Machine Vision: new horizons for quality control using vision analysis

Machine vision is a rapidly growing area of technology that plays a key role in quality control processes in many industries. To achieve optimum performance from machine vision systems, it is important to understand not only their components, but also to analyse the impact of the environment in which they operate.

This article discusses the key parameters of machine vision systems and the characteristic industrial infrastructure in which they are implemented. In addition, the various applications of these systems are presented, highlighting their importance in modern industry.

System architecture incorporating machine vision

An industrial infrastructure incorporating machine vision in the quality control process typically consists of the following components:

• Conveyor, robot or static inspection station
• Camera, sensor and lamp mounting system
• Illumination
• Cameras
• Computer running machine vision software
• Optical or inductive sensor
• Result presentation system (light or sound)
• Data archiving infrastructure (e.g. NAS drive)

The selection of the right vision system components (Figure 1) at the start of a project plays a key role in its successful application.

Vision Cameras

A key element of machine vision solutions are vision cameras (Illustration 2). COIG SA offers products from leading manufacturers in this field and, as a partner of the manufacturer HIKROBOT, can offer high quality products at very attractive prices.

The choice of the right camera depends on the specific application. The wide range of products available on the market can be divided into three main groups: matrix, line and smart cameras. Smart cameras have built-in software and lighting and are mainly used for less complex tasks such as counting parts or taking measurements.

The choice of camera type for a vision system also depends on the colour range to be captured. Monochrome (greyscale) cameras are ideal for tasks requiring high resolution and accuracy, while multi-colour cameras are more suitable for applications requiring colour identification.

These cameras are often equipped with CCD or CMOS sensors, which vary in terms of light sensitivity and readout speed. Cameras with different resolutions and frame rates are used depending on the application.

Video camera parameters

Resolution

In general, a higher resolution produces a clearer image. However, increasing the resolution also increases the processing power required and the price of the camera. It is important to determine the optimum resolution when selecting a camera and after the first images have been captured. The range of matrix camera resolutions on the market is concentrated between 1MP and 40MP. Cameras with higher resolutions are also available. However, with very high resolutions, in addition to the limitations mentioned above, the maximum frame rate drops significantly.

Communication interfaces

The image captured by the CCTV camera is sent via an appropriate communication interface to a computer for processing, analysis and interpretation. Manufacturers currently use the following solutions in the devices they offer:

• FireWire – a standard for serial data transfer defined in the IEEE 1394 document; the IEEE 1394a standard proposed by camera manufacturers allows data transfer at speeds of up to 400 Mb/s; the IEEE 1394b standard allows data transfer at speeds of up to 800 Mb/s;
• Gigabit Ethernet – standard that enables data transfer at speeds of up to 1 Gb/s; uses various types of transmission media: fibre optic, twisted pair, coaxial cable;
• USB 2.0 – a standard that enables data transfer at speeds of up to 480 Mb/s;
• USB 3.0 – a standard that allows data transfer at speeds of up to 4 Gbps;
• Camera Link – serial communication protocol designed for video systems to standardise scientific and industrial video equipment; capable of transferring data at a rate of 132MB/s.

Sensor

The size of the light-sensitive sensor has a significant effect on the image received by the camera. Smaller sensors are used with inferior lenses, and the limited size of these sensors means that the image quality of all lenses used is sharpest in the centre, but degrades progressively as you approach the edges. As the sensor size increases, the depth of field for a given lens decreases. This is because larger sensors require a much shorter distance between the lens and the subject, or a longer focal length.

Lighting

Adequate lighting is one of the essential elements for the successful operation of machine vision systems. Light can be used in different configurations depending on the application. Steady light is used for constant lighting conditions, while pulsed light works well for tasks that require the reduction of glare effects.

Light Source

The choice of possible light sources is very large and constantly growing. Currently, one of the most popular is LED lighting, which is characterized by energy efficiency and relatively small dimensions. In addition, it is characterized by long life, easy control, durability and wide possibilities of forming the lamps, for example, into rings. A popular alternative to LED lighting is the use of fluorescent or halogen lamps, which allow strong light to be emitted over a large area. Metal halide lamps, incandescent lamps and even lasers are also used.

Light input

The angle of incidence and reflection plays a key role in the lighting process. There are several positioning techniques, but the most common is to place the light source directly in front of the object to be inspected at such an angle that the reflected beam returns to the source. This allows the detection of any imperfections or scratches on the surface of the item being inspected.

