Study Of Control System Base On BP Neural Net And Computer Vision Measurement Technology

While a Quilting or embroider machine was processing flexible material at a high speed,the material will be distorted so visibly that actual maching track departure from the scheduled track because of the change of the direction and speed..The distortion is influenced by so complicated factors that the maths model is hard to establish,for example, the distance between the maching point and the frame, the elastic coefficient of the flexible material and the maching speed.So the system only using the feedforward control doesn't has the accuracy .But the system only using feedback control, because of the distortion changed with movement parameter and the hysteretic of the compensation,often results in violent surge and lag.The paper introduced neural net and computer vision measurement,and combined feedforward control with feedback control. According to the scheduled track, movement parameter,the materialan and current position, the rought distortion can be gained by means of BP neural net before the needle drops.The distortion will be used to correct the initial input. In the process after confirming the process track, the main controlling computer needs to run the image correlation procedure of match operating. The procedure functions in the compare between partial image of dynamic collection and corresponding image of the airscape. Basing on the digital image manipulation, we can gained the departure between actual maching track and the scheduled track.It will be used to train the neural net online. Simultaneously it will be also added to iuput of the neural net and the sum will be as total control input for resolving the problem of violent surge and lag to gain the accuracy of the system.For increasing the compute speed of the BP neural net,the paper introduces the scheme which use the FPGA based on the PCI bus to realize BP neural net and predesigns two important cell after analyzing the BP algorithm .