Traditional OCR vs Artificial Intelligence based OCR Software

Every day, around 20 million containers in the world transition through thousands of ports and terminals. Containers can be used in different modes of transport, from ship to rail and truck, without unloading and reloading their cargo.  

Due to its high value and the amount of assets passing through ports throughout the day, OCR (Optical Character Recognition) is an indispensable technology in ports, to accurately track alphanumeric codes on containers and their subsequent redirection. 

What is OCR? 

OCR is a technology focused on the digitisation of texts, applicable in different fields and sectors, which have the need to automatically identify symbols or characters in an image and then store them as data

To recognise the characters, the software inspects the image pixel by pixel, looking for shapes that match the character features. 

Ports have traditionally used the traditional OCR system, but with the passage of time and the increase in the transport of intermodal assets, this technology is becoming obsolete and does not provide the ideal results for proper traceability in ports.

Traditional OCR 

Traditionally, OCR has been a key automation technology in port facilities. Ports and terminals have invested millions of euros in installing arches equipped with cameras and sensors to trigger image capture once a vehicle passes through, relying on legacy OCR solutions for textual interpretation of these images.  

However, they do not live up to expectations. As a result, processes remain inefficient in the identification, monitoring and control phases. Moreover, these solutions are expensive to implement due to their high dependency on hardware and infrastructure. The financial investment required to implement these systems is very high, as well as the installation times, which cause disruptions to the ports’ activity due to the significant infrastructure required.

What is more: despite the significant trend towards digitalisation and robotisation, ports are moving more slowly than other sectors in this process. In fact, 97% of container port terminals are not automated, only 1% are fully automated and 2% are considered semi-automated.

How does it work?

To identify characters, the software examines the image pixel by pixel, looking for shapes that match the characteristics of the character. Depending on how complex or developed the software is, it will search for matches with the characters and fonts available in the program, or it will try to identify characters by analysing their characteristics so that their recognition is not limited to a certain number of fonts.

Although OCR can currently recognise characters, the accuracy of existing solutions does not yet reach a maximum accuracy when reading IDs of containers, wagons and moving number plates.

What elements does a traditional OCR require?

Metal door and rail infrastructure

To ensure that vehicles and assets always arrive in a predetermined position and a specific direction, physical gates must be built and installed. Only assets passing through these gates will be tracked.

Presence sensors and system triggers

OCR systems do not become operational until the presence of an asset is detected. These presence triggers are generally implemented through the use of photocell sensors, inductive sensors, weight sensors or radar. 

Lighting panels

Lighting conditions are controlled at the OCR doors by LED panels, strobe controllers and infrared lighting. 

Measuring systems

Laser measurement systems are used to detect the exact 3D position of different assets in order to guide the system “where” and “when” to read. 

Image capture

A configuration usually consisting of 4 to 8 cameras is activated synchronously to record various fields of view of the equipment (e.g. front, back, sides, top of a container). Area scan and line scan cameras are used to obtain different views of the equipment. Specific, high-end cameras are needed for each type of reading: Plates, Containers, Wagons, among others. 

Central CPU Unit

Computer hardware that controls the entire system and executes character recognition. It needs power and connection to the port’s local area network. 

OCR software

Based on a first stage of isolated character detection (with connected components or MSER techniques) and a subsequent letter recognition based on shape features and statistical classifiers such as decision trees or SVM. Such solutions are obsolete compared to current Deep Learning recognition systems. 

Data integration and communication with other systems

The recognised data and metadata are put into a structured format and integrated with the terminal operating system (TOS). 

Container leído por OCR

There are many elements required for OCR, and yet there are many disadvantages for ports and terminals. Approximately 3% to 5% of the IDs have to be corrected manually, as they are read character by character and not as a whole, which usually leads to reading errors.

Because of these drawbacks, ports lacking state-of-the-art automatic identification and data capture technologies ultimately suffer higher operating costs and competitive disadvantage.  

Recent research shows that annual bottlenecks represent costs of €4 billion in the EU transport network. Traditional OCRs, among other factors, contribute to this impact because trucks have to stop at the arches for the OCR to read the data. Causing a negative effect on productivity and customer satisfaction.

Artificial intelligence software

To meet these challenges, the Intelligent Identification Software for Intermodal transport was developed. A total disruption of the OCR paradigm, it enables the ubiquity of traceability in port facilities by breaking down the financial and technological barriers to its adoption. 

This technology generally offers the following features: 

  • Reading in complex environments, in dark, blurred or high-speed motion. 
  • Several readings at the same time, allowing the identification of different codes in the same shot. 
  • Reduction of hardware and maintenance costs, dispensing with large infrastructures. 
  • High percentages of accuracy, exceeding the standard of existing solutions, close to 100%.
  • Automatic integration of data into the port’s pre-existing information management system.
  • Faster ROI, making this alternative an optimal solution for all types of ports.
Ejemplo de casos de uso frecuentes para un OCR portuario de inteligencia artificial.

With these features, the software has the potential to democratise intermodal cargo traceability solutions for ports, terminals and intermodal platforms of any size.

Equipping artificial intelligence with technologies such as Deep Learning and Computer Vision is what allows the software to be trained with each reading it performs. This makes it possible to make readings in a general way and not character by character, which is more accurate than a traditional OCR.

Its use is usually aimed at access control, real-time monitoring of internal vehicle traffic flows or traceability of goods throughout the supply chain, although the uses of OCR based on Artificial Intelligence are very diverse.

Its unique neural network architecture makes asset recognition resistant to dust, dirt, movement, motion, blur, partial occlusion, damage or rotation, which are common identification problems with assets in port environments.   

      Nueva llamada a la acción

If you want to keep up to date with the latest news from about technology in port logistics and utilities, subscribe to our newsletter. You will regularly receive the most important content and news from our blog in your mailbox. 

Follow us also on our social profiles in LinkedIn, Twitter, YouTube, Instagram and Facebook

Categories:

Comments are closed