Automobile assembly action recognition method of multi-channel one-dimensional convolutional neural network 1DCNN model

Through the 1DCNN model and motion capture equipment of multi-channel one-dimensional convolutional neural network, intelligent segmentation and recognition of automobile assembly actions are realized, solving the problems of low analysis efficiency and insufficient accuracy in the existing technology, and improving the efficiency and accuracy of working hours verification.

CN119691607AActive Publication Date: 2025-03-25CHINA FAW CO LTD
6 Cites 0 Cited by

Patent Information

Application Number
CN202411572078.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-06
Publication Date
2025-03-25
Estimated Expiration
2044-11-06

AI Technical Summary

Technical Problem

The prior art has problems such as low analysis efficiency, difficulty in mastering methods and inaccurate artificial segmentation in automotive assembly labor-time measurement, especially in scenarios where labor-time results are required to be highly accurate.

Method used

The 1DCNN model of multi-channel one-dimensional convolutional neural network is adopted to collect the car assembly action data through the motion capture device, divide it into upper and lower body action categories, and build an action recognition model to realize intelligent segmentation and recognition of actions.

Benefits of technology

It improves the work efficiency of action segmentation, reduces the impact of technology and experience among engineers on the results, and improves the accuracy and automation of working hours verification.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention belongs to the technical field of automobile assembly man-hour analysis, and particularly relates to an automobile assembly action recognition method of a multi-channel one-dimensional convolutional neural network 1DCNN model. The automobile assembly actions are divided into upper half body actions and lower half body actions; acquiring sensor data of different actions executed by a worker, and building a multi-channel one-dimensional convolutional neural network 1DCNN recognition model; respectively training to obtain an upper body assembly action recognition model and a lower body assembly action recognition model; acquiring normal working sensor data of a worker, inputting the normal working sensor data into the upper half body assembly action recognition model and the lower half body assembly action recognition model, extracting a first column and a last column from prediction sequence data output by the upper half body assembly action recognition model and the lower half body assembly action recognition model, and converting the working time length to obtain a second column and a third column; the recognition of the automobile assembly action is completed; the method is applied to an automobile assembly scene, intelligent action segmentation and recognition are achieved, the working hour verification process is simplified, and verification analysis efficiency and automation are improved.
Need to check novelty before this filing date? Find Prior Art