SUCCESS STORY:
Access control, quick and easy: containers and license plates reading
«AllRead and TDT created a new slim and powerful integration, focusing on the results. Optimisation of back-office processes and cost reductions were achieved, as well as improved coverage of the vehicle monitoring system within the port facilities.»
The Challenge
Terminal Darsena Toscana (TDT) is the main container terminal within the Porto of Livorno, and serves the markets of central and north-eastern Italy, a key entry point to the European market and an important point also for the American and West African markets, with an annual operating capacity of 900,000 TEUs.
In recent years, TDT has adopted an important role for the transhipment of goods from North and South America to the Eastern Mediterranean and viceversa.
The Problem: containers exit control
TDT needed to implement a flexible solution to optimise the control network and obtain traceability of goods by road, from two of its exit lanes, and thus automatically monitor traffic.
In this use case, AllRead accompanied the terminal throughout the ful process of installing the two cameras required for the project, as well as the preparation of the server and the posterior integration of the data, without the need to work on large hardware installations or incur on large financial investments.
TDT’s requirements were:
- Detection and reading of European license plates.
- Detection and reading of BIC container codes.
- Association of containers with the corresponding license plate.
- Conservation of images as a proof of condition.
- Limited impact on infrastructure and budget.
- Quick implementation and integration with existing systems.
Prior to the integration of AllRead project with the TDT’s TOS (Terminal Operation System) called JTIS, the Terminal carried out these tasks manually. TDT’s team used rugged tablets to check out the containers from the Terminal. This manual process was responsible for 5% of errors in data collection.
The Solution: AllRead’s plug and play software
AllRead’s deep learning-based software has minimal technical requirements to track assets with the highest accuracy. AllRead simultaneously detects and reads several codes and symbols from the same video stream. Starting with a single IP camera, connected to the existing server of the terminal, AllRead operates to deliver real time and accurate results.
Two cameras were installed on each exit. The first one was installed at the front to read the license plates, and the second camera was installed at the back, to read both lateral and rear sides of each container.
Thus, a fisheye lens was used to read containers codes. This lens was essential to ensure both lateral and rear container sides. A fisheye lens was used to read containers codes. This lens was essential to ensure simultaneously both lateral and rear container sides. Thus, no additional camera was necessary, even in cases where two containers are loaded on the same trailer. Besides, when a code is too damaged on a side, providing a second opportunity to read the rear code is mandatory.
IP cameras were connected through fiber to the existing local server, where AllRead’s software was installed. Only 12 CPUs and 12 GB RAM were allocated in each lane, for AllRead to perform. TDT could monitor the CPU Load to ensure stability of the Server.
The generated data – license plates, asociated containers, time stamps and confidence levels – were easily integrated with TOS. Additionally, a standalone web interface provided real-time access to the data for control, monitoring and search queries.
The project started with one lane, and then, rolled out to the second one, using all configuration elements to clone the implementation in a record time of two weeks.
During the Setup, more than 10.000 trucks were identified, read and correctly delivered to ensure the stability of the configuration and the quality of the data output.
Key success factors: AllRead’s innovative Technology
AllRead´s technology is the result of 5 years of research in the Computer Vision Center of Catalonia. Its unique deep learning recognition models are not only highly accurate, even in unconstraint environments, but are also lightweight, with low needs for processing power.
The success of the results was due to several key factors:
- AllRead’s software is hardware agnostic, working perfectly with different brands of ordinary cameras, adapting to the clients existing infrastructure.
- The software detects and read several objects simultaneously with the same image capture, reducing the need for multiple cameras (and associated budget).
- AllRead used the existing servers of the client, given its low needs for processing power.
- AllRead works with standard protocols and formats for data transfer, making integration with existing information management systems easier.
The Results: accurate readings and fast process automation
AllRead reached TDT’s expectations of accurate control of its traffic through a cost efficient, frictionless and fast implementation.
With only two cameras, it was possible to perform:
- 99 % success rate in associating the vehicle with the load.
- 97 % in container reading accuracy.
- 99,5 % in the reading of license plates.
«Most importantly, based on the situation prior to AllRead, the number of data collection errors has been drastically reduced».
Conclusion
The key point of this project was the integration by JSON messages with TDT TOS (not using a database) so Gate Out OCR became a new module of the TOS.The software module has therefore been enriched with real-time
alarm and monitoring functions managed by the TOS of the Terminal.
AllRead and TDT created a new slim and powerful integration, focusing on the results. Both teams worked hardly together, achieving excellent results, and being aligned on what each of us must do was key to this project’s success.
As a result, an optimization of the back-office processes and a reduction of costs were achieved, as well as better coverage of the vehicle control system within the port.
If you want to know more about this success story, contact our team.