Wood defect classification based on image analysis and support vector machines

This paper addresses the issue of automatic wood defect classification. A tree-structure support vector machine (SVM) is proposed to classify four types of wood knots by using images captured from lumber boards. Simple and effective features are proposed and extracted by partitioning the knot images into three distinct areas, followed by utilizing a novel order statistic filter to yield an average pseudo color feature in each area. Excellent results have been obtained for the proposed SVM classifier that is trained by 800 wood knot images. Performance evaluation has shown that the proposed SVM classifier resulted in an average classification rate of 96.5% and false alarm rate of 2.25% over 400 test knot images. Future work will include more extensive tests on large data set and the extension of knot types.     

基于空间灰度共生矩阵的木材纹理特征提取

讨论了生成步长、生成方向、图像灰度级对灰度共生矩阵特征参数的影响，在对国内50个树种计算分析的基础上，得出适于描述木材纹理的灰度共生矩阵构造方法。结合了灰度共生矩阵特征参数间的相关性及木材纹理自身的特征提取出了一套表征木材纹理的特征参数。     

一种新的针叶材自动识别方法

提出通过横切面显微图像对针叶材树种进行计算机识别的方法。该方法通过提取图像的PCA特征,生成特征树,然后采用SVM对样本进行分类。使用8种针叶材,每种12个样本,并采用留一交叉验证,对图像的分割方法、最近邻与SVM分类算法和不同范数距离下的识别效果进行试验。结果表明通过部分木材微观的纹理结构进行木材识别的可能性。     

基于图像纹理特征的木材树种识别

利用木材图像的颜色、灰度、纹理等内容实现树种的相似性匹配检索,提取色调、饱和度、亮度、对比度、二阶角矩、方差和、长行程加重因子、分形维数、小波水平能量比重共9个特征参数,依据最大相似性数学原理,基于最小差值参数判别法和综合特征阈值法来检索样本.结果显示:基于图像纹理特征能够实现木材树种的检索和识别,综合特征阈值法的检索正确率与唯一性通常要好于最小差值判别法;但当被检索样本图像的纹理较弱或不呈现纹理特征时,检索结果的唯一性并不理想.综合而言,基于图像纹理特征最大相似性的木材树种检索识别较易实现,是一种值得继续发展和应用推广的木材树种识别方法.     

小波变换的木材纹理在线分选

以180幅木材样本图片为对象,研究以小波变换方法提取特征参数,分析几种小波基的特点和性质,最终以对称性为依据,选择使用sym4小波对图像进行二级小波分解,可以得到一级水平细节HL1、垂直细节LH1、对角细节HH1,二级的近似LL2、水平细节HL2、垂直细节LH2、对角细节HH2共7个子图,提取整幅图像的熵和每个子图小波系数的均值及标准差作为特征参数。将木材纹理按照直纹、抛物线和乱纹3种纹理的分类标准,以BP神经网络作为分类器进行了木材纹理分类的验证,并与灰度共生矩阵的方法进行了对比。试验表明:采用小波变换的方法对木材纹理特征进行描述,不但提高了分类的准确率,重要的是缩短了运算时间,可以达到在线监测的要求。     

基于SVM的多特征融合的天牛图像识别

天牛虫的图像特征提取对于天牛虫灾害的防治和监控有很重要的意义,针对目前图像识别在这个领域的应用中存在的问题,提出进行图像的特征提取和识别的方法:首先,对天牛虫图像进行二维小波变换分解,用低频图像进行特征提取,减少噪声的同时可以提高识别的准确率;然后,提取低频图像的SIFT(尺度不变特征变换)特征向量集,解决大范围的仿射失真、3维视角的改变、噪音的增加和光线的改变等造成的影响;为了提高复杂光照条件下的图像识别率,引入了颜色特征,将图像从RGB转换到HSV空间,提取图像的颜色矩作为颜色特征向量;最后,将所提取的特征作为SVM分类器的训练样本集,进而对目标图像进行识别,实验结果表明,提出的方法能够得到较好的识别效果。     

基于高分二号遥感数据的森林自然度等级划分

以地面样点为基础的森林自然度评价方法很难获得区域范围森林自然度等级,针对该问题,提出了利用高分遥感卫星影像数据,划分区域范围森林自然度等级的方法。以湖北竹山县九华山林场为试验区域,在选取研究区典型样地的基础上,结合高分二号(GF-2)遥感影像数据的特点,从GF-2影像上提取遥感光谱、纹理等特征并结合地形特征,采用随机森林算法在大尺度范围对九华山林场森林自然度等级进行分类研究。结果发现:以GF-2数据为基础提取的植被指数、光谱、纹理等特征与地形特征结合,采用随机森林算法可较好地划分森林自然度等级,总体分类精度高达93.97%,Kappa系数为0.91。对森林自然度等级影响最重要的6个特征因子为高程、坡向、坡度、纹理均值、光谱主成分变化分量和归一化植被指数(NDVI)。结果表明,基于遥感影像提取的特征和地形特征结合进行森林自然度等级划分的研究方法具有可行性,为大面积区域的森林自然度等级划分奠定基础。     

