Artificial vision

La artificial vision or computer vision It is a technique that can be used for a multitude of applications outside and within the industry. It allows understanding images, processing information, analyzing and producing a series of actions based on said data. And they can do it in a more efficient way than a human, since you give machines a great capacity to understand and interpret the images of the environment they are observing.

With the advance of AI (Artificial Intelligence), it has been possible to improve a lot these artificial vision techniques to achieve things that were unthinkable up to now. In addition, artificial vision techniques can be applied in-situ at the same time, or analyze images or videos already recorded. There is also a 3D aspect of this type of vision that provides new capabilities to emulate human vision by computer.

What is computer vision?

La artificial vision is a compendium of tools and methods to obtain, process and analyze images of the real world through computing. In this way, certain tasks can be treated and automated, from image corrections and restoration, to decision-making for other industrial applications such as those that will be studied later.

It is precisely the industrial sector the one that benefits the most from this artificial vision, since it allows the manufacturing or selection processes to be automated and to take much higher speeds than if they were done by a human. In addition, with the improvement of techniques, it has more and more applications and low costs, which allows its expansion from the automotive industry, to electronics, agriculture and even logistics.

Technique

What is done during the process it is basically having a camera / s or sensor to capture images of objects or the environment, process them quickly using software that runs on a computer, extract relevant information from those images and be able to apply them in some way. For example, objects that pass in front of a camera on a conveyor belt can be analyzed to detect those that are damaged and have a mechanical actuator discard them so that they do not continue in the chain.

All artificial vision system goes through the following steps:

• Catchment: The sensor will capture the image of the real object. That is through an optical sensor, CCD camera, CMOS, INGAAS, X-rays, IR, thermography, etc. This also has some associated accessories, such as lighting. In that case they can be fluorescent, LED, polarized light, laser, Backlight, etc.
• Digitization: converts the information captured by the collected images into digital format in order to be processed by the computer.
• Computers and Tablets: thanks to the control software it allows to process this information and obtain data on which it will act / decide in later stages.
• Results: results are obtained and acted upon.

All those steps need various modules or parts to be able to function as:

• Image module: the person in charge of capturing the signal or image of the object or environment.
• Digitizer module: the one that converts the camera's analog signal into digital.
• Display module: it should not be confused with the first one, it is the one that converts the digital signal resident in a buffer into a visual signal to be shown through a monitor or screen if monitoring is necessary.
• Image processor: can be software or hardware. Regardless of its implementation, it is responsible for interpreting the digitized images captured by the camera. Of course, in any case you need a computer.
• I / O modules: input and output manage image capture and control output based on the data obtained.
• Communication: it is the bus or interface through which the artificial vision system can communicate with the rest of the elements. They can be wireless, Ethernet, RS232, ...

Purpose or functionality

If a few objects pass through that conveyor every minute, a human could do it efficiently. But if tens, hundreds or thousands of them are happening, it becomes very complicated or impossible. This is where computer vision can speed up these processes and carry them out.

Therefore, computer vision is a great tool to accelerate industrial production. All thanks to processes based on solutions that are adapted to the production processes of each industry. With capacity for scalability, updating and customization if necessary.

For this, a multitude of devices can be used, from simple optical sensors, to a more advanced camera or a group of them to achieve 3D.

Advantages and disadvantages

In addition to the above, there are also a number of advantages and disadvantages of artificial vision systems. The most notable is the improvement in production performance in the industry, but there is more.

Between the advantages can be highlighted:

• Eliminate the subjectivity of inspection: By implementing artificial vision systems, an improvement in this regard can be achieved and performance improved when quantifying and evaluating parameters per unit of time.
• Flexibility: the systems themselves allow adaptation and scaling to be better adapted to production processes if they have been changed. This saves a lot of time and allows a quick start after every change, without having to train staff for the change or anything like that. Just a simple setup.
• Affordable: Although they are not cheap items for most of the pockets of individuals, but it saves a lot of money in the long run for the company. Furthermore, this technology is mature enough and understandable enough to be getting cheaper and cheaper. Computers, software or opto-electronic components are increasingly cheaper and more efficient.
• Costs: these artificial vision systems reduce costs in many ways, such as costs for returning orders, personnel replaced by these systems, temporary costs, increased production (higher profits), etc.
• Metrology: Allows you to extremely quickly measure or obtain information on the physical magnitudes that appear in captured images. For example, you could determine in a fraction of a second the dimension of a part, its area, the distance between parts, diameters, angles, position, etc. Something that a human cannot do so quickly.
• Classification: Thanks to the previous advantage, there is another one, such as fast and efficient industrial classification. This allows you to classify and automate tasks at breakneck speeds based on those dimensions, patterns, barcodes, color, area, shapes, etc.
• Best end product: artificial vision also has a great advantage that can have an impact on the end customer, and that is the improvement of the quality of the parts. By being able to be analyzed in a more efficient way, even in areas inaccessible to a human, it allows to produce higher quality parts. That translates into a more satisfied consumer and customer loyalty.
• Other: it also needs less attention, it is not as susceptible to visual errors as humans (inattention, carelessness, distractions, ...), it is not affected by absenteeism from work, it improves verification in places inaccessible to the human eye (eg: by rays X to see inside parts).

Between the disadvantagesThe most notable is the price of these systems, since it has practically no weak points. Only in some cases where a somewhat less objective and more subjective evaluation is needed can it fail, since in those cases there is nothing better than the people themselves to be able to evaluate each case.

