Industrial vision

Which method for which project?
Industrial vision

Tailor-made computer vision technologies, sized for each camera control project

The success of a vision project depends of course on the analysis power of the algorithms, but also and above all on the judicious combination of different technologies. 


Whether for applications of : Detection, classification and localization of defects or anomalies - Event detection - Dimensional measurement - Aspect control - Character and code reading... each project has its own specificities: type of defect or process, level of precision and resolution, required conformity thresholds, imperatives related to the tool or the production environment... All these parameters need to be taken into account when drawing up the specifications, in order to be able to dimension an adapted and tailor-made response to each computer vision project.
Thus, while such an application will require parametric methods, for another, it will be necessary to train an artificial intelligence model, and for yet another, to combine several methods.

Our technology, the result of more than 10 years of R&D, embeds the best of the latest generation of algorithmic - parametric and AI (artificial intelligence) technologies. Our AX Vision and AX AI software suites combine different technologies and allow the creation and implementation of customized image processing and optical measurement applications.

Unsupervised AI anomaly detection method

  • How it works: The model is trained, based solely on compliance images. It learns to recognise an OK cap. Any differences with the training images are reported as anomalies.
  • Information:
    • Allows the approximate location, count and size of an anomaly.
    • Does not allow the classification of an anomaly (to recognize it with and name it).
  • Required: Compliant parts for learning and some non-compliant samples to validate samples to validate the effectiveness of the application.
  • Recommendation: Method adapted to controls for which it is not useful to name the It is not useful to name the anomaly. Method to be preferred for the inspection of products with little variation in a very stable environment.
  • Training and deployment in production: AX VISION software suite.
Unsupervised AI anomaly detection method
Unsupervised AI anomaly detection method
Unsupervised AI anomaly detection method
Detection of anomalies on plastic cap (pollution and tear). OK : conforms / NOK : non conforms. The detected anomalies are indicated by a heat map distinguishing the zones of compliance (blue) from the zones of non compliance (red).

Supervised AI defect detection method

  • How it works: The model is trained, based on images of compliance and non-compliance. By labelling the database, it learns to recognise and name specific features. Only trained defects are reported.
  • Information:
    • Allows the detection, location and classification of the desired characteristics in a very precise and more reliable way: shape, position...
    • Allows for precise trimming of defects by Bounding box or by Mask
    • Allows to parameterize the conformity thresholding to meet the inspection requirements.
    • Does not detect untrained defects.
  • Required: A batch of samples containing all the defects to be recognised in production.
  • Recommendation: A method that is necessary when trying to detect something defined with a high degree of accuracy.
  • Training: AX IA software suite.
  • Deployment in production: AX VISION software suite.
Supervised AI defect detection method
Supervised AI defect detection method
Supervised AI defect detection method
Detection, localization and classification of defects on plastic caps (pollution and tears). OK : compliant / NOK : non-compliant

Defect detection method Parametric

  • How it works: Algorithms parameterised and calibrated according to a threshold of
    defined by the specifications.
    Only out-of-tolerance measurements are reported as NOK
  • Information:
    • Allows dimensional measurements (micron, mm, m) and colorimetric measurements (LAB, HSV, RGB) with precision, in absolute values.
    • Allows multiple inspections to be managed, even at very high speeds.
    • Does not allow the classification of a defect, nor work on cases that have not been that have not been parameterised.
  • Required: A batch of samples of non-conforming parts with indication of the value of the defects.
  • Recommendation: Preferable method for dimensional and colorimetric measurements.
  • Parameterisation and deployment in production: AX VISION software suite.
Defect detection method Parametric
Defect detection method Parametric
Defect detection method Parametric
Detection and dimensional measurement of defects (pollution and tearing) on plastic caps. OK : compliance / NOK : non-compliance with the compliance threshold defined in the specifications.

Technology comparison table

technology comparison table

This website uses cookies

Acceptez-vous le dépôt et la lecture de cookies afin d'analyser votre navigation et nous permettre de mesurer l'audience de notre site internet ?