局部区域描绘子和SOM聚类算法在LED管芯计数中的应用  

APPLYING LOCAL REGION DESCRIPTORS AND SOM CLUSTERING ALGORITHM IN LED DIE COUNTING

作　　者：唐亮[1] 刘烨[1] 陈雁秋[1] 

机构地区：[1]复旦大学计算机科学技术学院,上海201203

出　　处：《计算机应用与软件》2012年第10期11-13,21,共4页Computer Applications and Software

基　　金：国家自然科学基金项目(611750 36);上海市教育委员会科研创新基金项目(10ZZ03);上海市重点学科建设基金项目(B114)

摘　　要：传统的图像处理算法往往受到光照干扰、角度变化、噪声等因素的影响,无法得到准确快速的识别结果。提出一种新颖的局部区域描绘子刻画目标点群的灰度值特征,并结合自组织映射SOM聚类,对发光二极管(LED)芯片中的管芯区域进行高效识别和准确计数。实验结果表明,该方法能够对LED管芯进行有效的识别计数,准确度高、鲁棒性好。Traditional image processing algorithms tend to be affected by light interference, angle variation, noise and other factors, making it impossible to get identification results accurately and quickly. This paper presents a novel local region descriptor to describe the gray value feature of the target point group. When combined with self organising map （SOM） clustering, it is able to identify the die area in lightemitting diode （LED） efficiently and to calculate the die number accurately. Experimental results show that this algorithm is capable of counting the LED die number effieiently with higher accuracy and better robustness.

关 键 词：局部区域描绘子 SOM聚类 LED管芯计数 过检测策略 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]