In this paper, a feature selection and a twostage classifier for solder joint inspection have been proposed. Using a three-color (red, green, and blue) hemispherical lightemitting diode array illumination and a charge-coupled device color digital camera, images of solder joints can be obtained. The color features, including the average gray level and the percentage of highlights and template-matching feature, are extracted. After feature selection, based on the algorithm of Bayes, each solder joint is classified by its qualification. If the solder joint fails in the qualification test, it is classified into one of the pre-defined types based on support vector machine. The choice of the second stage classifier is based on the performance evaluation of various classifiers. The proposed inspection system has been implemented and tested with various types of solder joints in surface-mounted devices. The experimental results showed that the proposed scheme is not only more efficient, but also increases the recognition rate, because it reduces the number of needed extracted features.Index Terms-Bayesian classifier, feature selection, solder joint, support vector machine (SVM).
A novel four-eyes robust optical fiber distance and orientation measuring integrated proximity sensor is presented in this paper. The transducer can measure the distance and orientation of the object's surface at the same time, and is insusceptible to the fluctuation of the optical power output and variation of the object surface's reflectivity. The transducer's measuring attitude control is one of the important content studied in this paper. Base on the prior research, which verifies that the measuring pose does exist and there is only one such attitude, further study on optimizing the attitude adjusting control is done to improve the system's real time performances. In view of the complexity of the transducer's mode, an improved BP neural network data processing method is proposed to solve the difficulty in real time data processing. By partition the multi-output BP network into several single output ones and subsection the range of input variables, the sensor' s accuracy of pose estimation is significantly increased
A novel optical fiber proximity sensor for robot is proposed, which has fairly strong compensation ability to such factors as variation of optical source power and object reflectance. Composed of four receiving optical fiber, the sensor has a simple structure. Utilizing rotation freedom degree of robot's manipulator, it can measure the orientation and distance of the manipulator relatives to the object. This paper introduces the operating principle, the necessary and sufficient condition of the sensor's measuring pose, the optimum structure designing method which taking the minimum of measuring values relative error as objective function, emphasis is lays on the detecting technology based on Self Scan Photo Diode Array (SSPA) and the method for measure range enlarging, the sensor's measuring attitude adjusting and real-time data processing of sensor are also gone into detail.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.