When it comes to product inspection, more and more companies are faced with problems such as “visual inspection leads to quality change” and “inspection accuracy decreases due to lack of manpower”. Visual inspection automation is drawing attention to solve these problems. To be successful in automating visual inspection, it is important to understand the pros, cons, and flow of the application in advance. In this article, we will explain the basics of visual inspection automation, introduction steps and tips, and specific examples.
What is visual inspection automation?
Appearance inspection is the process of checking the appearance of products and parts to ensure their quality, mainly done in the manufacturing industry. This is done to detect problems such as dirt, scratches, and deformation and to eliminate defective products.
Traditional visual inspection was done visually, but human eye inspection had many disadvantages such as human error and high cost. Therefore, “automatic external inspection”, which prevents losses caused by inspections made by machines, is on the agenda.
The early systems had the problem of not being able to make flexible decisions, but now with better AI technology, the accuracy of image analysis has increased and automation of visual inspection is advancing across industries.
Benefits of moving to visual inspection automation
The introduction of visual inspection automation will help address several disadvantages of traditional visual inspection. Here are four key benefits of automation.
Save time
Mechanical inspection can significantly reduce inspection time compared to visual inspection. Also, unlike humans, machines do not change their working speed depending on the conditions, so they can continue the inspection at a constant speed.
Cost reduction
There is a perception that automating visual inspections costs a lot of money, but in the long run, there are many cases where automation can reduce costs. In addition to the personnel costs of inspectors, the cost of hiring and training new inspectors is no longer necessary and production efficiency has improved, so overall cost reductions can be expected.
Avoid human error
Visual inspection is prone to human error, which can be prevented by mechanical inspection. There is no difference in quality assessment due to differences in each person’s senses or daily conditions, and inspection accuracy can be maintained equally.
Elimination of labor shortage
Labor shortages in production facilities, decreasing birth rates, and increasing population make it difficult to keep up with skills. Automating visual inspection helps reduce labor shortages by eliminating the need to assign inspectors. Since employees can focus on other tasks, the risk of workers being overloaded or product quality deterioration is reduced.
Disadvantages of visual inspection automation
Although visual inspection automation has many advantages, it also has disadvantages that should be considered when implementing it. The main difficulties in automating visual inspection include the need for specialized knowledge and initial investment and the possibility of missing defective products even in mechanical inspection.
In visual inspection involving artificial intelligence, it is first necessary to create a large amount of sample data and train it to make defect decisions. If the quality or quantity of this data is insufficient, the accuracy of the defect decision will decrease.
It should also be noted that creating sample data itself is expensive and time-consuming. In addition, you may need to retrain the system when there is a change in the production process or when you want to adopt a new product.
In light of these issues, some companies are building visual inspection automation systems using AI in addition to deep learning. It’s important to consider what methods can be taken to avoid the pitfalls and the measures that are appropriate for your company’s inspection materials.
Flow to introduce visual inspection automation.
Automated visual inspection requires effort to design the system according to the business context and have sample data for AI learning, so it cannot be implemented immediately after the decision to implement it. In general, the installation process is carried out by following the steps below.
Define the requirements.
Typically, the first step when developing a system is to formulate a “requirements definition” to clarify the purpose of the development and the desired requirements. When setting up an automated visual inspection system, we first follow these steps to determine the direction of the project.
At this point, clarifying the goal, such as “reduce costs”, “improve quality” or “increase operational efficiency”, is the key to successful implementation.
Know and confirm.
Once the direction is determined, we move on to the stage of actually installing the system and adjusting the already built system for our company. A visual inspection automation system using AI also learns from sample data.
Collecting sample data and performing learning requires specialized knowledge, and any problems in this process will reduce the accuracy of the test. It is also recommended to review learning support materials when choosing a vendor.
Pre-implementation testing
After the system is built and set up, we perform pre-implementation testing to verify that the desired inspection accuracy has been achieved. Even if you work correctly, there may be cases where you cannot handle an unexpectedly bad product, so it is important to check before proceeding with the actual operation.
