Computer Vision Solutions in Warehouse Automation

Until recently, the warehouses of the largest industrial companies were perceived as a kind of secondary subsidiary segment. As a result of such attitude, the warehouses were ignored, even while the implementation of corporate-wide technical modernization projects. Meanwhile, a warehouse is one of the production sites for important stages in the preparation of the final product.

Some technological processes are carried out in the warehouse, namely, the acquisition of goods, accounting of receipts, package and so on. Employers monitor the number of goods received, manually count the boxes, upload all the information to the database, move the boxes, distribute them throughout the warehouse, and also, from time to time, carry out the inventory of all products stored in the warehouse.

Although the human factor significantly affects the quality of these tasks, the main problem for the owner is the expenses of the warehouse managed exclusively by people. And only when the costs of delivery, storage, and processing of commodity flows reach half the cost price of the finished products, managers start thinking about the ways to modernize and automate their storage and logistic processes.

The good news is that today there are quite comprehensive solutions designed to automate these processes.

Large companies around the world are implementing such systems for several years already. For example, the following graph shows the companies that are involved in the modernization, and in 2017, have already adopted the largest number of warehouse automation technologies.

To assess the rapid pace of development and use of robotic systems for processes optimization, look at the table below.

Itar that the number of such systems in the world has increased from 461 to 3200, and the average number of implemented in a single warehouse increased from 3 to 25. And all these have happened just in 4 years.

The most popular and rapidly developing way to automate logistic processes is computer vision systems.

Computer vision is pure automation, that’s why it can completely exclude the human factor. Warehouse inspection with the help of computer vision is safer and much more productive compared to performing the same task by people. When the rows of boxes move along the conveyor, the person physically does not have time to inspect each of them, read the barcode and distribute them to the destination. The only way to do this is to stop the conveyor every time or get rid of it at all and move each box manually.

Computer vision in warehouse management, in turn, can cope with this task. Besides, today some systems can almost wholly replace a person.

Here are a few figures once again proving the feasibility of adopting computer vision systems.

Thinking about the need to automate storage facilities, it is worth mentioning the problems associated with this step.

The first problem arising when thinking about the computer vision in the warehouse is the prejudice of each business owner regarding the complexity and the amount of investment in this step. Indeed, complex projects often require considerable investments, but it is necessary to clearly understand that the automation of the existing volume of processes can be relatively inexpensive. Prioritizing the major issues, implementation of all the tasks can be divided into several stages. Moreover, sometimes these stages may require individual solutions or can be implemented with the help of solutions that already exist in the market, which significantly reduce costs.

Here are three tasks of the major importance which is worth considering in the first place. These are tracking, handling and transportation of goods within the warehouse.

Further, we take a closer look at how modern computer vision systems can cope with each of these tasks.

Computer vision for tracking

As we mentioned before, automating the warehouse with computer vision can solve a number of tasks which cannot be qualitatively and quickly resolved by employees. Computer vision systems are already able to cope with several tasks of pallet and boxes tracking, and much more.

Marking recognition

The primary task of the warehouse staff is to identify all the goods arriving at the warehouse every day. Manual performing of these manipulations is a time-consuming process that often does not exclude elementary human errors and, as a result, financial losses.

Today, computer vision systems can easily cope with this task thanks to automated barcode (and other types of marking) reading systems.

When a pallet is loaded with boxes arrives at the warehouse, it is captured by the camera, which in turn captures the data from the marking. The software, which is the part of the system, processes the received information, sends the data to the database, which will further help to account all the goods.

In this case, the camera serves as a scanner, and the increased speed of the received images and data processing improves the productivity and frees up the workforce for more productive use.

Counting boxes by volume

Another method by which computer vision systems can record cargo is counting boxes according to their volume. The camera captures the entire pallet with boxed, creates a 3D image reconstruction and calculates the volume of the whole load placed on the pallet. Knowing the areas and volume of one box, the system can easily calculate the number of packages send the answer to the database.

Although, this method also has problems that preclude its use in most existing warehouses.

• All boxes must be perfectly aligned on the pallet. Otherwise, the camera will not be able to define the boundaries clearly and, as a result, will not be able to detect all the boxes.
• Poor lightning creates much noise in the image and several times complicates the process of identifying the edges.
• Incorrectly exposed light can create shadows on objects and thereby mislead the computer vision system of the boundaries recognition.
• Reflections from the packaging film, which is almost always used to pack the boxes, modify the shape of the objects, and, again, make the process of recognition more difficult.
• For the system to perform its tasks adequately, the types, sizes, colors, and shapes of the boxes on the pallet must be the same.

