AI vision system for quality control
K-QUALITY
Simple, Modular and Performing
K-QUALITY is the AI-based vision system for quality control with neural networks that detects on production lines any defective parts or parts, contaminants, cosmetic defects, missing components, assembly errors, etc. with the same reliability as human inspectors and higher, but with the speed of a computerised system.
Simple to use and scalable, KabVision’s machine vision systems operate in high-precision manufacturing industrial environments.
Machine vision systems based on deep learning can solve the most complex inspection challenges in industrial environments.
The wide variety of software packages, cameras, optics and lighting systems that can be integrated allow K-QUALITY to adapt to all kinds of production requirements in order to eliminate waste while optimising production.
The hardware used in K-QUALITY’s vision systems enables applications where spectra other than the visible (infrared, x-ray, etc.) need to be studied.
The goal is always fewer recall events and lower rework costs.
The scalar architecture allows the realisation of the simplest applications up to more complex machine vision systems, using a single programming language.
Deep Learning
The K-QUALITY system involves data collection as a first step.
This is referred to as the classification system, i.e. the collection of a dataset of images to be labelled. These images should cover all categories or classes to be recognised by the model. The more precise this step is, the better trained the neural network will be.
The images are then reworked (normalisation of pixel values, resizing and other transformations) before being used to train the model.
The next step is the creation of the model. Here, a deep convolutional neural network (CNN) is created that is capable of capturing higher-level hierarchical features and inherently makes the application robust to discrepancies between the dataset and reality.
Model training, validation and optimisation, and model testing are the last steps before integrating the model into an application or system to actively perform real-time image classification.
The choice of the most suitable hardware, and specifically the choice of the right lighting solution, in 75 per cent of the cases solve the application easily. We rely on the close cooperation and advice of IMAGE S an Italian distributor of components for vision systems.
The algorithm adapts to any situation, fully customisable to the application.
K-QUALITY uses our K-VISION vision platform that analyses and classifies complex problems in real time, even on high volume lines.
Simultaneous Quality Checks
Several quality checks can be requested simultaneously, e.g. the presence of cosmetic defects and correct assembly.
Ease of use
Fully configurable through easy-to-use operator panel checks;
total control of the vision system and complete view of the analysis in progress, as well as real-time results;
inspection recipe editable in real time without the need for expert operators;
interoperable: easily integrated with any PLC and SCADA via established communication protocols.
The system is Industry 4.0 and 5.0 eligible.
