Design of quality inspection machine vision system based on virtual instrument

introduction

Image intensifier is a photoelectric image enhancement device that amplifies weak light signals. People can use it to observe external scenes or targets under weak light conditions. The quality of the circuit system is directly affected by its manufacturing quality. Unqualified electrical systems can easily cause black spots (circuit cold soldering), bright spots (circuit short circuit), flashes, and flickering (circuit instability) in the image intensifier when it is used. Therefore, before the image intensifier is put into use, its reliability index must be evaluated and tested. However, according to the current relevant regulations and standards in my country, the reliability evaluation test of the image intensifier has the disadvantages of long test time and complex test conditions. The standard requires that when the image intensifier is tested, the test shall be divided into multiple test cycles. Each test cycle works for a total of 16 hours, and in 16 hours, it rests for 5 minutes every 55 minutes of work. There is a 2-hour interval between two adjacent test cycles, and the online working time of a test is required to be no less than 600 hours. In addition, during the test process, it is necessary to repeatedly apply various stresses (light stress, electrical stress, etc.) to the sight, and to identify and record the faults of the image intensifier under various test conditions in real time for subsequent analysis. It is precisely because of the complexity of the reliability assessment test that so far, there is no test equipment in China that can conduct reliability assessment on image intensifiers.

In recent years, with the continuous development of computer technology and digital image processing technology, machine vision has been widely used in medical imaging, industrial production, quality inspection and other fields. Virtual Instruments technology can quickly form a reliable test or measurement and control system by combining general-purpose computers with hardware through software. If the two are combined, the analysis function of machine vision and the control function of virtual instruments can be used by the system at the same time, making the system have a high performance-price ratio. Therefore, the image intensifier reliability detection machine vision system based on virtual instruments (hereinafter referred to as reliability test system) developed by combining machine vision technology with virtual instrument technology has achieved good results.

System structure and working principle

The whole system is divided into an optical-mechanical subsystem and a monitoring and recording subsystem, as shown in the figure. The optical-mechanical subsystem simulates the light stress and electrical stress in the actual working environment for the image intensifier and provides a bracket for placing the image intensifier during the test, including a light source, a large and small two-stage integrating sphere, a frosted glass, an aperture, a transmittance plate, a collimator, a night vision device bracket, a light stress switching motion device, and a luminous intensity detector.

The monitoring and recording subsystem not only identifies and records the black spots, bright spots, flashes, and flickering faults generated at the eyepiece of the image intensifier in real time, but also records the test environment parameters corresponding to the fault image. Finally, these test data are analyzed and processed to give a reasonable evaluation of the quality of the image intensifier. Considering the real-time requirements and efficiency of the system, the monitoring and recording subsystem is designed as a distributed structure, which is connected to a star network by four image machines and a management machine via a HUB. The image acquisition card PCI-1407 installed on each image machine is connected to the CCD camera to cooperate with the fault image recognition and processing software to monitor and record the fault image at the eyepiece of the corresponding image intensifier. In order to solve the problem of real-time storage of fault images, a disk array controller is also installed on each image machine. The management machine is equipped with a multi-function data acquisition card PCI-6024E to cooperate with the management machine software to monitor and record various parameters during the test, control the optical stress switching, electrical stress switching, increase and decrease of the optical machine part, etc. The control box and adapter are the interfaces between the optical machine subsystem and the detection and recording subsystem. On the one hand, it converts the control signal from the monitoring and recording subsystem into a signal that can be recognized by the motion mechanism, and on the other hand, it converts the test parameters of the optical machine part and other parts into electrical signals that can be recognized by the monitoring and recording subsystem, so that the two subsystems form a whole.

When the system is working, the operator first sets the test conditions (such as the required electrical stress) on the management machine, and then the management machine coordinates (through communication between processes on the network) the entire system to perform self-checks to ensure that all equipment is ready. After the self-check is completed, the management machine automatically sets the test conditions according to the test conditions set by the operator before starting the test. In each working cycle of the test cycle, each image machine first obtains a standard image without any fault (guaranteed by the algorithm and the operator's visual inspection). After that, the CCD camera connected to the image machine continuously converts the image at the eyepiece of the image intensifier into a standard video signal and inputs it into the image acquisition card. The image acquisition card decomposes and collects the video signal, converts it into a digital signal and transmits it to the computer for processing. The fault image recognition and processing software on the image machine processes the digital image signal in real time and identifies whether there is a fault in the image. If there is a fault, it is saved, otherwise it continues to judge the next frame. During the test, the management machine synchronously monitors the test environment parameters corresponding to each frame of the image and records them in the database. After each working cycle, the management machine controls the electrical stress applied to the image intensifier to turn off to ensure that the image intensifier rests. At the same time, it controls the motion mechanism of the optical-mechanical subsystem to change the aperture and transmittance plate, and switches the light stress to ensure that the light stress is ready before the next working cycle starts. Repeat this process until multiple test cycles of the entire test are completed.

