Study Of Vision-based Tracking And Positioning System For Moving Target In A Large-scale Space

As requirements are getting increasingly stricter in the precision and efficiencyof assembling of large scale equipment in many countries, how to measure theprofiles of details of a super large object in any kind of shape to meet the highassembling precision requirement has become a difficulty drawing keen attention inthe world. With the development of photogrammetric technology, imaging trackingtechnology and automatic control technology as well as emergence of high-precisionmeasuring devices catering for large space application such as laser theodolite, totalstation and laser tracker, it is possible to solve the above problem: on the basis ofphotogrammetric technology, with the help of a handheld terminal unit, the profile ofa remote detail of an object can be measured without contacting it; while imagingtracking technology enables high-precision large space measuring device toautomatically track the handheld terminal unit if guided by an external optical lens soas to measure the position and pose of the handheld unit in the global coordinatesystem on real-time basis; if such two technologies are combined, desired solution canbe achieved to provide accurate measurement of details of large size objects.Theodolite is characterized by sound precision and non-contact measuring mode,and therefore it is widely used in the field of large size object measurement. However,as a result of its low automation degree and small vision field, theodolite is unable tomeasure moving objects in a tracking way. On account of this, we introduced aphotoelectric servo tracking guiding system outside the theodolite, which consists ofhigh-precision camera holder and camera in conjunction with multiple-threadedcontrol program. The core of arithmetic includes: orientating and guiding theories ofperspective projection model, bundle adjustment method and optimization algorithm;target tracking theory of multithreading technology and Kalman filtering algorithm.The experiment proves that the method suggested in this paper realizes high efficientmeasurement of moving objects. This paper mainly involves the following contentsand work done: 1. The orientating theory of theodolite was analyzed theoretically, withemphasis on introducing the bundle adjustment method, and the program wasmade a Gauss-Newton algorithm based orientating program.2. Introduce was made to the guiding matrix constructing method that isbased on D-H model and bundle adjustment.3. The basic framework of tracking program was formulated: by means ofrecognition based imaging tracking technology, specific gray level distributionand thresholds were taken as the features for frame matching, and position ofcenter of mass and dimensions of the target were taken as the dimensional limits,and satisfactory results were achieved.4. The arithmetic of tracking program was optimized. With Kalman filterserving as the device and target observation as the correction of filteringestimation, the weights of observation and estimation are repeatedly calculatedand weighted iteration of the two was done, finally optimal estimated valueswere obtained. Due to allowance of free weight setting, such arithmetic is of bigrobustness, thus it effectively enhanced the tracking performance of the system.5. The communication module for the system was designed by MSCOMMcontrol of VC, leading to the possibility of remote setting of theodolite andremote real-time reading of measurement data via upper computer, so that theworking efficiency of the system was significantly improved.