Computer Vision

Trend Overview

Impact:

Moderate

Realization:

< 5 Years

Focus Areas:

Automation & Efficiency Health, Safety & Future of Work Physical & Digital Security Visibility & Resilience

Trend Clusters:

Artificial Intelligence Robotics

Sectors:

The trend of Computer Vision utilizes cameras to capture photos or videos and applies artificial intelligence (AI) algorithms to analyze data extracted from this digital imagery. Rudimentary visual AI systems are trained to simply differentiate objects from each other, while more advanced versions can track objects across viewpoints and learn on their own.

Emerging from AI as a trend in its own right, Computer Vision has developed in conjunction with the advancement of deep machine learning, leveraging the rising quality and decreasing cost of camera devices. In 2020, the computer vision market globally was worth 9.4 billion USD and – as AI, vision systems, and computer processing continually improve – it is anticipated this market will more than quadruple to 41.4 billion USD in 2030. Today, advanced computer vision technology is perfecting depth perception, 3D reconstruction, and dark and blurred image interpretation, all of which will unlock more opportunities in supply chains.

The Computer Vision trend is set to highly impact logistics in the coming years. Its technology will underpin and drive future logistics, enabling more efficient processes as well as sustainable and safe operations. However, more investment is needed for this trend to be fully realized. As experienced in the early days of sensor adoption, computer vision applications must be scalable for logistics organizations to maximize benefits.

Computer Vision can process single or multiple object detection and classify a number of elements from a single image.

Source: Allied Market Research

Relevance to the Future of Logistics

Order & Shipment Processing

Many tasks in the warehouse involve humans to visually assess or confirm elements of a product, parcel, or pallet as it passes through the facility. As computer vision technology advances, we here at DHL recognize such tasks can be gradually delegated to computer vision-enabled AI systems.

One practical use case is dimensioning a shipment, or measuring its area or volume. This can be difficult when shipments are large or oddly shaped, or they are on fast-moving conveyor belts. Companies like German-based Metrilus have recently developed low-cost solutions that automate this dimensioning process, capturing measurements in milliseconds and sending this data to the warehouse management system. Accurate measurement and volume detection enables best fitting packaging selection to avoid shipping air, thereby reducing waste and keeping sustainability at the forefront of logistics.

Another use case is object identification for picking and packing. Currently, most orders are picked by hand, and those that utilize computer vision in robotics solutions tend to identify a product by its barcode or QR code. But developments in computer vision have enabled robots to identify tens of thousands of products with high accuracy, regardless of the presence of identifier codes.

With this level of visual AI, logistics providers can optimize processes, reducing cost while increasing throughput.

Next use case

Asset Tracking

Millions of dollars are lost every year across sectors due to tools, equipment, and other assets going missing, and many hours are spent by workers searching for them. Computer vision technology can provide useful solutions to help track such assets and save time and money.

When a worker picks up a wrench and walks around a facility with it, a computer that processes connected visual feeds can locate where the wrench was last put down, while also tracking dozens of other tools at the same time. Vehicles like forklifts in a warehouse or trucks in a yard can also be tracked; this data can be used to determine if vehicle movements are performed in an optimal way. Furthermore, visual AI software can follow pallets of goods in inventory sections and keep accurate records of shelf vacancies and how long a pallet has remained on a shelf, in addition to tracking any falling or shifting parcels in the back of courier vans.

In implementing computer vision solutions in their facilities, logistics providers can avoid wasting time searching for misplaced assets and manually confirming asset locations. They can also use the analyses from visual AI to further optimize operations and accurately determine inventory stock levels.

Previous use case Next use case

Safety in the Workplace

Maintaining and improving safe working conditions is a top priority for the logistics industry. The COVID-19 pandemic increased attention to this and visual AI technology will be relevant and helpful to workplace safety in the years ahead.

During the peak of the pandemic, computer vision was used to ensure workers adhered to personal protective equipment (PPE) regulations. At DHL, we now see this technology expanding into other workplace safety use cases. Cameras and the AI behind them may one day detect if employees are utilizing ergonomic best practices to minimize injury risk, identify lone-worker emergency situations in less-trafficked areas, ensure vehicles are complying with local speed limits, and even determine if predefined walking pathways are being respected in a facility.

To ensure personal and data privacy, advanced computer vision solutions can blur out faces or separate personal identifying factors from analyses. This is helpful as it demonstrates to all parties that the goal of computer vision is not finding fault with individual workers but ensuring their safety and optimizing the workplace experience by assessing workflows, mapping out areas of high incidences, and triggering process change.

Previous use case Next use case

Trouble-Free Operations

With today’s global supply chain networks strained and under immense pressure, logistics operations must remain functional and unhampered by incidents and broken-down equipment. Computer vision can help with this.

Digital side mirrors like those designed by Israeli startup Brodmann17 can highlight nearby vehicles, especially those in blind spots, to truck drivers, enabling safer lane changes and turns while also reducing incident risk. Meanwhile, visual AI via cameras, whether on gates or drones, can detect irregularities in various objects like airplane wings, shipping containers, and warehouse rooftops, initiating predictive maintenance procedures to assess any potential damage or need for repair. Computer vision can also be applied to repair or emergency stocks, keeping inventory count, flagging items like lubricants and spare wheels that are running low, and triggering timely replenishment.

With computer vision as a preventative tool, logistics organizations can better protect supply chains from avoidable delays.

Previous use case

Challenges

Employees may be highly suspicious and resistant to computer vision solutions, believing they are being constantly watched and their personal data is at risk.

While approximately 90% of internet protocol (IP) cameras can be easily integrated and configured into new software platforms, older camera technology must be upgraded to scale computer vision solutions.

Unless cybersecurity measures are considered and strengthened, there is a risk of hacking and malicious manipulation of data, resulting in skewed analyses and altered AI performance.

Compliance with GDPR and other data privacy and protection laws must be considered and may limit the scaling of a computer vision solution across regions.

Challenge 1 Challenge 2 Challenge 3 Challenge 4

This trend should be ACTIVELY monitored, with imminent developments and applications.

Outlook

The trend of Computer Vision has use cases in practically every segment along a supply chain. The technology can be a helpful to logistics organizations in optimizing operations, improving worker health and safety, and reducing cost. Highly valuing data privacy and security from the beginning of visual AI implementations may relieve initial employee resistance to the technology. As today’s pilot projects pass their trials, rapid implementation is anticipated in the next few years.