Measurement Of Feature Region On Curved Surface Based On Coordinate And Image Technique

Surface measurement is an important criterion to judge product quality and evaluate status in use. Now manufactory and scientific research are exigent of surface measurement on 3D surface in which there are many difficulties. A non-contact, high accuracy and automatic method combining coordinate measuring technique with digital image technique was introduced to measure feature region on curved surface. New principle and methods were proposed.Analyzing and generalizing various surface measurement methods, concept and representations of surface feature region are presented in favor of standardization at measurement condition, means and evaluation. The determine method of datum plane as well as regulation and items about measurement standard establishment was discussed. They were very practical for quantificational mensuration of surface feature region.A new image-coordinate measurement is presented for automatic scan measurement of 3D surface. The method adopted new techniques such as information fusion, non-orthogonal coordinate measuring and measuring robot. And it combined contact measurement with non-contact measurement, rough orientation with precise measurement as well as coordinate measurement with digital image technique. The design principle of software and hardware was summarized. Full basic theory of data processing was established. According to system characteristic, a plane glass templet was designed and a simple calibration method of camera partial parameter was developed. The key technique of restoring surface information using projection concept in computer graphics and measuring dimensions was expatiated. The image capture, extraction, coordinate transform, representation and data processing for surface feature region were discussed in detail. Error analysis and evaluations at scan-measure were given. In addition, with the aid ofδ₅ criterion, quality of image capture and rationality of threshold selection can estimate quickly. And then two engineering applications show that this method is feasibility and practicability. Practical tests show that this is high effective, accurate, reliable and universal method.Image processing capability was enhancive by modem information processing theory. A method combining self-organizing feature map （SOFM） neural network and error back-propagation training （BP） neural network was used to segment color image more accurately in favor of improving accuracy for object region extraction.ESSA （EMD Search and Snakes Approach） method combining empirical mode decom-position （EMD） with Snakes model was presented. Firstly, feature region was separated from complicated background by remodel EMD. And then, detection region was farther reduced by towards-center search method at feature region around. Finally, feature region edge was detected by iterative approach based on Snakes model. In this way the purpose of complicated object extraction from complicated background was achieved.To detect defects on smooth hemispherical shell surface with specular reflection, a novel fast method according to distortion image of reflection strip was proposed. This method overcomes disadvantage of considerable errors caused by background noise during image processing and is easy to analyze the defect features such as size, concave and convex.