基于矛盾解决理论与物质——场分析的实木地板节子识别新方法研究

针对传统区域生长方法识别实木地板节子存在准确率低且速度慢的问题,运用TRIZ中矛盾解决理论分析与物质——场分析,提出一种结合分水岭、区域生长以及边缘检测的新的实木地板节子识别算法。算法首先将原图像转换为灰度图像;其次,运用形态学分水岭的方法对灰度图像进行分割;再次,选取满足条件的种子区域进行区域生长,得到节子区域;最后,运用Sobel算子对图像进行梯度运算,并找到节子的边缘。仿真实验表明,该算法较传统方法能够找到更合适的种子区域和区域生长的阈值,实现了对节子的快速、完整提取,节子分割平均用时60ms,平均辨识准确率在90％以上。     

边缘识别法在节子整体视频检测中的验证

通过对原木节子轮廓的数学描述和对节子边缘点的确定,提出一种比较准确的寻找节子边缘点的视频识别验证方法———边缘检验算子法,对节子的形状进行有效的分析和验证。提出了一种有效的节子数学描述和视频处理方法,为节子的视频检测技术提供了新的理论。     

梯度算子与灰度阈值融合的实木地板节子识别方法研究

实木地板表面节子的存在直接影响其质量与等级。针对节子区域与周边存在差异,传统边缘检测方法存在漏检、误检的问题,提出了一种新的边缘检测融合算法。算法首先对原图像分别进行灰度与阈值处理,然后对两幅图像进行梯度运算,并对结果进行操作;最后运用形态学方法得到节子边缘。研究证明,该融合算法保证了节子提取的完整性,可提高节子的识别率。     

世界节子研究进展

通过对大量文献的整理分类,从节子自身研究状况、对木材加工利用的影响和节子的无损检测技术几个方面,对截至目前世界上对节子研究的进展状况进行了综述,并提出了今后的一些研究发展方向。     

基于AOS的扩展C-V模型及背景填充耦合的单板节子缺陷识别

分析木材单板节子缺陷图像的特点,提出一种基于AOS的扩展C-V模型及背景填充耦合的单板节子缺陷识别算法。首先,对Chan-Vese提出的基于Mumford-Shah模型的水平集图像分割算法进行改进,使分割速度得到提高;其次,用AOS算法改进原模型的差分格式,使得差分格式无条件稳定;最后,结合背景填充技术,使得到的新图像缩减了目标与背景间的特征差别。通过对比试验,表明该分割方法能够快速识别单板单个节子缺陷,充分说明该耦合方法比Chan-Vese方法及其改进方法有更好的分割效果。通过用多水平集作为初始轮廓演化曲线,结果表明该方法也可快速实现对单板多节子缺陷图像的识别,实现对单板节子图像的多目标分割。     

基于Faster R-CNN的实木板材缺陷检测识别系统

我国木材资源有限,为了提高木材的利用率,采用机器视觉来实现木材缺陷快速而稳定的检测,不仅可以克服人工检测的低效率和木材缺陷识别的低准确率,而且对提高木材加工企业的智能化水平具有重要意义。为了高效、快速、准确地进行无损检测,采用深度学习方法,建立了一种基于快速深度神经网络的实木板材缺陷识别模型。首先采用Resnet V2结构对采集到的实木板材缺陷图像进行特征提取,然后应用该模型对节子、孔洞等实木板材缺陷进行训练学习,最后构建了Faster R-CNN检测框架,并使用tensorflow开发平台对节子、孔洞等实木板材缺陷进行预测输出。具体选取了2 000块杉木样本,通过旋转对原始的实木板材图像进行数据扩充,扩充后图像的80%作为训练集,20%作为验证集来进行仿真。仿真结果表明,该模型对实木板材节子缺陷检测正确率为98%,对实木板材孔洞缺陷检测正确率为95%,验证了将深度学习算法应用于实木板材缺陷检测中的有效性。     

