Fast and accurate object recognition by optimizied Hough-Transformation
- Maximized speed by using pyramid images
- Optimized precision in position and rotation ...
Single Runtime License for SHAPEFINDER
- delivery: softwarecode (Magic Number)
Object recognition using geometric information
CVB ShapeFinder uses the contours of an object to provide highly accurate position, scale and rotation information. The software is very resistant to partial occlusion ‒ where an object is partly covered or subject to noise or dirt. This resilience extends to dealing with reflections and changes in illumination.
Since CVB 2019, CVB ShapeFinder can take advantage of NVidia GPUs to provide accelerated search in 64b Windows and NVidia embedded platforms. With the right GPU, an improvement in search speed of 250% has been achieved.
CVB ShapeFinder uses a generalised form of the Hough transformation where patterns are located solely on the basis of localised contrast in the greyscales of an image. These contrasts correspond to pixels around which the greyscale values change significantly. In contrast the algorithm ignores image areas with only a low level of greyscale variation.
Users can define an appropriate contrast threshold and thus control the computational demands and the quality of the results. Another interesting characteristic is its ability to search for more than one pattern type at the same time, with minimum increase in search time as the search window only needs to be scanned once when searching for multiple models.