Product Application | Intelligent visual recognition has been successfully implemented, solving the problem of signboard recognition
In industrial manufacturing sites, signboard recognition is a crucial part of material tracking and process control. However, practical factors such as metal reflection, character wear, angle deviation, and changes in lighting pose significant challenges to visual recognition. Recently, Changdaotong Technology deployed an intelligent visual recognition system for a client, successfully resolving these issues.
Visual recognition faces multiple challenges.
The signs are mostly made of metal and their surfaces tend to reflect light. Over time, the characters may become worn or be obscured by oil stains. Additionally, there is inevitably a deviation between the camera's installation angle and the sign, resulting in image distortion. The production process is fast, requiring the system to complete data collection, processing, and output within an extremely short time. Therefore, high stability and response speed of the algorithm are required.
The multi-protocol centralized control architecture supports the system operation.
To address these challenges, the system adopts an industrial multi-protocol centralized control integrated server, with a high-performance industrial computer as the core, to build a full-link closed-loop architecture from the access of on-site equipment to the connection with the upper system. This architecture supports various mainstream industrial Ethernet protocols such as Profinet, Modbus/TCP/RTU, EtherNet/IP, and EtherCAT, and can act as both the master station or the slave station, enabling two-way data interaction between different brands of PLCs, drives, and on-site instruments, effectively solving the problem of unified access for heterogeneous devices.
Image processing and algorithm optimization enhance the recognition rate.
The visual recognition system acquires clear images through high-resolution industrial cameras combined with optical zoom lenses. It effectively corrects angle deviations and position fluctuations by using image positioning and offset compensation techniques. By integrating camera calibration and feature detection algorithms, it enhances the worn-out characters, significantly improving the recognition success rate under complex conditions.
The visualization interface is connected with the data to achieve closed-loop management.
The built-in web server provides a visual interface based on HTML5, supporting direct access via browser both locally and remotely. It enables equipment status monitoring, parameter adjustment, alarm handling and data query, reducing the complexity of operation and maintenance. Meanwhile, the system connects to the upper-level secondary database through the OPC UA standardized interface to complete standardized data collection, encrypted transmission and unified distribution, providing stable data support for management systems such as MES and ERP, and facilitating the digital transformation of factories.
Currently, this visual recognition system has been implemented in the customer's production line and is operating stably. It has achieved automatic collection and verification of signboard information, effectively reducing the error rate of manual data entry.
Tianjin Changdaotong Technology Co., Ltd.
Please first Loginlater ~