Ruminant Action Detection Using Skeleton-Guided Image Cropping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Accurately determining the actions of ruminants, such as cattle, from photographed images is challenging due to their dense breeding within limited areas, making it difficult to distinguish individual orientations, postures, and movements.
Innovation Solution
A determining apparatus equipped with a skeleton estimating unit, region extracting unit, cutting unit, and determining unit to analyze and determine ruminating actions by estimating the skeleton position, extracting the ruminant region, cutting a determining area, and performing action determination based on the cut area.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If dense breeding of ruminants is performed within limited areas, then productivity is improved, but measurement precision deteriorates due to difficulty in distinguishing individual orientations and postures
Solution Approach 1:
The patent divides the determination process into multiple stages: first extracting individual ruminant regions from the image, then estimating skeleton positions for each individual, and finally determining actions based on skeleton movements. This segmentation allows accurate individual identification even in densely bred groups, resolving the contradiction between high breeding density and measurement precision.
2Measurement precision
If full photographed images are used for action determination, then measurement precision is improved, but device complexity increases due to need to process entire images
Solution Approach 1:
The patent extracts only the relevant portions of the photographed image by first identifying individual ruminant regions and then focusing analysis on those specific areas. This extraction approach maintains measurement precision by preserving all necessary visual information while reducing device complexity by eliminating processing of irrelevant background and other animals' regions.
3Measurement precision
If skeleton position estimation is performed, then measurement precision is improved for action determination, but device complexity increases
Solution Approach 1:
The patent performs preliminary actions by first extracting individual ruminant regions from the image before estimating skeleton positions. This preliminary region extraction simplifies the subsequent skeleton estimation by providing predefined boundaries and focusing computational resources on relevant areas, thereby improving measurement precision while controlling device complexity.
Data Source
AI summary
A determining apparatus includes a skeleton estimating unit configured to estimate a skeleton position of a processing-target ruminant from a photographed image; an extracting unit configured to extract a region of the processing-target ruminant from the photographed image; a cutting unit configured to cut a determining area for performing action determining of the processing-target ruminant based on the estimated skeleton position and the extracted region; and a determining unit configured to perform the action determining of the processing-target ruminant based on the cut determining area.