Preface The development of modern industry has placed increasingly higher demands on the surface quality of industrial components and materials, and has also raised the bar for the inspection, identification, and evaluation of these surfaces. The evaluation of the microscopic characteristics of surface processing has shifted from a qualitative comprehensive evaluation stage to a quantitative, standardized parameter evaluation stage. However, the measurement and evaluation of these parameters are difficult, and related testing processes during production increase production costs. Taking the production line of TLG-23, the core material for the gas diffusion layer of proton exchange membrane fuel cells, as an example, the production of this material involves five processes: initial processing, molding, carbonization, catalysis, and PTFE treatment. The quality of each process directly affects the quality of the final product. The traditional production process involves sampling the material after each process has completed a certain quantity and sending it to a testing laboratory for measurement. Even slight changes in the testing environment can severely impact the test results. This production method significantly reduces production efficiency, and the products fall far short of market demand. Therefore, finding an efficient solution for the detection of surface microscopic characteristics has become an urgent problem to solve. To address the aforementioned issues, we designed a vision inspection system for fractal characteristic analysis of production material surfaces using NI's virtual instrumentation technology, successfully achieving online product inspection. The system hardware platform utilizes NI's mature computer technology and modular, high-performance hardware, providing robust support for our system implementation. The PXI system offers optimal performance, while the Compact Vision System (CVS), a product within NI's PAC (Programmable Controller) series specifically designed for vision acquisition and inspection, provides the most robust hardware architecture. The entire inspection system is a high-performance integrated real-time system built using NI CVS and PXI products. The PXI system, composed of the embedded real-time PXI controller PXI-8146 RT and PXI chassis modules, plays a crucial role as the control center. The real-time compact vision system uses the CVS-1450, handling image acquisition and preliminary processing at the front-end industrial site. Distributed across multiple production processes, the CVS system acquires data via 1394 cameras and performs front-end processing before transmitting the data to the PXI control center via industrial Ethernet. (As shown in Figure 1) Figure 1 Composition of a machine vision inspection system based on fractal theory **[b]Implementation of System Software Platform[/b]** The software of this inspection system was mainly developed using LabVIEW 7.1, a visual graphical programming software from NI. Its user-friendly interface, convenient and fast programming methods, and rich and powerful function library provided support for the development of this complex visual inspection system. The inspection software runs at the control center, utilizing the powerful performance of PXI to perform fractal analysis on the images, obtain the corresponding inspection parameter results, and provide a user monitoring and operation interface. The core of the entire inspection system lies in its software algorithm, which utilizes the currently popular fractal theory. Fractal theory has been successfully applied to describe the randomness, disorder, self-similarity, self-affineness, and multi-scale properties exhibited by engineering surfaces. This allows the identification and evaluation of the microscopic characteristics of industrial processing and production surfaces to be applied to practical engineering. During the system construction process, the Weierstrass-Mandelbrot fractal function (as shown in Equation 1) was used to identify the surface morphology through power spectrum. The final machine vision inspection system analyzes the surface morphology change process by calculating characteristic parameters such as the fractal dimension of the processed workpiece surface combined with the structure function. The core algorithm of the system utilizes a fractional Brownian incremental field model and employs multifractal spectrum for in-depth surface analysis. The calculation of fractal characteristic parameters in the core algorithm is already written in Visual C++ 6.0. LabVIEW, an open software platform, provides various tools for interaction with external code or programs, making it easy to load the dynamic link library files of existing fractal characteristic parameter calculation algorithms into LabVIEW. For system performance testing and verification, the entire system was deployed on the TLG-23 production line for the core material of the gas diffusion layer in proton exchange membrane fuel cells. The detection and control software developed using LabVIEW ran on the PXI platform. Figure 2 shows the interface of a real-time test conducted on June 29, 2005. Figure 2 shows the test interface of the machine vision inspection system based on fractal theory on the TLG-23 production line. The production line's industrial network also includes