Font Size: a A A

Robot Visual Position Based On Self-Calibration

Posted on:2008-09-12Degree:MasterType:Thesis
Country:ChinaCandidate:X J WuFull Text:PDF
GTID:2178360212979505Subject:Pattern Recognition and Intelligent Systems
Abstract/Summary:PDF Full Text Request
Vision feedback control loops have been introduced in order to increase the flexibility and the accuracy of robot system. Specifically, eye-in-hand visual servo realized through the so called "teaching-by-show" approach has received a large and increasing attention in the last decades. This dissertation investigates the problem of dynamic self-calibration procedure which is carried out in parallel to visual servoing in the following aspect:An improved self-calibrating algorithm for visual servo based on adaptive genetic algorithm is proposed. Our approach introduces an extension of Mendonca-Cipolla and G. Chesi's self-calibration for position-based visual servo technique which exploits the singular values property of the essential matrix. Specifically, a suitable dynamic online cost function is minimized using an adaptive genetic algorithm instead of gradient descent method. The primary advantage of our approach is shown to be less susceptible to the initial value of the camera intrinsic parameters during the online optimization process, with comparable accuracy in the result. It is not neccessary to know exactly the camera intrinsic parameters; instead, only coarse bounds of the five parameters are necessary, which can be done once and for all offline. Another merit of the algorithm consists of a fast convergence speed. Besides, this algorithm needn't knowledge of the 3D model of the object. Extensive simulations are carried out and the results demonstrate that the proposed approach provides significant better result in both robustness and convergence speed with respect to the existing standard gradient-based self-calibrating visual servo strategy.We introduce the Enhanced Mutative Scale Chaos Optimization Algorithm (EMSCOA ) into the self-calibration problem, aimed at reducing the size of the parameters distribution scope in GA algorithm and providing significant better convergence speed. First, the chaos variables are mapped to the range of the five intrinsic parameters, and then a couple of cycles are set, chaos search in the inner cycle and the range is reduced in the outer cycle in order to avoid blind and repeated searching of chaos optimization in searching space and improve searching efficiency.We carried out experiments on MOTOMAN-SV3XL robot to verify the proposed approach. First we obtain several corners using a harris corner detector, then in order to get a high accuracyFundamental Matrix, we use normalized cross correlation together with symmetrical matching algorithm to setup a exact corresponding relation in different images. At last, a disparity gradient constraint is considered, which provides exact correspongding points in the binocular images. Based on these work, we designed a positon-based visual servoing control law. The trajectory of the corner in camera view denotes that all of them reache at the desired position, which demonstrates the effectiveness of the proposed algorithm.
Keywords/Search Tags:Dynamic Self-calibration, Robot Visual Servo, Adaptive Genetic Algorithm, Enhanced Mutative Scale Chaos Optimization, Computer vision
PDF Full Text Request
Related items