Abstract
Object recognition and visual positioning frequently involve major difficulties like variable illumination, low contrast, occlusions and shadows. Experience has shown that binary methods do not give consistent results in the factory. A gray-level method is reported for recognition and alignment that exploits local features of object contours in matching the model with the scene. The method allows for recognizing two-dimensional objects, which differ from the modeled object, in nonoptimal conditions. The strategy of the method is to accumulate local evidences of matches using a Hough-type procedure.
| Original language | English |
|---|---|
| Title of host publication | Eighth International Conference on Pattern Recognition |
| Publisher | IEEE Institute of Electrical and Electronic Engineers |
| Pages | 773-775 |
| ISBN (Print) | 978-0-8186-8742-6 |
| Publication status | Published - 1986 |
| MoE publication type | A4 Article in a conference publication |
| Event | 8th International Conference on Pattern Recognition - Paris, France Duration: 28 Oct 1986 → 31 Oct 1986 |
Conference
| Conference | 8th International Conference on Pattern Recognition |
|---|---|
| Country/Territory | France |
| City | Paris |
| Period | 28/10/86 → 31/10/86 |