How to carry out testing for visual inspection projects
#01 What is Vision Inspection Equipment
Visual inspection refers to converting the captured target into an image signal through machine vision products (that is, image capture device, divided into CMOS and CCD) , and sending it to a dedicated image processing system. According to pixel distribution, brightness, color and other information, it is converted into Digitized signals ; the image system performs various operations on these signals to extract the characteristics of the target , and then controls the on-site equipment actions according to the results of the discrimination. With the rapid landing of artificial intelligence technology and the continuous development of the intelligent robot industry, visual inspection machines are bursting out with stronger vitality. The typical structure of visual inspection equipment design mainly includes five parts, namely: lighting, lens, camera, image acquisition and motion control system .
#02 What are the testing stages of visual inspection projects?
1. Software testing
Focus on whether the function implementation is consistent with customer needs , ensure the correctness of the software process, and the correct application logic relationship .
By controlling the input of the software , different test results can be achieved, and whether the test is correct and accurate through the difference between the input and output , so as to find the loopholes in the system, and start the cycle process of R & D, modification and test verification.
2. Hardware test
Carry out hardware reliability tests for the hardware itself and the environment , such as aging tests, compatibility tests, failure rate tests , etc.
It is necessary to test common hardware configurations to ensure that the software can run under various hardware configuration environments, such as compatibility tests with the whole machine, boards and peripherals .
3. Joint debugging test
Mainly test the software and hardware joint debugging function : verify the correctness of electrical and software signal communication logic, light source, camera and other hardware triggering functions such as photographing and scanning, and the statistics of detection status, etc.
In case of docking with external systems, it is necessary to check whether the basic functions of the modules between the systems, the detection capabilities of the algorithms, and the tempo meet the customer's expected standards.
4. Model testing
For model testing, focus on model functional testing, performance testing, evaluation of model indicators and analysis of indicator results .
Methods: Take A/B testing, robustness testing, offline testing, online testing and other means to evaluate the reliability and stability of the model.
Offline test/online test: Historical data is used for offline environment training and test evaluation, and an offline model version is generated. In order to ensure the stability of the model service, it is necessary to conduct online and offline consistency verification and model online stability verification tests for the model.
Consistency test: use the same batch of samples to request model services in offline and online environments, and the system compares and analyzes the model service decision results.
Stability test: The online new sample request model is used to evaluate the model according to the decision-making results of the model. Model evaluation index reference: KS, ROC, etc.
A/B testing: The algorithm model will iteratively generate multiple versions when ensuring stability. For online services such as using model A for decision-making, when model version iterations are required, model B needs to be trained in an offline environment and then deployed Go to the online environment, analyze the online and offline A/B test results, and use conversion rate and other methods to assist in deciding which version to use.
5. Performance testing
The software simulates different user volumes to carry out interface performance tests.
The hardware carries out continuous image acquisition to verify the stability of the program, the stability of signal communication, and the logical correctness of image storage.
Software and hardware joint debugging stability test: Based on the deployment scenario of the customer's production environment, implement a certain period of load operation, monitor and analyze whether the detection cycle and performance indicators meet customer expectations.
#03How to carry out the testing of visual inspection equipment?
1. Product demand analysis
Requirements analysis is the first step to start the testing work . According to the input of the requirements data of the project itself, the tester needs to extract the test scope, function points, business process, expected output, abnormal process and required time and resources of the project.
2. Test point dismantling
A test case is a set of documents consisting of preconditions, test inputs, execution conditions, and expected results to complete a specific requirement or target test, and to reflect test plans, methods, techniques, and strategies.
Test engineers need to design test cases based on the needs of specific projects, thinking in terms of mechanics, electricity, software, algorithms, etc.
3. Test execution and defect management
Differentiate testing stages: Execute test cases separately, use test tools to assist in testing, and improve test efficiency, such as: camera debugging tools, PLC debugging software, database tools, interfaces, performance testing tools, etc.
