扎捆棒材的图像识别计数算法

Counting Algorithm Based on Image Recognition for Bundled Bar

为了快速获得扎捆棒材的数量,在大量观测和实验的基础上,提出一种动态获取棒材形态学直径的方法;根据扎捆棒材的形态学分布,提出"米"字形搜索算法(RSS算法)和一种覆盖原则。先对棒材截面图像进行灰度化处理,然后根据灰度图片的灰度直方图两峰一谷的特点得到图片的全局阈值,将灰度棒材图片二值化;再对二值图片进行膨胀、腐蚀操作削弱粘连;最后基于二值图片,利用RSS算法进行计数。现场采集扎捆棒材的计数实验验证了该算法的有效性、可靠性和准确性。在加入人工矫正环节后,棒材计数准确度达到100%。试验结果表明,在计数对象满足两个分布特征的情况下,该算法都可以实现高准确度的计数。

In order to quickly obtain the number of the lash bar, the method of dynamically obtaining the morphology diameter of the lash bar is put forward on the basis of a large number of observations and experiments. According to the morphological distribution of the lash bar, the rice shape searching （ RSS ） algorithm and coverage principle is proposed. Firstly, gradation processing is conducted for section image of rod, then, according to the global threshold of the image is obtained in accordance with the two peaks one valley characteristics of the gray histogram of the gray scale image to get the binarization image,and then the adhesions is weakened by operations of expansion and corrosion,finally based on binarized image the RSS algorithm is used for counting. The experiments on site prove the validity, reliability and accuracy of this algorithm. The accuracy of bar counting is up to 100% when manual correction is added. The resuh of test shows that the conclusion is highly accurate counting can be implemented if the objects satisfy the two distribution features mentioned.