摘要 |
PROBLEM TO BE SOLVED: To improve discrimination accuracy and shorten the discrimination time in a work posture discrimination device. SOLUTION: This work posture discrimination device 12 takes in the plan image of a work W with a camera 11 and discriminates which standard pattern the work W posture is, from the plan image. In a standard pattern among various characteristic quantity of the input image, a characteristic quantity with the difference in the discrimination is selected. The discrimination device threshold processes input images and calculates each characteristic quantity. By optimizing the weight and threshold of each characteristics quantity by using a neural network, improvement of discrimination accuracy and shortening of the discrimination time can be attained. |