基于线性谱聚类的林地图像中枯死树监测

【目的】将基于线性谱聚类超像素的方法应用在森林病虫害防治领域,可智能监测无人机森林虫害图像中的枯死树,为森林有害生物的监测工作提供技术支撑。【方法】分别以湖北省受松材线虫与辽宁省受红脂大小蠹侵害的松林无人机图像为试验数据,首先使用线性谱聚类超像素分割算法将图像划分为多个超像素;然后基于枯死树木的颜色特征,初步提取可能为枯死树的超像素区域;最后基于枯死树木与其他干扰地物具有不同的纹理特征,计算超像素的区域密度和缝隙量,利用支持向量机对初步提取的超像素进行分类,从而检测出图像中的枯死树。【结果】基于线性谱聚类超像素和支持向量机的枯死树监测方法可有效排除与枯死树木颜色相近的其他干扰地物,较准确地提取出枯死树木。使用该方法与基于植被颜色指数的阈值分割方法、基于简单线性迭代聚类超像素和随机森林的方法,对35幅受灾松林无人机图像进行试验,并选用交并比、虚警率和漏检率3个评价指标对3种方法进行定量对比分析。结果表明,基于线性谱聚类超像素的方法监测出的枯死树区域最精确,其监测结果与人工检测结果的交并比均值大于58%,且虚警率和漏检率均优于另外2种方法。【结论】基于线性谱聚类超像素的枯死树监测方法能实现松林中枯死树的快速、准确检测及定位。     

支持向量机参数优化方法在木质粉尘火花探测中的应用

为提高木质粉尘火花检测的准确性,利用基于支持向量机(SVM)的物质分类方法检测木质粉尘火花。选取马尾松和杨木粉尘为研究对象,将两种粉尘分组点燃试验,获取火花和灰分的高光谱图像,提取感兴趣区域(region of interest,ROI)内的发光度数据进行预处理。利用感兴趣区域内的数据建立SVM分类模型,分别利用网格搜索法(GS)、遗传算法(GA)以及粒子群算法(PSO)对两类树种的SVM分类模型进行参数优化,并将三种参数优选方法的分类预测准确率进行对比。结果表明,三种优化方法均能够很好地检测两种树种的木粉火花,其中网格搜索法检测准确率明显高于其余两种,更适于木质粉尘火花探测,这为人造板生产过程中能够高效检测木质粉尘火花提供了一定的理论依据。     

Soft measurement of wood defects based on LDA feature fusion and compressed sensor images

We proposed a detection method for wood defects based on linear discriminant analysis(LDA) and the use of compressed sensor images. Wood surface images were captured, using a camera Oscar F810C IRF camera,and then the image segmentation was performed, and the defect features were extracted from wood board images. To reduce the processing time, LDA algorithm was used to integrate these features and reduce their dimensions. Features after fusion were used to construct a data dictionary and a compressed sensor was designed to recognize the wood defects types. Of the three major defect types, 50 images live knots, dead knots, and cracks were used to test the effects of this method. The average time for feature fusion and classification was 0.446 ms with the classification accuracy of 94%.     

基于分形理论的木材纹理特征研究

介绍了一种利用自相关函数来估算图像分形维数的方法，并将其应用到木材的纹理分类检测中。实验表明，分形维数值直接反映了木材纹理的粗糙程度，可定性地作为描述木材纹理粗糙度的一种度量。     

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.     

基于残差网络的遥感影像松材线虫病自动识别

松材线虫病是针对松树的特殊疾病,具有前期发病特征隐蔽、传播范围广、致病速度快的特点,因此面对林业病虫害问题,对受灾区域染病树木进行高效识别和分类,监测其他区域的树木生长情况,并且根据受灾情况确定损失,进行保险理赔是十分重要的.针对染病树木识别准确率低、识别速度慢的问题,本研究利用遥感影像和残差网络相结合的方法,并对残差...     

Automated classification of wood transverse cross-section micro-imagery from 77 commercial Central-African timber species

?Key message

Pattern recognition has become an important tool to aid in the identification and classification of timber species. In this context, the focus of this work is the classification of wood species using texture characteristics of transverse cross-section images obtained by microscopy. The results show that this approach is robust and promising.

?Context

Considering the lack of automated methods for wood species classification, machine vision based on pattern recognition might offer a feasible and attractive solution because it is less dependent on expert knowledge, while existing databases containing high-quality microscopy images can be exploited.

?Aims

This work focuses on the automated classification of 1221 micro-images originating from 77 commercial timber species from the Democratic Republic of Congo.

?Methods

Microscopic images of transverse cross-sections of all wood species are taken in a standardized way using a magnification of 25 ×. The images are represented as texture feature vectors extracted using local phase quantization or local binary patterns and classified by a nearest neighbor classifier according to a triplet of labels (species, genus, family).

