Papers by Keyword: Machine Vision

Abstract: Optical film is an important component for LCDs, cell phones and other mobile devices. It is crucial to ensure an optical films is defect-free. However, derive from the inherent translucent characteristic, many images of defects of optical films are low contrast. Low-contrast defects are notorious for their difficult to be detected. As some low-contrast defects are barely visible to human being without special training, it is indeed a challenging task to capture low-contrast defect images. To overcome the difficulty, we thoroughly investigated various of imaging and lighting technologies. After images have been captured successfully, we devided the captured defects, based on their sizes, into two categories, i.e. macro defects and micro defects. To detect both large and small defects, we set up a macro inspection system and a micro inspection system, respectively. In the study, we provide an easy to understand approach to reveal low-contrast defects. The experimental results showed that the proposed inspection system and method is capable of detecting low-contrast defects such as uneven coatings, streaks, white spots and foreign particles.

Abstract: This paper proposes a machine vision-based, servo-controlled delta robotic system for solenoid housing placement. The system consists of a charge-coupled device camera and a delta robot. To begin the placement process, the solenoid housing targets inside the camera field were identified and used to guide the delta robot to the grabbing zone according to the calibrated homography transformation. To determine the angle of solenoid housing, image preprocessing was then implemented in order to rotate the target object to assemble with the solenoid coil. Finally, the solenoid housing was grabbed automatically and placed in the collecting box. The experimental results demonstrate that the proposed system can help to reduce operator fatigue and to achieve high-quality placements.

Abstract: Grinding is an abrasive machining process which controls the surface finish of the materials in production of industrial components at the end stage. During the grinding process, wheel gets loaded with swarf and this loading will affect the cutting ability of the wheel and also the surface finish of the job. Hence the grinding wheel is to be dressed at appropriate time intervals and for necessary time duration. At present, the monitoring of grinding wheel loading and dressing is done by the operator i.e., the frequency of dressing and duration of dressing is determined by the operator based on his experience. Hence there arises a need to develop a systematic approach to this problem which can monitor the process and if not accurately should at least produce a realistic estimate of the percentage of wheel loading, frequency of dressing and duration of dressing. In machine vision system, Shutter speed is an important factor which will affect the image quality. Hence, in this work, attempts are made to monitor wheel loading using a low cost-high speed camera with variable shutter speed and the shutter speed is optimized for the better image quality.

Abstract: Generally after a cylindrical steel bar is made from hot forming process, its contour will look like ‘U’ or ‘wavy’ topography. The preformed steel bar after steelmaking process needs to remove the outside oxide layer by lathe or peeling machine (commonly known as stripping). As a consequence, the deformation of steel bar will cause subsequent problems, such as eccentric rotating of steel bar as the lathe is running. Although it is necessary to remove the oxide layer, the deformation will increase waste materials. To prevent the problems, a straightening machine is usually used to straighten the cylindrical steel bar before the pickling process. However, most current straightening machines cannot measure cylindrical steel bar contour on-line instantly. This study proposed a test method for measure the contour of a large cylindrical steel bar based on multi-line structured light and machine vision. Break line method were performed to measure the contour of the cylindrical steel bar. The experimental result shows that the measurement error is in the range of ±0.4 mm after calibration. The study result concludes that the proposed method can be applied to straightening machine for real-time online measurement to improve straightening efficiency.

Abstract: In this paper, a novel close to real-time artificial intelligent system for enumerating Total Viable Bacteria (TVB) in drinking water was developed by using pattern recognition and machine vision technology. In order to identify the viable bacteria accurately, four shape features including circularity ratio, eccentricity, rectangularity, and compact degree, and four color features (G_Rsd, B_Rsd, H_Rsd, S_Rsd) of the stained viable bacteria image were extracted. An optimal artificial neural network was used as the bacterial recognition classifier, whose inputs were the extracted feature parameters and output was bacteria signal or non-bacteria signal. By using this intelligent system, TVB counts in each sample can be enumerated within 1 h, but the traditional Aerobic Plate Count (APC) method will take us 48 h. The comparative test also indicated that the counting results by two methods are closely correlated (R²=0.9942). This close to real-time accurate information may contribute to melioration and instauration of drinking water safety systems and risk management for TVB.

Abstract: To improve measurement accuracy of safety belt pin, the size-sorting system which mainly includes a machine vision subsystem and a driving subsystem, is researched. The machine vision subsystem includes a industry camera, a double telecentric lens and a backlight. Image process steps that have sub-pixel accuracy are presented. The driving subsystem is designed with Atmega128L MCU. The state machine is designed for safety pin control logic. The two sub systems communicate with RS232 serial port. The test shows that the system has a maximum accuracy error of 0.05mm, repeatability error of 0.013mm.

Abstract: With the increase of requirement for the quality of raw materials in industry, surface defect inspection of steel bar has been an essential part of industrial production. The characteristics of vision-based detection technology for steel bar surface defect and the newest research development were introduced. The working principle of vision inspection technology and key issues were analyzed. Finally, the current domestic research emphases and development trends were proposed.

Abstract: Here we propose a new approach for measuring the environment for the mobile robot. Without using any old framework of sensors; we utilized the latest vision system -- the spherical camera system - which consists of several cameras around a certain geometric surface. Equipped with a lift device, the spherical camera system can measure the environment around itself. We present some results of simulated and real experiments showing the validity of our method. Our approach is acceptable and exciting, and will be applied in the mobile robot’s application in the future.

Abstract: Headlight detection was an important item of vehicle safety testing which main detection contents included light intensity and beam irradiation direction. It was to ensure the safe operation of vehicle at night or in adverse visual conditions. The basic concepts and testing standards of headlight were introduced, and the reasons of high failure rate for headlight detection were discussed. The main error correction methods of vehicle parking position in headlight detection were compared, and their advantages and disadvantages were analyzed. An error correction system of headlight testing measurement data was designed based on machine vision, and the process of system realization was given. It could provide a method to get more accurate measurement results of automobile headlight detecting.

Abstract: Based on the principle of machine vision technology, we designed a methodto detect the outline dimensions of automotive airbag quickly and accurately. We Used CCD camera obtain the airbag image, through the image processing method ofsmooth filtering andgray-scale transformationto complete pre-processing, finally applied Canny edge detection operator to extract the boundary of the airbag contour features,and then took the template matching methodto detect assemble error of the airbag image whether meet the requirement.The results show that the detection method have a higher precision, and the time is very short, it can improve the sampled positioningerror detection for the all checks image recognition detection, suitable for application in real-time online detection of airbag assembly line.