Zhang Guangjun

The application of machine vision in inspecting position-control accuracy of motor control systems

In this paper, a new structured-light machine vision technique based on a radial basis function (RBF) neural network is proposed and an inspection system is established. General structured-light machine vision techniques are usually based on accurate mathematical models and have some unavoidable and inexpressible errors. The proposed new technique is based on the training and learning of high-accuracy samples and overcomes the disadvantages of the general technique and considerably improves the accuracy of machine vision inspection systems. An experiment applying this new technique to inspect the position-control accuracy of a step-motor controlled stage with one linear translation axis shows that the RBF artificial neural network (ANN) is quite suited to structured-light machine vision inspection systems and that structured-light machine vision inspection techniques are really a novel and effective means for the inspection of the position-control accuracy of motor control systems.

An Application of Ant Colony Algorithm for Edge Feature Extraction in City Aerial Image for Building Recognition

Over the years, automatic and semiautomatic extraction of building from aerial and spatial images has become an outstanding issue. This paper presents an application of ant colony algorithm (ACA) for edge extraction in city aerial image for building recognition. The overall idea of this work is clustering pixels by a character complex of gray value, gradient and neighborhood of each pixel. The results of ACA will be compared with the results of Canny operator which is considered as the best edge feature extracting operator. And the comparison indicated that ACA is efficient in edge feature extraction, especially for the aerial images.

Estimation of projected circle centers from array circles and its application in camera calibration

A method of detecting projected circle centers (PCC) based on array circles is proposed. The perspective distortion of projected circle centers could be eliminated from coplanar circles. By utilizing the property of projected coplanar circles, lines connecting PCCs are computed from projective invariant points defined by the projected circles (PCs), and then the PCCs are determined from the intersection of the computed lines. Once PCCs are recovered, the camera could be calibrated from classic method such as Zhangs. Both synthetic and real data have been used to testify the proposed algorithm and good results have been obtained. The calibration results show that the estmation of projected circle centre bears high stability and accuracy, even when the image noise is massive.

Practice on cultivating computer application ability for the Opto-electronics Engineering students in graduation design

Economic progress and information technology development have spurred the demand of multi-talents. So cultivating the compound talents with computer application ability is one of the main duties for the Subject Construction of Opto-electronics Engineering. According to this, the hierarchical curriculum system of the computer application education for opto-electronics engineering undergraduates is briefly introduced. The problem of how to strengthen training of integrated capability especially the computer application skills of students through graduation design is expatiated emphatically. Taking an example of the graduation design based on LabVIEW, which is such an icon-based graphical programming language that allows engineers to express their ideas in a more intuitive and natural way, the following means are explained: enhancing the capability of problem analysis in starting phase; training of solving problem independently in execution phase; cultivating the ability of induction and summarization in the thesis writing period. The practices show that the above strategies can help students to improve multi-aspect capability and to meet the requirements of society better.

Robust and accurate star segmentation algorithm based on morphology

Abstract. Star tracker is an important instrument of measuring a spacecraft’s attitude; it measures a spacecraft’s attitude by matching the stars captured by a camera and those stored in a star database, the directions of which are known. Attitude accuracy of star tracker is mainly determined by star centroiding accuracy, which is guaranteed by complete star segmentation. Current algorithms of star segmentation cannot suppress different interferences in star images and cannot segment stars completely because of these interferences. To solve this problem, a new star target segmentation algorithm is proposed on the basis of mathematical morphology. The proposed algorithm utilizes the margin structuring element to detect small targets and the opening operation to suppress noises, and a modified top-hat transform is defined to extract stars. A combination of three different structuring elements is utilized to define a new star segmentation algorithm, and the influence of three different structural elements on the star segmentation results is analyzed. Experimental results show that the proposed algorithm can suppress different interferences and segment stars completely, thus providing high star centroiding accuracy.

A circular aperture-array structure optical system for digital sun sensor

AbstractsBased on the first type of Rayleigh Sommerfeld diffraction formula, an imaging model of circular aperture-array structure digital sun sensor optical system is developed. Then a 6×6 circular aperture-array structure optical system is designed. The results of numerical simulation show that the optical system is designed well and is conformed to the requirements of miniaturization and high accuracy of sun sensor.