LED Chip Detection Algorithm Research And Implement

LED chip detection system is one of the important steps in LED production. This system could make production automatic and chip classification effective; meanwhile, it could improve the detection accuracy rate, provide necessary information for the next production phase and save a lot of time, human and material resources. As for the existing equipment, image processing software is a key component of the LED chip detection system because the image processing quality affects the recognizing results of the entire system by means of influencing the image processing speed and accuracy rate. Currently, there is a great room for the improvement of the image processing algorithms, especially for the core algorithm. The existing system combined with some new algorithm methods could be used to distinguish images more accurately. For the various current overseas image recognition software systems, such as Belgium eVision, Canada Sapera, the guidance of determining the number of the target image in advance is needed in the practical application of processing, which greatly restricts the its application in LED chip testing machinery and fails to meet the requirements of the system.LED chip detection system is a comprehensive computer-based image processing system. This paper aims at rendering an in-depth study on the image processing algorithm in the process of LED chip detection to make it rapid and efficient, to make the whole detection process fully automatic and to localize the production of the LED chip.The requirements and objectives for the detection is to obtain chip array location of each block through a camera by means of high-speed and multi-similar target detection. This work is the core of image processing and is influenced by the quality of captured image, the complexity of chip design, the appearance of the chip and the recognition algorithm. It is planned to set the alternative testing condition and adopt artificial shielding state to complete the chip appearance detection concerning the geometry borders, surface scratches and electrode defects, and to classify the LED chips automatically.On the condition of ensuring the accuracy of LED chip detection machine, the efficient real-time image processing algorithms on the LED chip detection system plays a decisive role. This paper includes the research in four areas: the similar objective setting, edge detection, target detection and segmentation of the graphic images. The paper is divided into five chapters:Chapter one gives a brief introduction of the background of selection of the topic, purpose, significance and the main work which should be conducted relating to the topic during the research.Chapter two focuses mainly on the similar objective positioning algorithm, which can distinguish the precise positioning when the similar multi-objectives come together into the same area and can provide target positioning recognition method randomly. Compared with the traditional method, this method can accelerate the speed under the premise of guaranteeing the accuracy of the target positioning.Chapter three conducts a study mainly on the edge detection algorithm by analyzing the Canny edge detection algorithm. Pointing out the shortcomings of this method, the author provides a self-adaptive Canny edge detection method based on entropy theory and improves the quality and self-adaptability of the algorithm for the detection.Chapter four studies mainly the rectangular and circular detection algorithm. The author proposes a detection method combined with the OpenCV visual image library area, which can achieve a quick and accurate detect effect.Chapter five focuses on the image segmentation algorithm by analyzing the fixed threshold segmentation images and dynamic threshold segmentation algorithm, proposing dynamic image segmentation method. This kind of algorithm can achieve dynamic segmentation of the LED chip and appears a good effectiveness.This paper conducts a comprehensive study on a series of methods of LED chip detection to achieve following objectives: to recognize the numbers of the LED chip accurately under the premise that the number of the LED chips is uncertain; to detect LED chips boundary precisely by edge detection method on the condition that edge information is guaranteed; to find out the border which rectangular and circle are interested in without losing the target area; to give a dynamic segmentation in the LED chip target area so that the objective and background can be separated. Namely, in this paper, the author manages to find a method to detect the corner collapse, cracks and scratches of LED chip so as to better the classification of LED chips.