Applications in the machine vision industry

The applications of artificial vision in the industry go through three very specific fields such as process control and quality control, although some companies go further and are using it for other non-industrial applications.

Practical examples They range from temperature control, traffic control, verification of correct assembly, labeling and marking, inspection of welds, quality control of objects, selection and filtering, tooling control, control of surface finishes, pick-up & place systems for guide industrial robots, detection of foreign bodies in containers, etc.

Practical examples in the industrial field

The applications of artificial vision in the industrial sector is quite wide, as you can see. The range of applications in different sectors they go through:

• Electronics: In the electronics industry, artificial vision can be used in various manufacturing processes, such as the handling and identification of components, quality control, checking the correct welding and packaging of parts, for pick-up & place processes to place components in PCBs and solder them, etc.
• Automobile- Used for inspection in the manufacturing and assembly process of vehicle parts. As in the process of stamping, machining, welding, painting, burrs, extrusion, etc.
• Food: artificial vision in this industry allows to improve quality control. For example to see if the containers have been filled properly or that they do not have foreign bodies inside. They are also widely used to remove damaged or rotten fruits, remove branches, stones, peels and other elements that should not go to subsequent processes, categorize by size, etc.
• Packaging and packaging: In industry logistics and packaging, computer vision can inspect the presence or absence of certain markers. You can also catalog by barcodes or labels, inspect batches, expiration dates, correctly place caps, etc.
• Logistics and identification: allows you to quickly identify parts or products. It adapts very well to the needs of department stores and distributors, such as in Amazon logistics centers.

Machine vision and industry 4.0

La artificial vision, like many other digitalization and transition technologies towards the modernization of companies, such as Big Data, AI, IoT, and the cloud itself, fog and edge computing, has a crucial role in the so-called 4.0 industry.

All these paradigms together allow to improve all the conditions of this new emerging industry that aims to revolutionize the sector. And it is that after the industrial revolution with the introduction of machines (1.0), the introduction of electricity in the sector (2.0), the arrival of computing (3.0), now comes this new revolution thanks to these new techniques to give instead to version 4.0.

Machine vision, in fact, can group several of those improvements into one. Since it uses software and hardware to function, and may also include AI to give it greater intelligence and recognition capabilities. All this endows the industry with the great advantages and precision mentioned above.

But if this capacity is combined with other measures to improve and modernize other areas of the company that implant it, it can lead to an industry 4.0 with comprehensive solutions a lot. more efficient and competitive.

Companies like IBM, Red Hat, Marval, Telefonica, and many others have been trying for some time to help companies with this transformation so that they can achieve their objectives. In Spain, many important companies such as Santander, Cepsa, and many others have already begun to enjoy the great improvements of 4.0.

It is precisely Marval the company that has been developing artificial vision systems for the industry for more than 20 years and evolving its tools. Thanks to these projects and those of other competing companies, all the tools available to the industry have been improved.

For example, imagine a comprehensive system of Industry 4.0 in a factory where an artificial vision system can select the amount of raw material or valid parts needed. Based on this information, not only the unsuitable could be discarded and only the suitable ones could pass to the production chain.

With 4.0, this information could also be transferred to the cloud and use other emerging technologies for, for example, ordering from the supplier for parts or raw materials based on production capacity and the quantity of discarded parts, evaluating the inventory fully automatically. Or perhaps you report the detected failures to that supplier so that they can reduce those defects in future orders.

That is, in industry 4.0 technologies cover it all, from the first process to the last, and in all departments and sectors of the company.

Beyond traceability

Machine vision systems in an industry 4.0 can go beyond the measures of traceability (morphological analysis, defects, placeholders, color analysis, appearance, foreign objects, quality, code reading, etc.). It could also use that information OCR, OCV, or data obtained after processing to make other machines or processes in the factory ready or with more information about it.

For example, imagine that objects of varying consistency are produced. An artificial vision system could determine tolerance levelFor example, the consistency of each object through different systems and thus mark it so that a machine that has to stamp it in a subsequent process exerts the appropriate pressure depending on its consistency.

This is simply possible by understanding the artificial vision system as an IoT element connected and the next process machine as another connected IoT device. Therefore, they can communicate through the network and even between one and the other, other elements of the fog or the cloud can be used to analyze certain data.

Machine vision and industrial digitization

The new industrial digitization systems, emerging tools and computer vision will play a key role in the present and the immediate future in companies of all sizes. For example in combination with MES / MON systems (Manufacturing Execution System / Manufacturing Operation Management).

That is, the MES systems they are information management systems connected to industrial equipment and manufacturing lines. With them you can monitor and control the processes, the data flow of the plant and all in real time using ERP software. This is how the transformations from raw materials to the final product are tracked and documented.

MOS it is a methodology that allows to visualize the manufacturing processes from start to finish to optimize efficiency. That ensures efficient manufacturing execution and improves productivity.

Therefore, artificial vision systems have a crucial role in these cases, since they are a perfect complementary tool to draw up these industrial digitization strategies. Especially together with PLM (Product Lifecycle Management) developments, that is, software systems to manage the life cycle of products from their manufacture to their disposal, also passing through their commissioning.

As you can understand, for all this you need a lot of information stored in large databases in the cloud or locally, and that can be processed quickly and efficiently for analysis by Big Data. And those databases will be fed by those artificial vision systems that are the ones that can obtain quick information on all products.

All that without altering the TTM (Time To Market), on the contrary, you can get all that information and improve that parameter considerably. That is, if the time from when a product begins to be conceived until it is put on the market is longer.