基于图像处理提高木材识别准确性的新方法

发布时间：2018-08-27 07:44

【摘要】：为了提高机器识别木材的准确性,从木材图像预处理的角度出发,以复原木材图像纹理细节为目标,提出了基于SCN-MSE的木材图像超分辨率重建方法。将高分辨率图像经离散小波变换,把4个子带中相应位置、大小相同的碎片组成模块,再利用局部二值模式的邻域强度建立训练数据库;将低分辨率图像中的低频子带碎片,与数据库同类模块中的低频碎片进行比较,通过去除领域中心的均方差,寻找相似度最大的高频碎片,保留低分辨率图像的低频子带;再经小波逆变换得到超分辨率图像。选取樟子松及其树皮为识别对象,将基于SCN-MSE的超分辨率重建图像与经过传统预处理图像,利用SVM多项式核函数进行识别。识别结果表明,本研究提出的方法提高了樟子松及其树皮的识别率。
[Abstract]:In order to improve the accuracy of wood recognition by machine, a super-resolution reconstruction method of wood image based on SCN-MSE is proposed, which aims at restoring the texture details of wood image from the point of view of wood image preprocessing. The high resolution image is transformed by discrete wavelet transform, the corresponding position of the four sub-bands and the same size fragments are formed into the module, and then the training database is established by using the neighborhood strength of the local binary mode, and the low frequency sub-band fragments in the low-resolution image are taken into account. Compared with the low-frequency fragments in the same module of the database, by removing the mean square error of the domain center, the high frequency fragments with the largest similarity are found, and the low-frequency sub-bands of the low-resolution images are preserved, and then the super-resolution images are obtained by inverse wavelet transform. Selecting Pinus sylvestris var. mongolica and its bark as the recognition object, the super-resolution reconstruction image based on SCN-MSE and the traditional preprocessed image are used to recognize the image by using the kernel function of SVM polynomial. The recognition results show that the proposed method improves the recognition rate of Pinus sylvestris var. mongolica and its bark.
【作者单位】：内蒙古林业厅退耕还林工程管理办公室;内蒙古农业大学材料科学与艺术设计学院;
【基金】：国家自然科学基金(31460168) 内蒙古农业大学博士启动基金(BJ09-29)
【分类号】：S781.1;TP391.41

【相似文献】