基于BP神经网络的木材表面颜色特征分类的研究

本文针对木材表面颜色自动分类难题,在 RGB 颜色空间提取木材图像的颜色矩作为颜色特征参数,利用 BP 神经网络对特征参数进行分类,通过输入层、输出层和隐含层的设计,传递函数的选择,确定最终网络结构。实验结果表明,分类正确率达到98%,验证了本文提取的特征参数的有效性。     

基于直方图的木材表面颜色分类研究

颜色直方图是图像检索技术中一种有效的颜色表达方式，具有位移、旋转不变的特性。然而木材表面颜色较为单一，采用通常的量化方案不能达到很好的分类识别效果。提出了采用HSV颜色空间三个独立分量的直方图统计特征表达木材表面颜色信息的方法，利用色调、饱和度和亮度三分量之间的相互独立性，提取了各分量的直方图特征。最后利用BP神经网络对木材样本库的图像进行了分类仿真，其结果验证了特征的有效性。     

基于直方图和颜色矩方法的木材表面颜色特征的表达

简述了彩色图像处理中颜色空间的选择,分别用直方图和颜色矩方法表达木材表面的颜色特征。从H分量直方图可知,H分量在色彩上有较好的分类性,反映出木材彩色特征的变化。用颜色矩特征值作为BP神经网络的输入,对东北常见树种按颜色进行了分类,分级正确率达到了96.7%。     

基于空间灰度共生矩阵木材纹理分类识别的研究

以10种木材纹理样本为对象,研究了木材纹理参数体系的建立方法,并进行了分类识别的仿真实验。首先,针对木材纹理特点并结合类别可分性判据,构造了适于描述木材的空间灰度共生矩阵,并在此基础上提取了木材的11个纹理特征参数。其次,借助相关性分析对参数进行了特征选择,进而建立了能直接与人的感官对应的木材纹理参数体系。最后,利用 BP 神经网络分类器对木材样本进行了分类识别研究,识别率为87.50%,验证了参数体系的有效性,表明用本文提出的纹理参数体系对木材进行分类识别是可行的。     

基于BP神经网络木材纹理分类的研究

应用 BP 神经网络对常见的10种木材纹理进行了分类研究,获得了比较满意的效果。首先,应用灰度共生矩阵提取了木材的纹理特征参数;其次,在此特征参数体系下,应用 BP 神经网络对木材纹理进行了分类研究,识别率达89%。     

基于融合彩色特征木材表面颜色分类的研究

针对木材表面颜色自动分类的难题,在RGB颜色空间,将R、G、B三个颜色矩阵融合成一个特征矩阵,再对这个特征矩阵提取颜色三阶矩参数作为木材表面颜色分类的特征参数,设计了适合木材表面颜色分类的BP神经网络分类器,分类识别率达到98.67%,验证了提取特征参数的有效性。     

基于时延神经网络的木材干燥模型辨识的研究

木材干燥是一个复杂、多变的过程，传统的系统辨识方法很难建立一个理想的木材干燥模型。利用时延神经网络的特点，本文提出了基于时延神经网络的木材干燥模型辨识方法，并给出了辨识网络和结构。对辨识得到模型的仿真结果表明，利用时延神经网络所建立的模型是可靠且有效的。     

基于BP神经网络的针阔混交林TM遥感图像自动分类技术研究

在对标准BP神经网络试验分析的基础上，通过输入矢量归一化处理、主成分分析、增加验证集、改进训练学习算法、扩大隐层和输出层规模等措施，对BP神经网络自动分类系统进行改进；利用改进后的BP系统对吉林省汪清林业局的典型针阔混交林TM遥感图像进行辩识、分类试验研究。结果表明：改进后的BP网络分类系统自动分类精度提高了19．14％，比传统无监督自动分类精度提高8．55％，达到了区分森林类型的分类要求。研究还显示了该改进系统应用于针阔混交林TM遥感图像自动分类识别的精度随网络规模增大而提高。     

基于L*a*b*颜色空间对木材分类的研究

L*a*b*均匀颜色空间具有等距性和色差高分辨力的特点,非常适合色差较小情况下的颜色测量和比较.木材材色分布范围较窄,利用L*a*b*颜色空间中的颜色特征表示木材表面颜色,有利于木材材色之间的比较和划分.基于L*a*b*颜色空间,提取了东北常见五种树种木材图像的颜色特征进行分类研究,通过仿真试验得到了满意的分类结果.     

