Contour Detection | ingb-rts.de

Lighting invariant Contour Detection - lighting invariant Change Detection -Hyperbola Detection - Information Space Analysis - CI- based Identification - Data Categorisation - Alarming Routines - Machine Monitoring - Harmonic Analysis - DLS-Analyser - FD-Analyser - CI-Alarming Routines - Trend Analyser - Neural Based Quality Control - Sensor Fusion - Neural Based Prediction - Analysis of Biological Signals - Neural Based Identification - Acoustic Pattern Recognition - CI-based People Categorizer - Digital Twins

₪ …lighting-invariant contour detection

The basic structure of the retina is genetically determined to guarantee a quick and reliable analysis of the environment. While it was previously thought that the eye was "only" a type of camera, we now know that all essential object properties are already detected and analyzed here.

One of these properties is the object's demarcation from the background and from other objects using the contour property. This demarcation is realized in the retina by so-called on-off structures in combination with the so-called receptive fields, the latter of which adapt in shape and size depending on the lighting conditions.

The following shows some examples of the contour detections realized by IngB RT&S - based on the biological model.

For machine-assisted categorization, color and contour features were selected in such a way that, among other things, a reliable categorization with regard to the state of health is possible.

To analyze the condition of trains during operation and under difficult lighting conditions, camera images are first bionically enhanced and then transformed into an elevation representation using bionic contour filters, which are then analyzed by special neural networks.It should be noted that the original camera image is shown on the bottom along with contour detection, while the image optimization is shown on the top.

Another application of lighting-invariant contour detection is the "virtual safety fence". This involves monitoring the production area at intervals of 30 ms using a simple camera. Adaptive (mouse click) identification of endangered areas makes it possible to design and operate robot workstations in the context of safe human-machine interaction or safe object-machine interaction.