During the development process, the virtual instrument development platform NI LabVIEW5.0 PDS and the machine vision software development platform NI IMAQ Vision 5.0 were used in combination with NI SQL ToolKit to quickly develop most of the software modules. In order to improve the processing speed of the software, VC++6.0 was used to develop the underlying fault identification program, and the CIN interface of LabVIEW was used to embed and integrate the program written in C language into the software system. Power Builder 6.0 and MS SQL Server 7.0 were used to develop the status data management module. LabVIEW and NI DataSocket were used to write the data communication and system management modules. These software modules were installed on the management computer and the image machine respectively, and the software and configuration of each image machine were exactly the same. If the system needs to be expanded, it is only necessary to connect the computer configured according to the image configuration requirements to the network.

Technical features used by the system

The reliability detection system has the following technical features, which effectively ensure the normal operation of the system.

1. The system works under unconventional light sources

The image intensifier is used to amplify weak external light. The brightness of the image at the eyepiece is about dozens of lux, and the image noise is very large, which makes it very difficult to identify faults. The solution to the problem is to use LabVIEW and IMAQ Vision to write a program to automatically adjust the black and white level of the image acquisition card and the CCD exposure factor under different illumination to ensure that fault extraction is carried out at a higher signal-to-noise ratio.

2. The system has strong real-time performance

According to the index requirements, the system must perform a series of operations such as image acquisition, preprocessing, fault identification, and image storage within 80ms, which places high demands on real-time performance. There are two main ways to solve the real-time problem. One is to use disk array technology, and the other is to use VC to write fault identification software. According to the characteristics of the black spots, bright spots, flashes, and flickering faults to be identified, the grayscale thresholds decrease in sequence, and the area thresholds increase in sequence. The system uses grayscale and area as feature parameters for fault identification. Use VC to write a program to corrode the result of the subtraction between the fault image and the standard image, and then identify the fault according to the set threshold. Compile the written program into .lsb format and embed it into the LabVIEW program using CIN contacts. After testing, this program generally takes 30ms to identify a frame of fault image, which fully meets the system requirements.

3. High-speed image streaming

Another method used in the system to improve the real-time performance of the system is RAID technology. According to different storage performance, data security and storage costs, RAID has seven basic levels from RAID0 to 6 and some combinations of basic RAID levels. RAID0 allows multiple disks to execute a certain data request of the system in parallel, and distributes continuous data to multiple disks for access, which effectively solves the bottleneck problem between disk I/O and CPU processing speed. The hard disk group on each image machine in the system is connected to the system through a RAID interface card to improve the real-time performance of the system.

4. Distributed synchronous data acquisition and control

The whole system is composed of a management computer and four image machines that work together to complete the calculation work, and there is a strict timing relationship in the working process. In the communication module written with NI DataSocket, each time the sender sends a message to the receiver, it must get the receiver's confirmation before proceeding with the follow-up work. This mechanism well ensures the coordinated work of the entire system. On the other hand, in order to facilitate the subsequent distinction of subordinate faults, it is required to record the corresponding system status when saving each fault image. For this purpose, a synchronization mechanism of same frequency, same phase and simultaneous start is adopted. Same frequency means that the frequency of image acquisition and state acquisition is the same; same phase means that the synchronous video signal parsed from any image acquisition card is connected to the synchronous input terminal of the other three CCDs to ensure that the video signals sent to the image acquisition card by the four CCDs are in the same phase; in addition, the trigger terminals of the image acquisition card and the data acquisition card are connected together, and both work in the trigger state. After any image acquisition card sends a trigger signal, the whole system starts to move.

in conclusion

In the process of combining virtual instrument technology with machine vision technology to realize the whole system, in order to improve the real-time performance of the system, the fault identification part is completed with VC++. The written algorithm is compiled into a format supported by the CIN interface of the virtual instrument development platform LabVIEW and then embedded in the whole software system. After testing, the time taken by the system to process the fault image plus the time of image acquisition and storage using this software integration method and algorithm does not exceed 40ms in total, which fully meets the requirements of the index. At the same time, by using the virtual instrument development platform to complete the control functions it is good at, developers only need to focus on the integrity of the system functions without having to consider complex details. This greatly exerts the performance of the virtual instrument, makes the system highly flexible and extensible, saves the development costs, and improves the system's performance-price ratio.