other automated production equipment. The inspection system, composed of CVS and PXI, can seamlessly connect with them, ensuring real-time online inspection throughout the entire testing process. This not only eliminates the previously time-consuming and complex inspection process but also significantly improves production efficiency. Currently, the output of TLG-23 has increased from 100 sheets/day on a single production line to 835 sheets/day. Simultaneously, the inspection accuracy has also greatly improved. Figure 3 shows a recent production test report; the test results fully meet actual production needs and achieve the design requirements. **System Advantages** 1. Meeting the Needs of Industrial Environments While existing cameras can complete most image acquisition tasks, visual inspection systems built using these traditional image acquisition methods often experience detection errors or even malfunction when operating in harsh industrial production environments (such as high temperatures and severe electromagnetic interference). On our TLG-23 production line, many processes are carried out under harsh conditions such as high temperature and high pressure. In the initial stages of system design, traditional cameras were tried, but during testing on the TLG-23 production line, the carbonization process, conducted at extremely harsh temperatures, caused the vision system to completely fail. According to fractal theory, maximizing the stability and accuracy of the on-site visual acquisition system is crucial to significantly improving the detection accuracy of the machine vision inspection system. Later, we adopted NI's CVS 1450 system. This system is specifically designed for harsh industrial environments with high temperatures, electromagnetic interference, noise, and vibration, and can operate normally under extreme temperatures. The images acquired by the system throughout the carbonization process did not exhibit image distortion or blurring due to the high temperatures. 2. Thanks to NI's virtual instrumentation technology, this system boasts strong scalability and secondary development capabilities, offering exceptional flexibility, integration, and robustness. It can communicate with the vast majority of modern automated equipment, allowing the introduction of this advanced vision system without replacing existing production control equipment. On the hardware side, existing automated control equipment on our production lines can be integrated into our industrial network simply by selecting different configurations and options in the software settings, enabling collaborative operation and saving production costs to a certain extent while expanding the system's application scope. For a completely new production line, we can use NI's FieldPoint distributed I/O to replace PLCs and other devices. These FieldPoint distributed I/Os combine the reliability of PLCs with the functionality of PCs, providing accurate measurement and control even under the harshest working conditions. Furthermore, these FieldPoint distributed I/Os maintain consistent communication with the Compact Vision system and PXI control center, enabling joint monitoring of measurement data through a user-friendly LabVIEW software-designed user panel. On the software side, the system uses structure functions to distinguish different material surfaces. The system can be configured before operation, allowing for the selection of processing methods (such as polishing, grinding, precision turning, electroplating, etc.), offering greater flexibility and enabling its application in a wider range of industrial material production sites. Its user interface is user-friendly, maintenance is convenient, and upgrades are easy. Due to its broad application range, the system is currently being used for product testing on the production line of another material, WHQ-TJ-4G, and its measurement and analysis results fully meet design requirements. In the future, it can also be applied to the production and testing of high-precision engineering materials such as disk coatings and stainless steel parts. **Conclusion** The high-efficiency machine vision inspection system based on fractal theory was constructed using NI's CVS vision system, PXI controller, and LabVIEW 7.1 as its hardware and software platform. This inspection system successfully applied fractal theory to the identification and evaluation of engineering surface morphology in practical engineering, achieving satisfactory results. It can be said that the online inspection of this system fundamentally changes the original production mode of certain special industries, removing the previously restrictive link in networking production systems and realizing true production automation. Its outstanding advantages are: ① The inspection system adopts a simple bus-type network structure; ② The system structure is open and expandable; ③ It has a powerful human-machine interface, making maintenance simple; ④ It has strong secondary development capabilities and a wide range of applications; ⑤ Considering the system's stability and construction costs, its reliability is effectively guaranteed; ⑥ It is easy to integrate with other NI hardware and software to build more powerful and complete industrial control and measurement systems, facilitating standardization. Editor: He Shiping