Objects Geometry Comparative Analysis Method for Industrial Robot Vision System

摘要

At present, in computer vision systems, neural networks are used to process information received by the system from cameras. The recognition of all objects on the image is an extremely resource-intensive task, the solution of which consumes most of the computing power. For that reason, systems based on neural networks cannot be fully utilized for real-time systems due to limited computing resources. To build real-time computer vision systems, the authors suggested using the contour comparison method. The method allows to supervise the geometry of objects, conduct presorting and screen out defective parts, thereby the pressure on neural networks will reduce. The method is implemented in the Java. The created software performs image processing and objects search on it, that are the most similar to the template. The results of the experiment showed that the desired object is correctly determined on a noisy image and the proposed method can be used to solve the problem of pattern recognition in technical vision systems.