基于CNN的木材内部CT图像缺陷辨识

【目的】为获取木材内部构造形态,提高木材内部缺陷识别率,依据获得的计算机断层扫描图像,提出一种基于卷积神经网络(CNN)的木材内部缺陷辨识方法,以实现木材的高效化自动分类。【方法】首先,利用课题组自行开发的计算机断层扫描系统,采集样本木材内部CT图像800幅;然后,对样本图像进行处理,随机选取700幅原始样本图像,从中截取出单个缺陷区域和正常木材断层区域样本图像20 000幅,并利用图像增强等算法将数据集扩充到70 000幅,标准化图像大小为28×28像素,分为正常、裂纹、虫眼和节子图像共4类,取60 000幅图像作为训练集,10 000幅图像作为测试集,剩余的100幅原始样本图像用于试验验证。【结果】通过60 000幅图像来训练网络模型,对测试集10 000幅图像进行分类,分类正确率达99.3%;利用训练得到的网络模型对100幅原始样本图像进行验证,平均分类正确率为95.87%。【结论】基于卷积神经网络的木材内部CT图像缺陷辨识算法,克服了传统识别方法图像预处理繁琐、训练方法复杂、训练参数过多、耗时过多等问题,具有精度高、复杂度小、鲁棒性较好等优点,且辨识正确率和辨识时间都比现行常规算法精准并用时短,是一种无损、高效、准确的辨识分类方法。     

基于BP神经网络的中国进口俄罗斯木材贸易额预测

采用灰色关联度分析法筛选对中国进口俄罗斯木材贸易额影响较大的5个因子,运用影响因子及木材贸易进口额构建BP神经网络模型,利用GM(1,1)模型预测影响因子值,将其代入训练好的BP网络模型中对中国进口俄罗斯木材贸易额进行预测。预测结果表明,中俄木材贸易仍具有良好的发展前景。     

Application of single neuron adaptive PID controller during the process of timber drying

The paper presents a method of using single neuron adaptive PID control for adjusting system or servo system to implement timber drying process control, which combines the thought of parameter adaptive PID control and the character of neural network on exactly describing nonlinear and uncertainty dynamic process organically. The method implements functions of adaptive and self-leaming by adjusting weighting parameters. Adaptive neural network can make some output trail given hoping value to decouple in static state. The simulation result indicates the validity, veracity and robustness of the method used in the timber drying process     

Mechanical stress grading of tropical timbers without regard to species

Some reports have shown that for single species the correlation between modulus of elasticity (MOE) and modulus of rupture (MOR) in bending is quite high. Tropical timbers consist of hundreds of species that are difficult to identify. This report deals with the mechanical stress grading of tropical timber regardless of species. Nine timber species or groups of species with a total number of 1094 pieces measuring 60 × 120 × 3000 mm, were tested in static bending. The MOE was measured flat wise, while MOR was tested edge wise. Statistical analysis of linear regression with a dummy model and analysis of covariance were used to analyze the role of MOE and the effect of species on prediction of MOR. The analysis showed that using MOE as a single predictor caused under/overestimation for one or more species and/or groups of species. The accuracy of prediction would be increased with species identification. An allowable stress and reference resistance for species and/or groups of species were provided to compare with the prediction of strength through timber grading. The timber strength class for species and/or groups of species was also established to support the application of mechanical timber grading.     

Analysis of determination methods for characteristic timber properties as related to growth area and grade yield

The origin of the raw material is a key aspect for strength grading of timber. Large grading areas are favored by the sawmilling industry as they require less effort in handling and documentation during the production process. However, large growth areas can also cause problems, as too high mechanical properties can be declared or yields may become uneconomical. The presented study presents a method that should allow for timber from different countries to be combined into a single grading area. Additionally, the influence on the yield for guaranteeing timber properties for differently defined populations is analysed. In this process, a number of available calculation methods for characteristic values for modulus of rupture, modulus of elasticity, and density are considered as the determination method also influences the final yield. Non-destructive and destructive test data from 8487 spruce specimens from Europe tested in bending or tension are the basis for the presented study. Based on the grading results the presented method is able to simply identify countries that may be combined. The definition of pan-European grading areas seems problematic if characteristic timber properties need to be guaranteed separately for each individual country as it may result in a severe drop in yield. However, checking timber properties only for the European population is unsatisfying as calculated timber properties considerably vary depending on the origin. As for the calculation method, the preferred method itself seems to have less impact on bending class assignments then on tension class assignments.     

