PiXiT®

Images and signals classification

Industrial vision, and in particular image classification, is a key component of many manufacturing tasks such as products sorting, control, identification, etc. The vital elements of such a vision system are both the hardware component (CCD, lightning, etc.) and the software component, which processes the acquired image to draw control actions and alarms.

The key problem with current image classification systems is the lack of generic procedure and algorithms. For each classification problems you need to manually extract the relevant features that will be useful to classify the image. In some cases the images processing is so complex that it is not suitable for real-time applications.

In partnership with the University of Liège, PEPITe has worked in the area of automatic learning and computer vision in order to bring an answer to these key issues. The results of this work have been included in PiXiT, a toolbox developed and maintain by PEPITe.

Some highlights of PiXiT 1.0

• Builds easily and quickly image classifiers
• Uses generic methods to build the classifiers
• Manages very larges images in many different formats
• Deploys the classifiers very fast (suitable for demanding real time applications)
• Constructs robust and relialable classifiers

This technology is adaptable to any indexed information classification (temporal signal classification, 3D images classification, etc.).