Livestock identity inspection method and system based on multi-modal image analysis

CN121789252BActive Publication Date: 2026-06-09AIEASY
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN Β· China
Patent Type
Patents(China)
Current Assignee / Owner
AIEASY
Filing Date
2026-03-04
Publication Date
2026-06-09

AI Technical Summary

Technical Problem

In existing technologies, traditional livestock identification methods are prone to wear, detachment, and tampering, and are cumbersome and costly to operate in large-scale ranches. Identification methods based on single visual features are sensitive to changes in lighting, have limited three-dimensional feature discrimination, and are difficult to achieve non-contact, automated, large-scale, rapid inspection.

Method used

A multimodal image analysis method is adopted to acquire lateral body video streams and near-field images of the head and neck. By combining a three-dimensional convolutional neural network and a two-branch feature extraction network, dynamic contour features and local biological features are extracted, target features are generated, and matched with the database.

Benefits of technology

It achieves contactless and automated livestock identification, improves the accuracy and robustness of identification verification, and overcomes the limitations of single feature recognition in complex scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121789252B_ABST
    Figure CN121789252B_ABST
Patent Text Reader

Abstract

The application discloses a kind of livestock identity inspection method and system based on multi-modal image analysis, belong to image processing technical field.The macroscopic dynamic profile and micro local biological characteristics of livestock are analyzed cooperatively to realize contactless identity recognition.It includes: synchronously obtaining side body video stream and head and neck near-field image;From video stream, extract key posture frame and generate the target single body mask sequence after shape correction, extract dynamic profile feature vector by three-dimensional convolutional neural network;Head and neck image is processed by a double-branch feature extraction network, the network extracts deep semantic features and micro-texture features enhanced by directional gradient filtering in parallel, and generates local biological feature vector by fusion;Finally, the confidence of two types of features is evaluated respectively, and adaptive weighted fusion is carried out accordingly to generate target features and database for matching.The application effectively overcomes the limitations of single feature recognition in complex scenes, improves the accuracy and practicality of identity inspection.
Need to check novelty before this filing date? Find Prior Art