基于过程神经网络的木材生长轮密度预测

提出一种基于过程神经网络的木材生长轮密度长期预测方法.本方法利用输入输出均为时变函数的过程神经网络输出为时变函数的特点,将原始数据拟合为输入函数并表示为一组正交基的展开形式后,使用混合遗传算法训练过程神经网络,得到过程神经网络的输出函数,以此实现木材生长轮密度的一次多步长期预测,通过与传统时间序列预测方法比较,预测精度得到显著提高,并为时间序列长期预测问题提供新方法.     

用材林林地定级与估价技术研究

在构建福建省用材林林地定级评价指标体系(经济指标、地利指标和立地指标3大类10个指标)的基础上,采用层次分析法确定各评价指标的权重,以小班为评价单元计算总分值,采用聚类分析法划分用材林小班林地级别。在定级的基础上,提出了两种用材林林地估价方法,即基准地价修正法和地价模型法。     

Grading and mechanical characterization of small-diameter round chestnut (Castanea sativa Mill.) timber from thinning operations

ABSTRACT

In forestry, thinning operations result in the extraction of young trees with small dimensions. The evaluation of the potential end use of these small-diameter logs (currently mainly used as poles or fence posts) for added-value products such as structural timber is of considerable economic and industrial interests. In the present work, 216 pieces of small-diameter logs of chestnut timber obtained from thinning operations were evaluated in order to determine their mechanical properties and assess various visual or non-destructive grading systems. The two visual standards evaluated (EN 1927 and DIN 4074) were ineffective in grading according to mechanical properties. On the other hand, a grading system based on a non-destructive measurement (acoustic wave velocity) resulted in better classification by structural quality. The characteristic values of the small-diameter round chestnut timber, determined according to the standards EN 384 and EN 338, achieved similar values as standard rectangular sawn timber with respect to modulus of elasticity and density, and higher values for bending strength.     

常用竹材使用性能分级评价研究

采用竹材直径、竹壁厚度、竹材天然耐久性能和可处理性能为指标,将每项指标设定四个等级及相应等级下的参考分值,对17个竹种在作材用的功能上进行使用分级,并将分级结果与竹的传统用途进行比较,结果表明这种分类方法有一定的可取性,就是得分高的竹材较适合作材用竹。     

泡桐材色变异规律的研究

泡桐材色在树干或侧枝圆盘年龄上从里向外呈现“深-浅—深”的径向变化形式。在不同高度上,泡桐材色随树高增加逐渐变浅,在靠近梢部时又有变深的趋势;泡桐枝材的材色变化规律与干材相似,可以用6年生的成熟枝材来评估干材材色。     

Multivariate product adapted grading of Scots pine sawn timber for an industrial customer,part 1: Method development

ABSTRACT

Rule-based automatic grading (RBAG) of sawn timber is a common type of sorting system used in sawmills, which is intricate to customise for specific customers. This study further develops an automatic grading method to grade sawn timber according to a customer's resulting product quality. A sawmill's automatic sorting system used cameras to scan the 308 planks included in the study. Each plank was split at a planing mill into three boards, each planed, milled, and manually graded as desirable or not. The plank grade was correlated by multivariate partial least squares regression to aggregated variables, created from the sorting system's measurements at the sawmill. Grading models were trained and tested independently using 5-fold cross-validation to evaluate the grading accuracy of the holistic-subjective automatic grading (HSAG), and compared with a re-substitution test. Results showed that using the HSAG method at the sawmill graded on average 74% of planks correctly, while 83% of desirable planks were correctly identified. Results implied that a sawmill sorting station could grade planks according to a customer's product quality grade with similar accuracy to HSAG conforming with manual grading of standardised sorting classes, even when the customer is processing the planks further.