In addition to inspection accuracy, we also check various functions, including the way the inspected product is photographed, the type and configuration of cameras and lighting used for photography, and the detection of defective products.
Start of implementation
After testing and confirming that there are no problems, we will start the actual implementation. In some cases, defective products that cannot be detected after operations have started may occur, so we constantly monitor the inspection accuracy and make fine adjustments based on trends to stabilize the accuracy.
In addition, to achieve the goal of automating visual inspection and further improving quality, it is important to analyze problems and consider improvement measures even after inspection accuracy has stabilized.
Comparison Points When Choosing a Visual Inspection Automation System
When choosing a visual inspection automation system, you can reduce the risk of failure by paying attention to the following points.
Is the content identifiable?
Things that are easy to detect, such as irregular color, foreign matter, and defects, vary from system to system. Therefore, it is important to check in advance whether the system in question can detect what you want to inspect.
In terms of system features, check how well it can support situations such as “I want to introduce not only an inspection machine but also peripheral devices” and “I want to create an AI model and adjust the inspection accuracy”. This will make it easier to implement as desired.
Test decision method
There are two types of methods for identifying defective products in automatic visual inspection systems: “rule-based type” and “AI type”. The rule-based type inspects by estimating product information such as color intensity, area, and width, while the AI type inspects based on data learned by the AI.
A principle-based approach is suitable if you want to make clear decisions based on numerical evidence. The AI type is suitable when you want to deal with ambiguous situations where it is difficult to draw a line numerically, or when you use many types and shapes, and it takes time to adjust the numerical values.
Implementation and operating costs
Automated visual inspection is often touted as a cost-cutting measure, but it is said to cost several million yen or more to implement. It is important to check in advance that the implementation and operating costs, including the cost of outsourcing support for system tuning, are within budget.
Costs can vary depending on the scale of the site and the content of the inspection, so if you want to evaluate based on a specific budget, it is best to request a quote rather than relying on information on the internet.
Operation/maintenance system
Once a system is put into operation, it needs to be operated and maintained to ensure that it continues to operate smoothly. In particular, in cases such as visual inspection automation, adjustments need to be made to maintain and improve accuracy even after the system is operational.
The important thing here is the support system for post-implementation operation and maintenance. It is important to check the adjustment method, whether there are functions to support it, how much support you can get from the vendor, and choose a system that you can continue to use with confidence.
Introduction example
Sky Corporation is involved in the development of inspection systems for production lines using artificial intelligence and image recognition technology and researches and develops useful technologies for visual inspection. Applications that perform visual inspection using image recognition can detect defective products and calculate dimensions.
Specific examples where this technology can be used include measuring product dimensions in millimeters on factory production lines and detecting scratches, dents, dents, and misaligned labels on packages. In addition, rotation correction can be automatically applied based on contour information even if the captured image of the inspection target is tilted.
System development for Sky Corporation
Sky Co., Ltd. Supports the development of various business systems, including visual inspection in factories. Using the technology and know-how gained through system development in a wide range of industries, we can provide flexible support from PoC (pre-certification) to systemization.
Artificial intelligence and image recognition technology are used in all areas, including visual inspection in factories, abnormality detection in equipment, autonomous driving of vehicles, classification of products and devices, image diagnosis support in medical care, and active vehicle monitoring in logistics. industry
If artificial intelligence can perform tasks suitable for automation, it will be possible to save labor and reduce the burden on employees. If you are considering developing a system using image recognition technology, please do not hesitate to contact Sky Corporation.
Summary
Thanks to the development of artificial intelligence-based image recognition technology, automated visual inspection systems that can handle all types of cases have now been created, and inspection accuracy has also been improved. On the other hand, it is very important to understand the risks, costs, and knowledge required for long-term operation before deciding to implement a system.
First, carefully consider how to incorporate the most effective system to solve your company’s problems.