关于古建筑图像中破损点优化提取仿真
发布时间：2019-03-08 14:44
【摘要】：对古建筑图像中对破损点的优化提取,对古建筑后续原貌恢复具有重要意义。对图像破损点的提取,需要获得破损点分量特征信息,计算图像的破损点超像素级视觉特征,完成图像中破损点优化提取。传统方法保留破损图像关键区域特征点,删除古建筑图像背景区域特征点,但忽略了计算图像的破损点超像素级视觉特征,导致提取精度偏低。提出基于小波阈值自适应修正的古建筑图像破损点提取方法。基于RAC约束的两步法来标定系统,对图像进行灰度化转换,进行图像的边缘检测和小波降噪处理,对小波降噪提纯后的古建筑图像进行破损点深度超像素特征分割,获得破损点分量特征信息,计算图像的破损点向量量化区域的超像素级视觉特征,实现破损图像破损点提取优化。仿真证明,所提方法从视觉对比和量化分析的角度实现了破损区域的轮廓分割与检测,破损点提取效果优越。
[Abstract]:The optimum extraction of the damage points in the image of ancient buildings is of great significance to the restoration of the original features of the ancient buildings. In order to extract the broken points of images, we need to obtain the feature information of the components of the broken points, calculate the hyperpixel-level visual features of the broken points of the image, and complete the optimal extraction of the broken points in the image. The traditional method preserves the key region feature points of damaged images and deletes the feature points of the background region of ancient building images, but neglects the hyperpixel-level visual features of the damaged points of the calculated images, which leads to the low extraction accuracy. Based on wavelet threshold adaptive correction, a method for extracting damage points of ancient building image is proposed. A two-step method based on RAC constraint is used to calibrate the system. The image is grayscale transformed, the image edge detection and wavelet denoising processing are carried out, and the wavelet de-noising purified ancient building image is segmented by the ultra-pixel feature of the depth of damage point, which is based on the wavelet de-noising and purification. The feature information of the broken point component is obtained, and the hyperpixel level visual feature of the vector quantization region of the broken point is calculated to realize the optimization of the broken point extraction of the damaged image. Simulation results show that the proposed method achieves the contour segmentation and detection of damaged areas from the visual comparison and quantitative analysis, and the extraction effect of damage points is superior.
【作者单位】： 长春大学旅游学院;
【基金】：吉林省教育厅“十三五”社会科学研究规划项目(吉教科文合字JJKH20171024SK)
【分类号】：TP391.41;TU-87
本文编号：2436911
[Abstract]:The optimum extraction of the damage points in the image of ancient buildings is of great significance to the restoration of the original features of the ancient buildings. In order to extract the broken points of images, we need to obtain the feature information of the components of the broken points, calculate the hyperpixel-level visual features of the broken points of the image, and complete the optimal extraction of the broken points in the image. The traditional method preserves the key region feature points of damaged images and deletes the feature points of the background region of ancient building images, but neglects the hyperpixel-level visual features of the damaged points of the calculated images, which leads to the low extraction accuracy. Based on wavelet threshold adaptive correction, a method for extracting damage points of ancient building image is proposed. A two-step method based on RAC constraint is used to calibrate the system. The image is grayscale transformed, the image edge detection and wavelet denoising processing are carried out, and the wavelet de-noising purified ancient building image is segmented by the ultra-pixel feature of the depth of damage point, which is based on the wavelet de-noising and purification. The feature information of the broken point component is obtained, and the hyperpixel level visual feature of the vector quantization region of the broken point is calculated to realize the optimization of the broken point extraction of the damaged image. Simulation results show that the proposed method achieves the contour segmentation and detection of damaged areas from the visual comparison and quantitative analysis, and the extraction effect of damage points is superior.
【作者单位】： 长春大学旅游学院;
【基金】：吉林省教育厅“十三五”社会科学研究规划项目(吉教科文合字JJKH20171024SK)
【分类号】：TP391.41;TU-87
【相似文献】
相关期刊论文 前1条
1 冯引安;李引贤;侯瑞宁;;图像型火灾空间定位研究[J];系统仿真技术;2009年03期
相关博士学位论文 前1条
1 李森;火灾初期建筑内图像清晰化及人员检测技术研究[D];中国科学技术大学;2014年
，本文编号：2436911
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2436911.html
