AI-supported image processing: Computer Vision

Computer vision (CV) is a branch of artificial intelligence and computer science that deals with the automatic processing and analysis of images and videos. The goal is to give AI models the ability to interpret and understand visual information similar to how the human eye does. AI-supported image processing enables companies to capture conditions and data in real time and use this data to optimize their processes. This data includes the current positions of load carriers (e.g., pallets), inventory levels, storage space fill levels, and the load status of trucks.

A computer vision camera captures a wooden pallet with an EPAL stamp, while a screen in the background displays information.
© Michael Neuhaus - Fraunhofer IML

Computer vision for logistics

Computer vision enables the recognition and tracking of goods and the measurement of load carriers using cameras, right from goods receipt to pallet management and storage. Thanks to artificial intelligence, these systems gather increasingly accurate and informative data from digital images and videos, which companies can then use to make informed decisions. For this reason, computer vision is an extremely promising digitalization approach for the logistics industry, offering increased process transparency, optimized workflows, and thus savings in terms of effort. New methods from the fields of machine learning and AI can be used to automate a wide range of processes in distribution centers and warehouses.

Our services for computer vision

Based on many years of research in the field of computer vision, we support companies with the following areas of expertise:

Process mapping and conceptualization to determine which processes can be optimized through AI-supported image processing

Assistance with hardware and software selection

Implementation of operational computer vision solutions

Development of new algorithms and computer vision solutions

Workshops and tech-deep-dives on the topic of computer vision

Computer vision for greater efficiency in intralogistics

Computer vision starts with the recognition of logistics objects. Yet its potential is enormous – the following overview shows possible applications using the example of intralogistics:

Detect, classify

What is a logistics object and what object class does it belong to?

Localization, context recognition

In which storage area and at which storage location is a load carrier located? Where in the yard is a truck located?

Counting

How many pallets are in the goods receiving and issuing areas? How full are the transfer points and areas?

Identify

Which pallet is it specifically? Which IDs are attached to the load carrier? What is the license plate number of the truck, or which ILU code can be found on the loading unit?

Tracking

Movement profile: From which starting point to which destination has an object been moved? How long has an object stayed at its current location (dwell time)?

Measuring, volume, and fill level detection

What percentage of a truck or swap body's cargo space is being used? Can a pallet be loaded any further?

Computer vision for companies: The ML Toolbox 

The importance of computer vision for logistics is reflected in research, as the topic was addressed in the large research project Silicon Economy. Here, researchers developed open source base components for AI algorithms. They then implemented two exemplary logistics use cases.

The basic components include the “ML Toolbox”, a software ecosystem that simplifies the use of various computer vision algorithms controlled by machine learning (ML). This is made possible by the provision of various services and the MLCVZoo software development kit. “MLCVZoo enables the integration of ML models from different frameworks via a uniform API, eliminating the need to integrate the frameworks individually for each new image processing project,” says Maximilian Otten, lead developer of the ML Toolbox. In addition, an intuitive AI-based do-it-yourself image processing service has been developed to facilitate the use of the ML Toolbox. “For our new Guided Training Service, companies do not need specialized IT expertise, just a basic understanding, so they can train an AI with just a few steps and minimal recruitment, depending on the complexity, in a matter of minutes,” explains Julian Hinxlage, who heads the “CV on Edge” project in which the service was developed. Overall, the ML Toolbox services cover the entire ML Ops pipeline for AI-supported image processing solutions.

Access the open source components of the ML Toolbox

Practical applications of computer vision in logistics

 

Optimizing yard logistics with computer vision

The project “Yard Lense on Edge” seeks to bridge the digital gap between the factory gate and intralogistics. Within this framework, researchers are developing a solution for automatically monitoring yard logistics.

 

Automated material requisition

Automating hospital processes: “AI4Demand” is a device that automatically requests materials from module cabinets in hospitals. AI image analysis is used to detect empty module compartments in real time and automatically trigger material requests.

 

Digitalizing rail freight inspection with AI

The research project “DIMI” is revolutionizing the process of technical vehicle inspection with a comprehensive digital solution. Camera and AI systems are used for visual inspection and detection of damage or anomalies.

Intelligent camera for computer vision

At Fraunhofer IML, researchers are working not only on software but also on hardware for computer vision. They have developed a compact and inexpensive modular AI camera. Its key feature is the fact that the devices require no infrastructure other than Wi-Fi and electricity. The captured data remains largely on the camera, which also ensures the protection of personal data. The camera components are modular, commercially available parts. This allows the device to be customized to the individual use case.

A person is holding a computer vision camera in their hand, while a laptop with programming code can be seen in the background.
© Michael Neuhaus - Fraunhofer IML

Benefits of computer vision for logistics

Fraunhofer IML is leading the way for this innovative technology across numerous current use cases. The use of computer vision in logistics offers several advantages:

Agenten Übersicht
Auftragsagent

Optimization of everyday operations

Computer vision makes it possible to optimize processes without having to interrupt them. For example, ongoing processes in the warehouse can be recorded using cameras and analyzed for optimization potential without disturbing ongoing operations.

Routingagent

Rapid implementation of ideas and use cases based on solid data

Cameras are a straightforward means of data collection. They can be used to collect large amounts of data, which can then be used to start optimizing processes.

Fahrzeugagenten

Data collection with consideration for data protection

Since the processing of the collected data can take place directly on the camera, the protection of personal data is ensured.