A dedicated inspection neural network is trained for each decor. This enables Edge InspectiON to learn the typical appearance and allowable structures of the respective decor and precisely detect deviations.

This decor-specific approach requires a higher initial setup effort compared to purely geometric 3D inspection. However, it forms the foundation for a vastly more comprehensive quality assessment: Edge InspectiON evaluates not only shape or height deviations, but the actual visible appearance of the furniture edge.

As a result, defects such as chip-outs, open glue joints, overhangs, underhangs, pressure marks, or damages can be detected alongside surface or decor foil imperfections—even when they do not cause a relevant geometric change in the component.

Especially across varying colors, decors, and surfaces, this AI-based evaluation delivers reliable quality control tailored specifically to the given application.

In edge inspection, the key factor is not the measurement principle itself, but which relevant defects are reliably detected.

3D systems capture the geometry or height profile of the edge. Defects such as chip-outs or overhangs can be detected if they produce a corresponding geometric change.

Edge InspectiON, on the other hand, directly evaluates the visible appearance of the edge. This makes it possible to detect both typical edge defects—such as chip-outs, open glue joints, or overhangs—and defects that do not cause any relevant geometric change, such as damage to the surface or decor foil.

The additional AI-based evaluation allows the system to distinguish complex decors and permissible surface textures from actual defects, thereby reducing false alarms.

For new decors, images are captured and processed through an automated training procedure. Only defect-free parts are used for training. During the production process, defects are then detected as deviations from this trained nominal state. Typically, a dataset of approximately 20 to 30 images is sufficient.

For structured or patterned decors, 40 meters of defect-free edge are required as a training dataset.

This data does not need to be produced separately. Edge InspectiON automatically captures the images during normal production and saves them under the respective decor. Once a sufficient edge length has been processed, training can be initiated.

For solid-colored or highly uniform decors, significantly less image data is typically required, as the surface structure is much simpler.

As a result, the training dataset is generated effortlessly alongside the active production process.

Does the teach-in process work with batch size 1?

Yes. The required image data does not need to be captured within a single, continuous production batch.

Edge InspectiON automatically stores the captured edge images sorted by decor. The required edge length can therefore be collected across multiple orders and production runs.

This makes the system ideal for high-mix production environments and batch size 1 operations.

Yes. The required image data does not need to be captured within a single, continuous production batch.

Edge InspectiON automatically stores the captured edge images sorted by decor. The required edge length can therefore be collected across multiple orders and production runs.

This makes the system ideal for high-mix production environments and batch size 1 operations.

Once sufficient image data is available, the training simply needs to be initiated once in the software. The actual training takes about 12 hours and runs automatically in the background.

Alternatively, HECHT can handle the training as a service. In this case, the new network model is typically available within less than 48 hours after transferring the required image data.

There is no need to program inspection parameters or manually teach individual defect types.

No. Training a new decor does not require any defect images.

Edge InspectiON uses defect-free edges to learn the baseline appearance of each decor. Any deviations from this standard are then detected and evaluated against your defined quality limits.

Consequently, a database of previously known defects is not required.

Alternatively, HECHT can handle the training as a service. In this case, the new network model is typically available within less than 48 hours after transferring the required image data.

There is no need to program inspection parameters or manually teach individual defect types.

No. The appropriate inspection method depends on the specific decor.

For structured, patterned, or visually demanding decors, AI-based vision systems offer distinct advantages because they learn the typical structure of the decor and distinguish it from actual defects.

For solid-colored or highly uniform edges, conventional vision technology already delivers excellent results. In these cases, AI training is not required.

HECHT does not strictly rely on a single technology, but rather applies the method that makes the most sense for each specific inspection task.

Conventional machine vision is exceptionally well suited for clearly defined and uniform features. For solid-colored decors or simple structures, it delivers highly reliable inspection results.

However, with patterned and textured decors, traditional rule-based inspection quickly becomes complex because acceptable decor structures must be differentiated from actual defects.

This is where AI-based inspection offers a decisive advantage: it learns the baseline appearance of the decor, enabling it to evaluate complex structures far more effectively.

Edge InspectiON therefore combines conventional machine vision and AI, deploying each technology where its strengths lie best.

Edge InspectiON evaluates not just individual geometric features, but the actual visible appearance of the edge.

For textured decors, the system must learn which color gradients and patterns belong to the normal surface. Through this teach-in process, Edge InspectiON reliably distinguishes intentional decor structures from actual quality deviations.

For simple or solid-colored decors, this setup effort is reduced accordingly—or can be omitted entirely by using conventional machine vision.

Edge InspectiON operates at an image resolution of over 200 DPI (approx. 8 pixels per millimeter), allowing it to capture extremely small anomalies down to a technical minimum defect size of 0.016 mm².

However, for industrial applications, the key factor is not the theoretical minimum defect size, but identifying which deviations are genuinely relevant to product quality.

A practical baseline for production is approximately 1 mm². This benchmark is based on operational data from more than 20 installed HECHT 4i inline scanners using the same core AI technology across leading manufacturers in the kitchen, office, and home furniture industries.

The allowable defect threshold can be configured individually for each decor to match customer-specific quality requirements.

Yes. Quality thresholds can be set according to the specific requirements of the product and decor.

For example, you can define the exact size threshold at which a detected anomaly is classified as an actual defect.

This ensures that the inspection system does not generically decide what is "good" or "bad." The customer defines their own quality standard—Edge InspectiON enforces it reproducibly.

Edge images are captured during normal production and automatically saved by decor. For textured decors, sufficient data is typically available after processing roughly 40 meters of defect-free edge. For simple, solid-colored decors, significantly less data is required—or conventional machine vision can be used instead.

The training process is launched once and subsequently runs automatically in the background.

This allows Edge InspectiON to combine deep inspection capabilities with manageable setup effort, even across high product variety.