一种自适应逐级开运算的原木端面识别方法

A Log Statistical Method Based on Step by Step Opening Operation

为提高原木检尺的工作效率,提出一种自适应逐级开运算的成堆原木端面识别方法,以实现自然环境下成堆原木端面的快速识别。获取自然环境下堆放的成堆原木图像,利用PSO改进的K-Means聚类算法基于图像的RGB色差值,对图像进行背景和目标的分割,用PSO算法寻找K-Means的初始聚类中心,通过背景和原木堆的聚类实现背景和目标的分割。对于分割后的图像,设计了一种自适应原木端面大小的逐级开运算原木端面识别方法,该方法根据图像中的目标原木大小,以圆面积计算公式中面积和半径的正比关系为依据,用半径作为结构元素,从而自适应地寻找出结构元素的最大值和最小值,在2个最值的区间内设定1个可变的结构元素,从最大值逐渐减少地对目标图像进行开运算,从而实现对图像中不同直径大小的原木识别,计算出原木堆图像中所有原木的根数。结果表明,使用逐级开运算进行原木识别,对清楞原木的检测达到平均正检率97.26%,漏检率2.74%,错检率8.10%;混楞原木检测正检率93.32%,漏检率6.68%,错检率7.02%;与使用Hough变换进行圆检测的方法相比,本算法清楞原木和混楞原木的正检率平均提高分别为16.24%和19.29%。自适应逐级开运算的原木端面识别方法,可以识别不同直径的成堆原木端面,从而实现原木根数的自动计算,能有效提高人工检尺的效率。

In order to improve the efficiency of measuring the log volume,and realize the rapid recognition for piles of logs in natural environments,an end face recognition algorithm for piles of logs was proposed based on adaptive step-by-step opening operation.Firstly,the background and target of the log image were segmented by the RGB difference value which was calculated through PSO's improved K-Means clustering algorithm.Secondly,a log-end recognition method was illustrated based on adaptive step-by-step opening operation to the segmented images.According to the size of the target log and the proportional relationship between the area and the radius,the radius used as the structural element,the method could adaptively find the maximum and minimum values of the structural element.A variable structural element was set in the interval of the two maximum values,and the opening operation was done gradually from the maximum value to the minimum value for the target image.Lastly,the log-ends with different diameters were recognized,and the number of all logs was calculated in the piles of logs image.The experimental results showed that the average positive detection rate of the same size logs images was 97.26%,the missing detection rate was 2.74%,the false detection rate was 8.10%,and the positive detection rate of the different size log images was 93.32%,the missing detection rate was 6.68%,and the false detection rate was 7.02%.Compared with the Hough transform method,the average positive detection rate for two types of logs images increased by 16.24%and 19.29%,respectively by the algorithm in this paper.The log end face recognition method with self-adaptive step-by-step operation can identify the end faces of piles of logs with different diameters,thereby realizing the automatic calculation of log volume and effectively improving the efficiency of manual rulers.The studies show that this method can identify the end faces of piles of logs with different diameters,automatically calculate the number of logs,and improve the efficiency of manual measuring.