基于MEA-BP神经网络对木材内部缺陷诊断的研究

为了提高木材内部缺陷的自动识别率,采用电阻层析成像(ERT)的方法获取电导率波动信号,通过小波包变换对采集的数据进行3层小波包分析,对八维特征向量进行提取,利用思维进化算法(MEA)优化权值和阈值,孔洞、节子、腐朽试样各45组数据,进行BP神经网络训练,每种缺陷20组作为测试集,识别木材内部缺陷。结果表明:MEA-BP神经网络对木材孔洞、节子和腐朽的识别率分别为96.92%、95.38%和92.31%,该模型解决了复杂组合的优化问题,提高了搜索效率,并且达到最佳的预测效果。     

基于模糊聚类分析的木材缺陷CT图像分割

为了提高木材的利用率,在木材加工之前对木材缺陷CT图像进行分割,将节子和空洞等缺陷分割出来,通过观察缺陷的位置便于工人师傅下锯。利用计算机断层扫描（CT）技术获取木材缺陷图像,将数字图像处理技术与模糊聚类算法相结合,在标准的模糊C均值算法的基础上改进,采用半模糊聚类的分析方法对木材缺陷图像进行分割检测。实验结果表明：基于半模糊聚类的图像检测方法在木材图像检测上取得了较好的效果,缺陷边缘处很平滑,细节保留完整,更多的保留了边缘上的信息。从而证明了半模糊聚类分析法在木材缺陷CT图像处理方面具有可行性。     

基于空气耦合式超声波的锯材节子缺陷检测

针对锯材的节子缺陷,基于空气耦合式超声波的传输系数与待测样本密度之间的关系,探索了空气耦合式超声波的节子缺陷检测新方法,并采用自行研制的空气耦合式超声波检测仪对杉木锯材试样进行检测,结果表明,空气耦合式超声波技术不仅能检测出锯材表面可见的节子缺陷,也可以有效检测不可见的内部节子缺陷。     

木材缺陷检测理论及方法的发展

通过对国内外板材节子视频检测理论现状和发展的研究，对计算机视频技术在木材缺陷检测上的应用与发展进行了综述，并对视频技术在木材缺陷检测上的应用前景进行展望。     

A comparison between three different methods of measuring knot parameters in picea abies

Both foresters and sawmillers are interested in the knot structure of trees; in particular, position and number of knots, knot diameter, knot length and dead knot border. For research purposes, it is possible today to carry out non‐destructive measurements using computer tomography (CT) and image analysis. The aim of this study was to measure knot parameters on Norway spruce (Picea abies (L.) Karst.) using a non‐destructive method developed for Scots pine (Pinus sylvestris L.), and to compare the results of this method with the results of two different destructive methods. In order to do this, two Norway spruce stems were scanned by CT. Then five logs from one stem were cut into flitches 20 mm thick and the defects on the sawn surfaces were scanned manually. The other stem was cut just above every whorl and then each knot was split through its centre and the knot parameters were measured manually. The study showed that the CT method compares well with the destructive methods. It is a reasonably fast, non‐destructive method which measures position and diameter of knots and detects larger knots with acceptable accuracy. The study also showed that a large number of smaller knots were not found by the CT method and that the CT method measured knot length and dead knot border with low accuracy. This means that the CT method has to be adjusted to Norway spruce in order to improve its ability to measure knot length and dead knot border and to detect smaller knots.     

