Automation specialist Prime Vision has developed an object recognition solution for postal organizations that can not only read address labels, but also decipher stamps, franking marks, logos and other images printed onto letters.
The company’s research and development team has developed and refined the solution, so that the automated recognition of printed information on envelopes is faster and more accurate than ever. For one customer, Prime Vision achieved a 99% read rate and almost zero error rate based on a test deck of Dutch stamps.
Not only does the new solution reliably automate a manual process, it also provides revenue protection, as under-paid mail can be detected and customers can then be charged for the difference in price. As with all Prime Vision developments, the object recognition technology is supplied as a module that can be integrated into already existing systems.
The object recognition tool works by constantly comparing information. The speed of comparison was significantly improved by applying innovative search techniques developed by Prime Vision’s research team. As a result, more target objects can be stored in the database with even more image descriptors, which improves the applicability, increases the detection rate and reduces the error probability.
The validation of the detection result has been expanded by comparing the object with a reference image over the full image area. This ensures that even very similar objects can be distinguished with high certainty. Additional optical character recognition (OCR) techniques also read text in the objects for the purpose of validation or data gathering.
Customers maintain control over the system as they can manage their own image database so that all new objects can be added at short notice, via a configuration tool. Currently there are 850 stamps logged in the test system.