Empty bottle? Not with Deep OCR
Food & Beverages | HALCON | OCR & OCVTo identify products quickly and reliably, even in difficult circumstances, machine vision has proven its worth. The company Visione Artificiale also relies on machine vision for its system to trace aluminum bottles during production. Within this system, they utilize the standard machine vision software MVTec HALCON from Munich-based MVTec Software GmbH.
Visione Artificiale SRL, based in Bione, Italy, specializes in the integration of machine vision into fully automated robotic systems. The company's portfolio includes systems for the quality control of components, 3D vision systems, bin-picking applications, and deep-learning-based applications.
Visione Artificiale has developed an application for a company in the food industry that can be used to automate the tracking of CO₂-filled aluminum bottles that are used to fizz up still water. Laser-engraved information on the cylindrical bottles, such as serial numbers or filling dates, is used to automatically identify the aluminum bottles using optical character recognition (OCR). The engraved information is robustly and quickly checked for accuracy using machine vision. As the inspection is automated and takes place around the clock, the company can save costs in the long term.
Automated and accelerated inspection process
As part of the application, a line scan camera scans the rotating bottle and captures a two-dimensional image of the curved surface. The first step is to identify the areas on the image that contain letters and numbers. The network then determines the characters and checks the contained information for correctness. This automates and accelerates the entire inspection process. As the setup is equipped with two cameras and rotating devices, two bottles can be tested simultaneously per cycle for greater efficiency.
Clear identification thanks to Deep OCR
Due to the surface characteristics of the aluminum and as a result of the exposure, a variety of reflections and specks may occur during image acquisition. This makes it difficult to correctly segment the characters during image processing and interferes with OCR-based identification. To ensure robust recognition rates nonetheless, Visione Artificiale relies on the Deep OCR technology integrated in HALCON. The technology uses deep learning algorithms and can localize characters regardless of their orientation, font, and polarity. The automatic grouping of letters also makes it possible to identify entire words. Misinterpretations of similar-looking characters are completely avoided, which significantly increases recognition performance. HALCON's Deep OCR has been trained to reliably identify a wide variety of fonts.
"Due to the special nature of the bottle material, a conventional OCR system would not have been able to identify the engraved texts. To achieve robust recognition rates despite the reflections, we needed an intelligent OCR system that can rise to this challenge. Deep OCR proved to be the optimal solution for our requirements. Thanks to comprehensively pre-trained deep learning networks, even difficult-to-read texts can be recognized with high accuracy. MVTec's HALCON libraries offer an impressive range of deep learning algorithms that enable us to successfully accomplish this complex task," confirms Fazio Saverio, founder and owner of Visione Artificiale. Saverio and his team were supported by iMAGE S. This MVTec Sales Partner assists its customers in all aspects of machine vision and provides its own products and technologies for this purpose.
Machine vision for higher quality and productivity
The use of the HALCON machine vision software including Deep OCR made it possible to trace the CO2 bottles using serial numbers in the first place. The process can only be implemented quickly and economically through automated inspection and verification. In addition, the employees who previously carried out the process manually are relieved of this monotonous task and are available for more demanding tasks. Thanks to optimized traceability, the productivity of the entire process chain has been increased and the quality of the products has been significantly improved.