Research On Some Key Technologies Of Close Range Photogrammetry System Based On Marker

With the development of industry in our country,the requirement for the precision of the industrial parts is becoming higher and higher.The use of traditional measurement methods has some shortcomings such as long processing cycle,tedious operation,and low degree of automation.The use of close range photogrammetry technology to measure the key points of the parts has the characteristics of fast,high efficiency,high precision and so on.Due to the lack of industrial parts surface texture,the encoding mark points and non-encoding mark points are placed on the measured object surface to construct the measurement environment,then photogrammetry technology are used to reconstruct the marks,the parts quality are evaluated according to the coordinates of these marks.Among the process,due to that the non-encoding point has no unique encoding information,how to get one hundred percent correct matching of non-coding points is a technical problem.The method of beam adjustment with additional distance constraint can incorporate the distance constrains into the overall beam adjustment,which can make the calculation result more accurate.The incremental reconstruction method allows person to observe the result of reconstruction after each photo shoot,and Timely feedback can be provided during the process of taking photos.The incremental Photogrammetry System based on the mark points are studied and discussed systematically and the related testing experiments is completed.The main research contents and results are as follows:First,the related basic theory of the photogrammetry are studied systematically.The marker extraction,recognition,and mark encoding and decoding method,the camera calibration method and random sampling method are briefly introduced,which provide a theoretical foundation for the implementation of incremental Photogrammetry System based on marker.Second,the matching problem of non-encoding point is studied,the disadvantage of the epipolar constraint two view matching and the conventional three view matching is analyzed.The better matching results can be obtained by using back projection error constraints without the occlusions problem of non-coding points.In this paper,the back projection error constraint and the minimum depth value constraint are used to further optimize the results of the conventional matching results.The results show that the proposed method can obtain higher accuracy matching whether the baselines of three views are approximately parallel or welled distribution,which provides a robust method for the implementation of the incremental photogrammetry system based on the mark points.Third,the bundle adjustment with the additional distance constraint is studied.To get a rough scale model by using the distance between the encoding point of the markers on the rule and the reconstruction process.using the zoom scale to get the approximately camera position coordinates and space point coordinates,viewing those coordinates as initial value of the bundle adjustment,then using the similarity transformation model as the transformation model between the object coordinate system and the camera coordinate system,and the distance residual error weights and image point residual error weights are assigned,using bundle adjustment to optimize the internal and external camera parameters,the camera distortion parameters,model scale,object point,finally,to obtain high precision reconstruction results.Fourth,an incremental reconstruction process are proposed,the method construct the initial model using the two images through the relative orientation,then the images are added one by one,matching correspond image points and space points to get new image pose,the part of current image remaining match encoding points is found in the remaining encoding markers between in the current image and in other images.The bundle adjustment is used to refine the related parameters,non-encoding point that have been reconstructed is mapped to the current image to find the corresponding point,according to the matching method of non-encoding point proposed in this paper.The residual non-encoding point in the current image matched with that in the other two images which have most encoding point coincided with the current image according to the encoding points.And in this process,camera calibration is performed.Experiments are carried out by using actual data to verify the feasibility of the method.