Research On 3D Object Matching Method Based On Dual Quaternion Algorithm

3D object recognition is a challenge problem in computer vision, in which the most important task is how to identify the shape and the pose of the object rapidly and accurately. There are many practical applications in project. In visual navigation, for example, the motion between successive positions is usually either small or approximately known. But regarding to complex object especially free-form object, it is difficult to achieve proper registration result because of the complexity of representation and low segmentation reliability. The matching algorithm described in this report that is based on geometric principle of model meets this need. In process of recognition, objects are represented by free-form curves, i.e., arbitrary space curves of the type found in practice. A curve is available in the form of a set of chained points. The algorithm deals with 3D object coordinate directly without need of an explicit object description in terms of primitives. The idea of algorithm is lied in iteratively matching points on one curve to the closest points on the other, which based on iteratively closest point (ICP) algorithm. A least-squares technique is used to estimate 3D motion from the point correspondences, which reduces the average distance between curves in the two sets. The results of emulation experiment with synthetic and real data indicate that it is an efficient and accurate estimation method of 3D object recognition and registration.The report summarizes three problems including sensor types, 3D object representation methods and matching strategies during 10 years, which need to be solved by 3D object recognition system, furthermore, it classifies and sums up those of main methods. Then, we work out an accurate and efficient mathematics method, dual quaternion algorithm, which can be used to estimate 3D object motion parameters, and validate the algorithm's performance, precision and speed by experiments. Last, we apply the dual quaternion algorithm to matching algorithm of 3D object recognition and make use of least-squares technique to propose a matching method based on iterative closest point. The key technologies andmain operation steps of algorithm are confirmed by demonstrating several problems in practical applications. We apply emulation experiment with synthetic and real data to prove the validity and robust of our method.