基于机器视觉的智能破竹分片系统设计与验证

Design and verification of intelligent bamboo fragmentation system based on machine vision

【目的】破竹分片是影响竹材出材率的重要基础工序。针对竹段的直径大小跨度大、壁厚不等、圆度不一的个体化特性,提出基于机器视觉的智能分片方法,以提高竹材的出材率。【方法】基于竹段的天然特性和破竹工艺要求,以竹段的真实截面轮廓为基础,以刀盘可分片的数量为约束,建立出材率计算模型。通过工业相机采集竹段的截面图像,在Canny算子的基础上引入多尺度细节增强算法来消除竹材图像内、外径间的干扰信息,并通过融合迭代阈值与梯度直方图分析法自适应获取双阈值,以提升竹段截面轮廓提取的鲁棒性。采用边界排序生长算法快速计算剖分竹片的最大内接矩形。在智能破竹机上开展生产试验验证。【结果】改进的轮廓提取算法,有效去除了竹材截面图像的伪边缘特征,能获得完整的边缘轮廓特征,计算获得的竹材内外径与真实尺寸的平均误差为0.9%。同一随机剖分角度下,本研究方法计算的出料率与最大出材率相比平均偏差为1.3%,小于圆模型的6.3%和椭圆模型的1.6%。通过生产试验,验证了基于机器视觉的破竹机智能分片系统的出材率平均可达73%,高于传统破竹机出材率。【结论】设计的基于机器视觉的智能分片系统,能快速获取竹段的真实截面轮廓,精准确定剖分份数,有效提高了竹材的利用率。

【Objective】The process of bamboo splitting is a crucial step that affects the yield rate of bamboo materials.To address the individual characteristics of bamboo segments,such as varying diameters,uneven wall thickness,and inconsistent roundness,we propose an intelligent splitting method based on machine vision to improve the yield rate of bamboo materials.【Method】Based on the natural characteristics of bamboo segments and the requirements of bamboo breaking technology,the calculation model of wood yield was established on the basis of the real cross-section outline of bamboo segments and the number of cutter segments as constraints.By using an industrial camera to capture cross-sectional images of bamboo sections,a multi-scale detail enhancement algorithm was introduced based on the Canny operator to eliminate interference information between the inner and outer diameters of the bamboo image.The dual thresholds were adaptively obtained by integrating iterative thresholds and gradient histogram analysis to improve the robustness of bamboo section contour extraction.Using boundary sorting growth algorithm to quickly calculate the maximum inscribed rectangle of sliced bamboo.Conduct production testing and verification on intelligent bamboo breaking machine.【Result】The improved contour extraction algorithm effectively removed pseudo-edge features from bamboo material images and obtains complete edge contour features with an average error between calculated inner/outer diameters and actual sizes at 0.9%.Under random initial splitting angles,our method showed an average deviation from maximum yield rate at 1.3%,which was lower than 6.3%for circular model or 1.6%for elliptical model.Through production experiments,it is verified that the average yield of the intelligent segmentation system based on machine vision can reach about 73%,which is higher than the yield of traditional breaking machines.【Conclusion】The intelligent segmentation system based on machine vision designed can quickly obtain the true cross-sectional profile of bamboo segments,accurately determine the number of segmentation parts,and effectively improve the utilization rate of bamboo.