摘要
为了解决人工检测瓶装饮料生产线倒瓶时存在的耗费人力和效率低等问题,提出一种基于快速图像增强的生产线倒瓶自动检测方法。针对原始图像建立2层高斯金字塔模型,提取该模型中第2层图像的亮度分量并对其进行增强处理;将得到的增强矩阵映射到RGB空间,实现了HSV空间到RGB空间的快速转换,根据拉普拉斯金字塔模型重构出最终的增强图像;对增强图像进行二值化处理从而分离出瓶盖,通过提取瓶盖的轮廓特征来检测倒瓶。实验表明,在对瓶盖未被遮挡的倒瓶进行检测时,该方案检测的准确率为100%,被遮挡时的检测准确率为93%,且对单幅图像的检测平均用时为60 ms,能够满足工厂生产线倒瓶检测对于准确性和实时性的要求。
In order to solve the problems of manpower and low efficiency in the process of bottle inspection in bottled beverage production line,a method based on fast image enhancement automatic detection of inverted bottle in the production line is presented. A two layer Gauss Pyramid model is built for the original image to extract the luminance component of the second layer image in the model and enhance it; the obtained enhancement matrix is mapped to RGB space to achieve the fast conversion from HSV space to RGB space. The final enhanced image is reconstructed according to the Laplace pyramid model. The enhanced image is binarized to separate the cap,and the inverted bottle is detected by extracting the contour features of the cap. Experiments show that,when the inverted bottle of the cap is not blocked,the accuracy of the detection scheme is 100%; when blocked,the accuracy rate of detection is 93%,and the average time for detecting the single image is 60 milliseconds,meeting the requirements of accuracy and timeliness of bottle inspection in the production line.
作者
郝蓓
杨大利
侯凌燕
肖勃雷
HAO Bei YANG Dali Hou Lingyan XIAO Bolei(Computer School, Beijing Information Science ＆ Technology University, Beijing 100101, China Beijing Centigram Co. Ltd, Beijing 100085, China)
出处
《北京信息科技大学学报（自然科学版）》
2017年第5期39-44,共6页
Journal of Beijing Information Science and Technology University
基金
国家自然科学基金资助项目(61401290)
关键词
生产线
倒瓶检测
快速图像增强
二值化
轮廓特征
production line
inverted bottle detection
fast image enhancement
binarize
contour feature
