Image Processing for Mouse Identification From Time-Series Outlines
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional technologies struggle to identify individual animals, especially mice, when they are constantly moving or in groups, and require costly equipment like cameras and thermographs for health monitoring, limiting their applications beyond health management.
Innovation Solution
An information processing device that uses a skeleton estimation model to detect animal outlines, individual identification models to recognize animals as individuals, and activity determination models to analyze their behavior, employing machine learning to process images of mice in a predetermined activity range, allowing for identification and activity analysis without additional hardware.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional biometric processing technology is used to measure animal data contactlessly, then measurement precision is improved, but individual identification capability deteriorates when animals are moving or in groups
Solution Approach 1:
The patent segments the animal's body into multiple feature points (head, body, tail, legs) and tracks their positions independently across frames. This segmentation allows reliable individual identification by creating unique movement patterns for each animal, even when they are close together or moving rapidly
Solution Approach 2:
The patent performs preliminary actions by detecting animal outlines and extracting feature points before full individual identification. The system first identifies potential animals in the scene, extracts their key feature points, and then uses temporal tracking of these feature points to establish individual identities, enabling reliable identification in dynamic conditions
2Measurement precision
If health condition monitoring is implemented using camera and thermograph, then measurement capability is improved, but device complexity and cost increase
Solution Approach 1:
The patent makes the single camera device multi-functional by using it for both individual identification and activity monitoring. The same image processing pipeline that tracks animal positions and outlines also enables detection of behaviors such as eating, drinking, and social interactions, eliminating the need for separate thermographs or specialized sensors
Solution Approach 2:
The system uses the inherent information already captured by the standard camera (image data and temporal sequences) to perform multiple functions including identification, activity monitoring, and behavioral analysis. The camera's regular image capture serves simultaneously as the basis for both individual recognition and health-related behavior monitoring, making the system self-sufficient
3Adaptability or versatility
If activity analysis is added to the identification system, then application versatility is improved, but processing complexity increases
Solution Approach 1:
The patent merges individual identification and activity monitoring into a single integrated processing system. The same feature point extraction and temporal tracking mechanisms used for identification are simultaneously utilized for activity detection, combining multiple functions into one unified processing pipeline without requiring separate systems
Solution Approach 2:
The system maintains continuous tracking of animal feature points across image frames, and this continuous data stream serves dual purposes: maintaining individual identification and enabling real-time activity monitoring. The uninterrupted temporal sequence of feature point positions provides continuous information for both identification stability and behavior analysis
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A challenge is to provide an art to identify mice as individuals from captured images of one or more mice that are doing activity in a predetermined activity range. An image processing device 2 includes an image acquisition unit 51, a part extraction unit 52, and an individual identification unit 54. The image acquisition unit 51 acquires a captured video of a state in which one or more animals are doing activity in a predetermined activity range. An outline detection unit 57 extracts the outlines of the bodies of the one or more animals from each of multiple unit images included in the video. The individual identification unit 54 analyzes the multiple outlines detected from each of the multiple unit images by the outline detection unit 57, in a time-series manner and identifies the one or more animals as individuals in each of the unit images on the basis of the analysis results. Thus, the challenge is solved.