Defect definition:
(1) Failure to meet the functional requirements agreed in the contract or technical agreement;
(2) There is an error that the product manual indicates will not occur;
(3) The function exceeds the scope of the product manual;
(4) Failure to achieve the goal that should be achieved although not specified in the product manual;
(5) It is considered difficult to understand, difficult to operate, slow to run, or considered bad by end users. Test engineers need to clarify the type, priority, and severity of defects, and use defect management tools or defect lists to record, feedback, and verify defects.
4. Verification of important inspection items
The verification before shipment needs to pay attention to the following points:
Safety inspection of the line: the ground wire must be installed first.
When the visual inspection equipment is in operation, there is a high voltage. In order to avoid safety accidents, the safety measures can be greatly improved by installing a ground wire when the visual inspection equipment is frequently used.
Voltage stability check: The working voltage of general industrial equipment is 220V, including 380V.
Voltage asymmetry is likely to cause serious damage to human industrial equipment and reduce the service life of mechanical equipment.
Interface checks: such as digital inputs and outputs, actuators for triggering, communication with PLCs, robots, linear axes, database systems, etc.
Commissioning Verification: Manual and automated system testing with a small number of artifact variables.
Documentation of results: Good and bad images and evaluation of test results should be documented in detail.
#04 sets of high-quality machine vision inspection system
must-have conditions
1. High quality imaging system
The imaging system is called the "eye" of the visual inspection equipment , so the recognition ability of the "eye" is the most critical index for evaluating the imaging system. Usually, the evaluation index of the imaging system is mainly reflected in three aspects:
(1) Whether the existing defects can be found
For detection based on image methods, the most original and only data that can be based on is the color (or brightness change) on the collected image . In addition, there is no other data for reference. Therefore, a high-quality imaging system should first be an imaging system that can fully express the color change of the surface of the detected object . Therefore, in addition to choosing a high-definition camera and lens, the lighting design used to create the imaging environment is also very important , and sometimes even a lighting system specially designed for special defects appears. The so-called 100% quality inspection system actually refers to 100% full inspection in images that can fully express various defects .
(2) The minimum size of defects that can be found
The smallest measurement unit of a digital image is a pixel (pixel) , which itself does not represent the actual size of the subject.
The relationship between the actual size of the subject and the pixels is accomplished through a physical quantity called resolution. Resolution refers to the actual object size represented per unit pixel.
The smaller the resolution value, the higher the fineness of the image , the smaller the defect size that the detection system can find, and the higher the detection accuracy.
(3) Whether the image can be ingested fast enough
Just as the human eye sees a moving object, when the object moves fast enough, the human eye can no longer clearly observe the entire object.
The "eye" camera of the machine vision inspection system also has an upper limit of shooting speed, that is, the main frequency of the camera.
When the running speed of the subject exceeds the upper limit of the main frequency of the camera, the camera cannot obtain a clear and complete image, and the detection cannot continue normally.
The higher the main frequency of the camera, the faster the acquisition speed , so that the detection can be carried out efficiently.
Therefore, whether a camera with a sufficiently high main frequency is used is also a key factor in evaluating whether an imaging system is of high quality.
2. Mature image processing and analysis algorithm
The image processing and analysis algorithm is equivalent to the judgment thinking of the human brain during manual inspection in the entire inspection system .
Since machine vision is a highly practical subject, evaluating the quality of an algorithm depends more on the verification of practical applications, rather than examining whether advanced or complex theories are used in the algorithm.
Therefore, what we need is a stable and efficient image processing and analysis algorithm that can fully simulate the judgment process and method of the human brain , which is the so-called mature processing and analysis algorithm.
Therefore, when designing a processing algorithm, it is necessary to fully analyze the human judgment process and convert it into the language of the computer.
3. Good operability
Good operability mainly requires that the application and operation of the testing equipment should be simple, convenient and easy to understand . For example, the system has a friendly human-computer interaction interface and a good oriented operation design.
4. Other stable supporting facilities
Other supporting facilities refer to facilities other than the detection system, such as transmission control platform, defect processing device (rejection, alarm, marking, etc.). The requirement is that the operation must be stable, and the signal response should be timely and rapid.