| 摘要 |
An object taking out apparatus is provided, capable of taking out randomly stacked objects (41) with high reliability and low cost. An image of one of the workpieces, as an object of taking out at a reference position, is captured by a video camera. Whole feature information (8) and partial feature information (10) are extracted from the captured image by a model creating section (6) and a partial model creating section (7), respectively, and stored in a memory with information (9) on partial feature detecting regions. An image of randomly stacked workpieces (41) is captured and analysed to determine positions/orientations of images of the respective workpieces (41) using the whole feature information (8). Partial feature detecting regions are set to the images of the respective workpieces (41) using the determined positions/orientations of the respective workpieces (41) and information (9) on partial feature detecting regions stored in the memory. Partial features of the respective workpieces (41) are detected in the partial feature detecting regions using the partial feature information (10), and priority of taking out the workpieces is determined based on results of detection of the partial features. A robot (31) is controlled to successively take out the object of the first priority using data of the position/orientation of the object of the first priority. <IMAGE> |
