Frequently Asked Questions
Answers
In principle, a range of products with different surfaces are suitable. The influencing factors are material properties, surface structures, and geometries on the product. Some examples of suitable materials include metals (fine-pored and coarse), ceramics, coarse-grained plastics, natural materials such as wood or gemstone, and leather. Unstable materials and glass are out of scope. For a definitive assessment of whether your products and specific materials have sufficient unique surface characteristics, you can order a one-time pre-examination study from us. Through our tests we can directly determine the most suitable area for a recording, as we require product specific engravings or geometries for the alignment of the picture.
Currently, an image area greater than 5mm² is required as a feature for identification and verification.
No, unfortunately this is not possible because we need close-up photos of the specific surface structure. Our technology captures a specific, unique region of each individual product. Unlike e-commerce product catalogs, our technology does not capture the entire product.
Yes, we pursue a 1:1 relationship between image and surface. Via single image operation, we are aiming to differentiate and trace each individual product via the surface structure. Compared to an artificial intelligence approach, this enables us to store more information about each product and leads to greater security for the manufacturer. Instead of putting a different serial number or tag on each product, our technology takes one image per product.
With an artificial intelligence approach, a general pattern is obtained, which is compared with the training data from the database. With this, a statement can be made about the probability of finding a product class match. This is computationally intensive and new elements lead to a manipulation of the overall model performance or image variations to noise. The reliability of the statement has time-dependent shifts.
To verify the authenticity of an image of an individual surface, our technology converts the image into a pattern signature, compares it to a previously stored pattern signature, and then it is uniquely matched to a specific pre-registered surface. This eliminates the effects from data noise which in turn leads to more reliability and performance. Authentication consistency over time is achieved by adding new images of the pattern after, for example, cleaning processes or other slight changes in the pattern.
In our initial examination, together we seek to determine with you the area with the least alterations in the form of wear or soiling.
The technology is robust against light scratches, but only up to a surface specific degree. If the surface is completely worn away, the information will also become unreadable, just as with a DMC or QR code.
This can be counteracted by, for example, updating the image a few times in a products life cycle. Through such action the history and wear and tear are considered too. Additionally, higher authentication security can be reached by registering several different positions on the product.
However, if your products are not traded as second-hand or you only want to allow customer verification when they buy a new product, this may not even be necessary for your use case.
Yes, we can take images of different surface areas and link them. This can be useful if, for example, you want a public pattern and a covert pattern that you do not share with your customers.
In the beginning, each new model must be visually evaluated by Bosch. For model changes that have no impact on the previously defined surface area, no further action is required except for the creation of a new model in the backend to be displayed in the frontend. In this case a simple picture of the new model will suffice. In case of a potential impact on the surface and consequently on the recognition, Bosch checks if the current settings need adjusting. Bosch can in this case quickly begin engineering services though this effort can be circumvented by selecting a suitable and robust surface area in the preliminary when the product is added to the inventory.
No technical line integration is required. To avoid downtimes and interventions in automated customer production lines, the camera Desk Device is designed as a modular stand-alone system, which works independently after initialization and use your LAN connection to connect to the Internet. The Desk Device must capture the finished surface of the product during registration (same structure as when the product is removed from the packaging) and must therefore be the last step in the manufacturing process such as during the visual inspection or at the packaging station. With this plug and play approach, the authentication solution is horizontally scalable with an increase in your production volume. Line integrations are a possibility but require additional in advance agreed upon development efforts.
In the smartphone app, we guide the user to take an optimal picture of their product for verification. In general, proper lighting must be ensured and the product must be photographed on a solid and neutral surface. The smartphone must also be held steady to avoid blurring. Many of the most common smartphones now have very good cameras with macro function, which are very suitable for the capturing the pattern. In a B2B environment it can be advantageous to work with a fixture and a ring light for particularly small and complex surfaces. The user can also activate their phone’s flashlight through the app during the capturing process and track the image quality in real time. Large areas of product surfaces such as brand logos help to generate reliable and easy to implement solutions for a smartphone app.
The final duration depends on the specific product, as various factors have an influence, such as the population of products to be tracked, the materials, the surface condition, etc. The sole duration of image registration is aimed at less than 3 seconds with the camera desk device. We aim duration for 3 seconds for the image recognition with our smartphone app. The time will vary dependent on the specified region of interest and the amount of to be stored datapoints, as well as network quality and the country from which the request is initiated. The handling time is highly user dependent.
The registration station, Desk Device, supplied with our system is designed as a stand-alone device. For high-volume products and automated lines, if required, Bosch has the technical know-how to integrate the registration into the existing production line.
By default, a timestamp, the signature of the respective camera desk device as well as the location are saved and displayed. Additionally, brand, model name, and a model image can be easily uploaded into our backend and displayed for selection in the frontend. For further metadata, new interfaces can be created.
You do not require a blockchain for secure authentication and in purely virtual terms, the Bosch Cloud is as secure as blockchains. Since the user wants to receive product information from a reliable partner (manufacturer, Bosch), a blockchain application also does not provide a direct trust advantage.
A major problem with current blockchain technology is the lack of ability to store image data but storing this data can be beneficial for documentation in some authentication use cases. The second disadvantage of blockchains are the fees incurred registering information and attributes of each product, which, given the current development of blockchains, are not priced consistently, and can be high compared to the product cost. A blockchain application further adds complexity and an extra audit trail. One advantage, however, can be found in the context of secondhand business: end customers could not only independently retrieve information and attributes about the product, but also enter them in the blockchains.
If you, as the brand owner, wish to take legal action against the presumed infringer to claim compensation for damages, it is important that you can produce your evidence conclusively. Often, the brand owner's experts are called in to testify at trial, but since both sides have this option, this testimony can be questioned in case of doubt. Technologies can be supportive here, especially when it comes to very good AAA counterfeit products, where even the experts have difficulties to distinguish an original from a fake product.
Now the question arises of whether our technology is admissible as conclusive evidence in court. Our solution registers each new product at the manufacturer via unique surface features of the natural surface structure and transforms them into a unique pattern signature , which is assigned to the respective individual item. This means that even products from the same tooling at the manufacturer are assigned a different pattern signature. The procedure, especially the registration, is strictly limited by technology to brand owners, so the provenance of the registered products can be reliably proven. During verification, the same predefined surface location is recorded, and the generated pattern signature is compared with the stored pattern signatures for consistency. A positive verification only occurs if the matching determines an identical pattern signatures. Randomly identical pattern signatures are technically excluded. Reproducing the entire surface structure 1:1 is beyond the capabilities of counterfeiters. Thus, if you as brand owner can demonstrate you applied our solution as intended for this type of product and the questioned products are new and not completely worn off, then this should help you to prove the case conclusively. This could enable you to prove the product was not manufactured in one of your facilities.