?Results

The classification combining both local phase quantization and linear discriminant analysis results in an average success rate of approximately 88% at species level, 89% at genus level and 90% at family level. The success rate of the classification method is remarkably high. More than 50% of the species are never misclassified or only once. The success rate is increasing from the species, over the genus to the family level. Quite often, pattern recognition results can be explained anatomically. Species with a high success rate show diagnostic features in the images used, whereas species with a low success rate often have distinctive anatomical features at other microscopic magnifications or orientations than those used in our approach.

?Conclusion

This work demonstrates the potential of a semi-automated classification by resorting to pattern recognition. Semi-automated systems like this could become valuable tools complementing conventional wood identification.

     

层次支持向量机在色木孔洞缺陷检测中的应用

针对现有木材无损检测中存在的问题,提出根据木材的声脉冲响应特点,通过自制的声波信号采集装置提取含有孔洞缺陷木材的声脉冲响应信号,再分别从时域和频域对信号进行处理,提取相关的统计信息作为识别特征,再输入到层次支持向量机（SVM）中进行识别的方法.结果表明,该方法对色木孔洞位置的识别准确率在95％以上,具有需构造的SVM分类器数量少、不存在不可识别域、训练和识别速度快的优点.对基于支持向量机的木材孔洞缺陷识别进行探讨,并对其有效性进行验证.     

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。     

Study of visual evaluations for wood flooring applying fuzzy logic

This study was aimed to investigate the visual evaluations of wood flooring. The selected 12 species of wood flooring were simulated by computer and combined with the 6 sets of visual image adjectives to design a questionnaire and survey the visual evaluations of consumers. Triangular fuzzy number of fuzzy theory was employed to obtain the scores of the 12 species of wood flooring in the 6 visual image evaluations. The results showed that the differences between different species of wood flooring in the visual evaluation were less significant in terms of “classical and primitive”, “durable and practical”, and “natural and original”. However, the differences were more significant in terms of “Elegant and Soft”, “close and comfortable”, and “tender and amiable”. Furthermore, six groups with relative overall visual images were induced from the comparison of wood flooring by qualitative classification. For example, the overall visual images of Quercus rubra, Acer saccharum and Quercus alba were similar, and they were more “Elegant and Soft”,“durable and practical”,“close and comfortable”, and “tender and amiable”. The overall visual images of Carya ovata, Pinus rigida and Castanea sativa were similar, and they tended to be “classical and primitive” and “natural and original”.     

东南亚阔叶树材数据库查询系统

国内研究或鉴定东南亚木材较早的为中国林业科学研究院木材工业研究所。已发表的有陈嘉宝的《十二种柬埔寨重要工业木材的粗视特征及其物理力学性质》一文;1989年又编译了《马来西亚商用木材性质和用途》一书(商用木材74种)。1988年,安徽农学院林产工业研究所卫广扬等主编的《东南亚木材——识别及用途》问世,以近百种进口的马来西亚和菲律宾原木作为木材构造和材性试验的主要材料,共记载针、阔叶树材90余种;因系进口原木,所以不少没有种名;供作材性试验的木材也多为1号标本。以上这些资料各有侧重,有的只有粗视构造,有的虽有显微构造记载但不少没有种名,不利于木材识别;有的虽然作了木材物理力学试验,但试材太少,代表性不强;就已研究的树种数量看也较少,不能满足广大用材者的需求。     

基于多特征提取与注意力机制深度学习的高分辨率影像松材线虫病树识别

为了提高松材线虫病树的监测效率,减少其对林业生产造成的损失,提出一种基于多特征提取与注意力机制深度学习的高分辨率影像松材线虫病树识别方法.该方法首先在高分辨率遥感影像上提取松材线虫病树的光谱特征、空间特征等多特征,然后进行Relief特征选择算法,取特征权重前8个特征进行病树识别,发现选择差值植被指数DVI(diffe...     

Texture analysis of stereograms of diffuse-porous hardwood: identification of wood species used in Tripitaka Koreana

Tripitaka Koreana is a collection of over 80,000 Buddhist texts carved on wooden blocks. In this study, we investigated whether six hardwood species used as blocks could be recognized by image recognition. An image data set comprising stereograms in transverse section was acquired at 10×?magnification. After auto-rotation, cropping, and filtering processes, the data set was analyzed by an image recognition system, which comprised a gray-level co-occurrence matrix method for feature extraction and a weighted neighbor distance algorithm for classification. The estimated accuracy obtained by leave-one-out cross-validation was up to 100% after optimizing the pretreatments and parameters, thereby indicating that the proposed system may be useful for the non-destructive analysis of all wooden carvings. We also examined the specific anatomical features represented by textures in the images. Many of the texture features were apparently related to the density of vessels, and others were associated with the ray intervals. However, some anatomical features that are helpful for visual inspection were ignored by the proposed system despite its perfect accuracy. In addition to the high analytical accuracy of this system, a deeper understanding of the relationships between the calculated and actual features is essential for the further development of automated recognition.