The Extraction Method Of Laser Centerline Using Laser Assisted Stereo Vision Measurement For Hot Forgings

Large parts are the core base equipment of many national major technical fields such as ships, nuclear power, and coal chemical industry. The service conditions of these parts is extremely complex, therefore they are generally manufactured by forging processing method. Detecting the key dimensions of forging during the process in real-time can optimize the process, reduce the rejection rate and increase energy efficiency. To resist the high-temperature factors in free forging, the active binocular stereo vision measurement system which is assisted with blue-violet semiconductor laser beams is used to detect the key dimensions of forging on-line.In the laser-beams-assisted stereo vision measurement system, the reconstruction results’ accuracy and stability after matching are directly affected by the extraction of laser bar’s centerline, and it determines the final measurement’s accuracy of the key dimensions. However, the environment during the forging process is very harsh, the complex surface situation of forging, the diverse reflection modes of the assisted beams and the poor quality of the semiconductor laser give hot forging stereo vision measurement system extremely difficult extraction conditions. Therefore, the centerline extraction algorithm is studied for the terrible conditions of hot forging stereoscopic measurement, the main research contents are:1. From the emitting principle of semiconductor laser, a large number of laser projection experiments were conducted to research the distribution characteristics of the light strip intensity. Three structures of the laser light intensity distribution are summed up. The laser speckle phenomenon is analyzed and the imaging properties of speckle described by statistical method. We propose an automatic segmentation method on this basis considering the characterization of laser speckle and backgrounds are dissimilar. The light beams and backgrounds are separated and the extraction of laser gradation information is completed through morphological methods. By analyzing the denoising effect of image filtering, smooth gradation distribution is obtained.2. On the basis of conducting numerous curve fit experiments of light intensity distribution in tangential direction, a Gaussian mixture model is determined to describe the characteristics. The feature point selection is discussed and determined as the center extreme points. The existing extraction methods are studied to accomplish the centerline extraction algorithm flow of light strip. First, one row gray data of the light is intercepted to fit Gaussian mixture model, and the light of the distribution structure, the center position, and the fitting range of extreme points are solved. Next, center extreme point of adjacent rows is calculated on the basis of known rows.Repeat this step until the light of the regional boundaries to obtain light strip centerline.3. Finally, replication experiment was carried out to verify the accuracy of the centerline extraction method proposed in this paper.The center line extraction experiments of different light stripe structures showed a stability under 0.22 pixel; the error of diameter measurement experiment of cylindrical sample is 0.33%; and an experiment at forgshop was conducted which result is relatively reliable.