Research On Visual Detection Sensor Of Welding Robot

The automation, flexibility and intelligence of welding technology ensure the high quality,high production efficiency and low production cost. Robot welding has become the main featureof automation welding technology. Because vision sensor has noncontact measurement, moreinformation, sensitive, accurate and high anti-jamming performance, it becomes the inevitabledeveloping trend of welding robot. In order to realize auto-track control, the detection is the coretechnology. Now the seam vision tracking still has poor real-time, poor intelligent and has theerror brought by the preposed sensor, so it is necessary to develop a high-intelligent, highanti-interference performance and fast detection system, the system also can reduce the error.This paper mainly completed the following contents:To solve the influence by all kinds of interference factors and the size of vison sensor issues,a structure light sensor is developed. The main hardwares of vision sensor are selected, and thelight path and structure is designed. As a result, the size of sensor is decreased, and theanti-interference performance is enhanced.Considering the input of seam tracking is actual value, the vision detection system iscalibrated. It mainly include camera calibration, eye-in-hand calibration and structure lightcalibration.The relationship between the two dimensional image to three dimensional actualcoordinate is established.Otsu based on background correction is developed for strong interference welding image. Acompounding process method including this developed method, median filter, solitary point filteris to be used to pre-process seam images. As a result, the welding seam information is extractedrapidly and stably. Experiments show that using this compounding process, the seam detectingsystem has powerful anti-jamming ability. Within25ms, a537(H)×597(V) image can beprocessed, so it can meet the requirement of real time seam tracking.A fitting methods for front and back welding gun distance is developed, aimed ateliminating the error brought by the preposed sensor. Experiments show that this method hashigh accuracy, and can effectively reduce error. A seam information detecting system based on structured light sensor and using LabVIEWand Matlab are taken in practice. The system human-computer interface is friendly. Experimentsshow that the system is with high accuracy, and fast detecting.