Unfortunately, in modern realities of large warehouses, such systems can rarely be applied. However, innovations are not static, and we are sure that solutions for most of these problems will be soon found.

Staff tracking

Although the majority of warehouse automation systems are aimed at speeding up processes and minimizing the human intervention, the complete elimination of human resources remains impossible.

In this regard, the issue of security has always been and remains relevant. Security is understood as a matter of preserving a person’s life as well as the entire the entire warehouse property. Here, technologies equipped with computer vision systems can become extremely useful.

• Cameras can monitor the warehouse perimeter and record an unauthorized location of workers, non-typical behavior, etc. This type of control will help to avoid thefts at the enterprise and warn you when an authorized person is spotted on the forbidden territory.
• In any enterprise, including warehouses, there is a safety precaution that employees must follow. There is one specific example of how to automate a warehouse with computer vision: the recognition of hard hats on the employees’ heads. An ordinary camera can record what is happening and save the recording. If an accident has occurred in the workplace, the recordings are brought up and reviewed to determine whether the worker was wearing the helmet. A smart camera, which is the part of the computer vision system, having detected an employee without the helmet who is located in a potentially dangerous are, will immediately report about the increased danger either to the employee himself or to the manager. This is a top-rated solution that is all over the world and is actively recommended for use in various organizations.
• Facial recognition technology – one of the most popular types of computer vision – allows you to identify and employee and each person entering and leaving the territory. For accurate identifying, it is enough to upload to the system several photos of each employee. The system will record the time of entry and leaving, as well as the total time spent by the worker inside the territory.
• Some systems, thanks to machine learning, can accurately recognize actions performed by employees. Recognition is possible for almost all operations that can be captured by the camera, but still, it all depends on the specific task.

Computer vision for item management

In itself, computer vision systems are capable of performing far non-trivial tasks. However, the combination of CV with other latest technologies can completely change the way you imagine a warehouse with machine vision.

For example, most of the storage and production areas are equipped with manipulators facilitating the human labor and performing the monotonous work of handling of boxes hour by hour, day by day. The only problem is that each such manipulator must be individually programmed to work with a certain type of objects of the same size and weight.

This issue has already been solved, and today it is being tested in a real warehouse performing the conditions for which such systems are being developed. Equipped with a machine vision system, the manipulator will be able to perform more complex tasks, such as unload different-sized boxes from the pallet and sorting them for further movement through the warehouse.

If the tests are successful, human participation in this type of activity will be minimized over time.

Computer vision for transportation

After the unloading of the received goods is over, it is time for its transportation and distribution to the warehouse for further storage. Usually, this is done by an employee who, sitting in a loader, drives around the warehouse and delivers pallets.

The use of robots to move objects is not an innovative solution, but if you equip them with an advanced computer vision system, the robots will be able to carry out an automatic analysis of the incoming tasks and, having got access to the cargo, they will understand where to take it and which shelf to put it on. They will also be able to do the same in the reverse order. That is, if you need to get a pallet, the robot drive to the right place, the computer vision system scans the barcode from the box or RFID tag, finds the needed object and delivers it to the destination.

Conclusion

As a conclusion, we would like to once again dwell on the benefits that automation gives today:

• Reducing the costs of human labor;
• Permanent control over the availability of goods in stock;
• Increased warehouse turnover rate;
• Reduced warehouse maintenance costs;
• Preservation of goods;

And, most importantly

• Establishing business processes.

On a daily basis, we see that many of the machine vision solutions are about to be achieved, but there are also those that are still far away from perfection. Thanks to the breakthroughs, robots will be able to show unprecedented results and surpass people in most processes.

Today, the imperfection of computer vision systems is due in part to technical reasons. However, we are watching a rapid development of information technology, and more and more solutions to these technical problems are being found. Machine vision systems are becoming more popular for solving not only logistics’ issues but also the ones arising in security, medicine, etc.

UDTech helps to implement ideas and develop unique solutions for businesses and consumers. If you do not want to be a witness of how your competitors bypass you, introducing more and more technological solutions to their business processes, do the same now.

Source: UDTech IoT Blog