基于机器视觉的数码印花工厂自动报工系统

Design of Automatic Work Reporting System for Digital Printing Factory Based on Machine Vision

摘要:

针对纺织行业中现有人工布匹生产订单报工管理困难的问题，设计并实现了一个基于机器视觉的数码印花工厂自动报工系统，将正在生产的布匹图像与订单库中的花型图案进行智能匹配，从而对生产订单进行快速准确的报工管理；首先系统客户端使用工业相机实时采集生产线上的印花布匹图像，并对图像进行滤波、归一化等预处理；然后将图像通过HTTP协议发送至系统服务端，利用基于Vision Transformer的神经网络模型进行特征提取和多层特征融合，再与订单库中的花型图案进行匹配，找出待报工订单；最后对匹配结果进行数据库管理，并将匹配结果返回客户端用于报工结果展示与确认操作；实验结果表明，该自动报工系统的订单匹配TOP-3命中率达到90.4%，满足工厂布匹生产需求。

Abstract:

Addressing the challenges of manual reporting management in textile production, a machine vision-based automatic reporting system was developed for a digital textile printing factory. This system intelligently matches the patterns of the fabrics currently in production with the designs in the order database, enabling rapid and accurate management of production orders. Initially, the client system uses an industrial camera to capture real-time images of printed fabrics on the production line, which are then subjected to preprocessing such as filtering and normalization. These images are then transmitted to the server system via HTTP protocol, where a neural network model based on Vision Transformer is employed for feature extraction and multi-layer feature fusion. Subsequently, the images are matched with pattern designs in the order database to identify the orders needing reporting. Finally, the matching results are managed in the database and returned to the client for display and confirmation of the reporting results. Experimental results demonstrate that this automatic reporting system achieves a TOP-3 match rate of 90.4%, fulfilling the needs of factory fabric production.