基于优化卷积神经网络的木材缺陷检测

针对深度学习中的卷积神经网络算法,在木材无损检测过程中存在缺陷定位不准确、缺陷轮廓和边界信息不完整、识别精度需进一步提高等问题,利用非下采样剪切波变换最优稀疏表示特性,以及简单线性迭代聚类算法能很好地保持像素紧凑度和图像边界轮廓的优点,设计了一种优化的卷积神经网络算法,以提高木材无损检测的准确率。首先采用非下采样剪切波变换对采集的木材图像进行简单预处理,保留木材图像的缺陷特征不丢失,降低图像处理的复杂度以及运算量;然后利用卷积神经网络对木材图像实现深层次的算法设计,同时应用简单线性迭代聚类算法对初步模型进行增强改进,提取出相对准确的木材缺陷轮廓;最后通过反复调整参数和调试优化器,优化卷积神经网络算法的收敛速度,提高学习和运算效率,完善卷积神经网络对木材缺陷轮廓的提取,在降低运算复杂度的同时,提高其精度,具有良好的鲁棒性。相比径向基函数(RBF)神经网络、向后反馈-径向基函数(BP-RBF)混合神经网络和卷积神经网络,本算法对木材缺陷具有更好的识别效果,其识别准确率达到98.6%左右,且识别时间相对更短。     

Applying multifractal spectrum combined with fractal discrete Brownian motion model to wood defects recognition

Wood nondestructive testing technology is a new and multidisciplinary industry scientific research. It has attained fast development and achievements in recent years. X-ray computed tomography (CT) scanning technology is a kind of wood nondestructive testing technology in practice. CT scanning technology has been applied to the detection of internal defects in the logs for the purpose of obtaining prior information, which can be used to reach better wood sawing decision. Fractal geometry and its extension multifractal are used for describing, modeling, analyzing, and processing of different complex shapes and images. A method in CT image edge detection using multifractal theory combined with fractal Brownian motion is applied in the paper. First, its multifractal spectrum is estimated. Then, different types of pixels are classified by the spectrum; they are smoothing edge points and singular edge points. From the images processed by multifractal spectrum theory and compared with each image by different spectrum values, it can be seen that the larger the range of threshold is set, the more exact the edge can be detected. The paper provides a new method to recognize the defect information and to saw it in the condition of nondestructive wood.     

基于人工神经网络的原木CT图像缺陷识别

以欧洲白蜡为例,利用训练好的神经网络识别原木CT图像中的各种木材缺陷.不同隐蔽层节点的神经网络可以正确地识别树皮、节子、腐朽和无疵木材;但是对于细小裂纹尚还不能准确识别.计算机快速、自动识别图像中的各种缺陷,有利于实现最优化的锯切方案.     

遗传神经网络在家具板材优化下料问题中的应用

阐明了遗传算法和神经元网络结合的必要性和可行性 ,通过用遗传算法训练前馈神经网络权值解决家具板材优化下料问题并给出实验结果。     

Comparison analysis of nonlinear segmentation methods for selected wood features in sugar maple

Two classification methods, a feed-forward neural network and a fuzzy logic algorithm, were used for the automatic identification of CT images for selected wood features in sugar maple, one of the most important hardwoods in eastern Canada. Three wood characteristics were selected for automatic identification together with the background as a default. Local features, such as position and local pixel values were used as the neural networks inputs. The fuzzy sets consisted of four different possible pixel values and four possible distances from the center of the log. The fuzzy method used in this study was of the Mamdani type. Five sugar maple logs were randomly selected for this study. One of the logs is used for the training of the neural network and the others for validation and comparison. The structure of the neural network was optimized and was used for the segmentation of the other logs. An efficiency function, consisting of the number of pixels correctly labeled, was defined for the evaluation of the segmentation process. This study shows that a segmentation based on a fuzzy method has better capabilities for generalization than one based on a feed-forward method.     

基于BP神经网络的森林植被遥感分类研究

如何解决多类别图像的识别并满足一定的精度,是遥感图像研究中的一个关键问题,具有十分重要的意义。本文使用Landsat7ETM+遥感数据和森林资源分布图等地理辅助数据,用BP神经网络方法对森林植被进行了分类,并与最大似然法的分类结果进行精度比较分析。结果表明地理辅助数据参与的BP神经网络用于森林植被遥感图像分类其效果是较好的,是一种有效的图像分类方法。