The introduction of a scattering element reduces the ability to detect imperfections on the surface of the object, but makes it easier to find the edges of the object. This causes the object to stand out more from its background.

Shifting the light source so that it falls at a non-perpendicular angle on only part of the object of interest makes it possible to better distinguish more protruding elements.

Lighting on the back of the object is also used as needed to detect holes or, in the case of semi-transparent objects, the uneven density or thickness of the object’s material.

Software 

There are many machine vision software solutions on the market, but in this article we will focus on three that COIG SA specializes in: SC MVS (from HIKROBOT), Aurora Vision and a proprietary application – MV COIG. The following tools are able to satisfy most of the needs arising from different production applications.

SC MVS and Vision Master is a machine vision software from HIKROBOT that provides customers with algorithmic tools to quickly create vision applications and solve visual inspection problems. It can be used for a variety of applications such as visual positioning, size measurement, defect detection, and information recognition. Vision Master includes both traditional algorithms and a deep learning module that can be used for tasks such as object detection, classification, or image segmentation. The software has an interactive graphical interface with intuitive and easy-to-understand function icons, simple and easy-to-use interactive logic, and drag-and-drop functionality for creating machine vision solutions.

Aurora Vision is a dataflow-based environment designed to support machine vision. Combined with its extensive library of well-optimized graphical filters, it allows the user to create both simple and very advanced traditional and deep learning algorithms. It also allows the application of logic and the handling of digital and analog signals. Aurora Vision has a set of tools for real-time presentation of results.

MV COIG, is a proprietary application designed for visual quality control using artificial intelligence, primarily in the industrial sector. The software is based on advanced deep learning algorithms and neural networks. Based on a training set of images, the model learns the product and then detects and classifies its defects and finds anomalies. One of the main advantages of the software is the advanced reporting of the system performance in the form of graphs, results matrix or text data.

Vision Camera Applications

Quality Control of Manufacturing Processes

Machine vision systems are well suited for quality control of production processes (Illustration 3). Using advanced image analysis algorithms, vision cameras detect even the smallest product defects. In the automotive industry, for example, cameras scan the surface of the car body to detect scratches, nicks, or other irregularities that could affect the aesthetics and functionality of the vehicle.

These systems quickly identify defects, eliminating the need for manual inspection, saving time and production costs, and increasing customer satisfaction by delivering only the highest quality products.

Barcode Scanning

In logistics and warehousing, machine vision systems are used to quickly and accurately scan barcodes on products and packaging. Using high-resolution cameras and advanced code-reading algorithms, goods can be identified instantly, resulting in faster and more efficient inventory management, more efficient order processing, and minimized human error.

The systems also allow goods to be tracked at every stage of the supply chain, enabling better planning and optimization of logistics processes.

Object Recognition

Using advanced machine learning algorithms, machine vision systems can recognize and classify various objects based on their characteristics. For example, in manufacturing automation, machine vision cameras are used to locate and manipulate objects in production environments.

These systems can identify items on a production line and perform specific operations on them, such as sorting, packaging, or assembly. This can automate production processes, increase productivity, and reduce operating costs.

Improving industrial safety

Machine vision systems play an important role in industrial safety by identifying and responding to potential hazards in real time. Using advanced image analysis algorithms, machine vision cameras can monitor the work environment and detect dangerous situations, such as the presence of people in hazardous areas or anomalies in machine behavior.

As a result, personnel can be quickly alerted to potential hazards and take appropriate countermeasures, reducing the risk of accidents and injuries to employees. In addition, machine vision systems can be integrated with access control and machine condition monitoring systems for even more effective workplace safety management.

Summary

Machine vision technology has evolved significantly over the past 15 years, becoming a very important, and in many cases essential, tool for quality control and production automation. Today, machine vision is used in a wide range of industries and sectors, including electronics (e.g., semiconductors), medical (e.g., medical devices), pharmaceutical, and automotive.

Machine vision provides a non-contact tool for inspecting and identifying components, accurately measuring dimensions, or guiding robots in the manufacturing process. The hardware and software solutions used in the machine vision systems described above are constantly evolving and improving.

Faster equipment, smarter tools, and more sophisticated software are making machine vision systems more prevalent in industry.