Detection Technology For Surface Defects In The Solid Rocket Motor Liner Based On Machine Vision

The liner plays an important role in the solid rocket motor. The surface defect of the liner such as pit defect and convex defect can cause deviation of thickness. The liner thickness over the technical requirement not only affects the rockets'performances, but also leads to an safety accident. Therefore, the research on detection technology for surface defects of rocket motor cladding has important practical significance. This paper takes a certain type of rocket motor cladding as research object, study on the detection technology for surface defects which is based on computer vision and develops an automatic detection system.According to the system detection requirements and structure characteristics of detection objects, this paper designs the overall structure of system and chooses appropriate hardware equipment to build the platform. In consideration of the camera's clear focus range is limited, a plan of image acquisition which photographing cladding through multiple equally intervals is chosen. The collection intervals of system is obtained by calibrating the camera.In accordance with the change rule of image grey value, a defect scanning mode based on circular scan has proposed, which taking image center as circle, and detecting the annular region of image which meet the system accuracy. In order to effectively reduce the influence of background noise on the judgment of defect feature, a method of combining with image schema matching and median filtering is propounded to process the image. The processing results demonstrate that the noise-abatements are effective and the features of defects are well protected. On the basis of self-adaptive threshold segmentation and binarization for the image, the defect area is filled by using seed filling algorithm, and the distance of defect to motor tail and its area size are calculated. The system has finished the automatic detection of surface defects of rocket motor cladding. The auto-testing results indicate that the system detection resolution reachesФ3mm, and the error of defect location is less than 2mm.