segmentation network for cow detection. We then built the multiple objects tracking (MOT) algorithm utilizing location and appearance cues (color and CNN features) to carry out cow tracking process. To leverage the robustness of the system, we also considered the recent features from the previous tracked cow.