石鑫华视觉网-中科院自动化所 计算机视觉课件

本程序用于对工件图像的预处理，边缘提取等。将改进遗传算法和Hausdorff距离结合用于工件识别。采用canny边缘为匹配特征，将修正的Hausdorff距离作为相似性度量准则，并应用遗传算法进行最佳匹配的快速搜索，成功实现了目标物体的匹配识别(This procedure used for the workpiece image preprocessing, edge extraction and so on. Will improve the genetic algorithm and Hausdorff distance combination for workpiece identification. Using canny edge for matching characteristics, will be amended Hausdorff distance as similarity measurement criteria and the application of genetic algorithm is the best match of the Quick Search, the successful achievement of the target object matching identification)

说明：  Donoho提出的小波阀值去噪的基本思想是将信号通过小波变换（采用Mallat算法）后，信号产生的小波系数含有信号的重要信息，将信号经小波分解后小波系数较大，噪声的小波系数较小，并且噪声的小波系数要小于信号的小波系数，通过选取一个合适的阀值，大于阀值的小波系数被认为是有信号产生的，应予以保留，小于阀值的则认为是噪声产生的，置为零从而达到去噪的目的。尺度不变特征变换（Scale-invariant feature transform，SIFT），是用于图像处理领域的一种描述。这种描述具有尺度不变性，可在图像中检测出关键点。(The basic idea of wavelet threshold de-noising proposed by Donoho is that the wavelet coefficients generated by the signal after passing through the wavelet transform (Mallat algorithm) contain the important information of the signal. After the wavelet decomposition, the wavelet coefficients of the signal are larger, the wavelet coefficients of the noise are smaller, and the wavelet coefficients of the noise are smaller than the wavelet coefficients of the signal. By selecting a suitable threshold, the wavelet coefficients of the noise are larger than the valve The wavelet coefficients of values are considered to be generated by signals, which should be preserved. The ones less than the threshold are considered to be generated by noise, which is set to zero to achieve the purpose of denoising.)

说明：  一、课题介绍 本设计为基于MATLAB的火焰烟雾火灾检测系统。传统的采用颜色的方法，误识别大，局限性强。结合火焰是实时动态跳跃的，采用面积增长率，角点和圆形度三个维度相结合的方式判断是否有火焰。该设计测试对象为视频，通过下一帧和上一帧的差异发现是否有火情，并可发出语音报警。本设计带有一个人机交互式GUI界面，界面友好。是个不错的bishe选题。 二、算法流程 结合火焰的面积增长率，角点和圆形度三个维度综合判断。并且得出每帧图像火焰部分的该三个参数，实时显示在GUI上。 三、GUI界面设计(This design is based on MATLAB flame smoke fire detection system. The traditional method of using color has large error recognition and strong limitation. Considering that the flame is real-time and dynamic jumping, the area growth rate, corner and roundness are combined to determine whether there is flame. The test object of the design is video. Through the difference between the next frame and the previous frame, whether there is fire is found and the voice alarm can be sent out. This design has a personal computer interactive GUI interface, the interface is friendly. It's a good choice for BiShe.)

目标物与摄像机距离已知的情况下，计算白天大气能见度距离。(The distance between the object and the camera is known to calculate the visibility distance in daytime.)

计算BBO晶体一类匹配时随非共线角变化的相位匹配角(Calculating the phase matching angle of BBO crystal with non-collinear angle)

数字水印的图像处理方法，很好，做图像水印的同行可以参考(Digital watermark image processing methods, well, do their counterparts in the image watermark can refer to the)

SIFT，即尺度不变特征变换（Scale-invariant feature transform，SIFT），是用于图像处理领域的一种描述。这种描述具有尺度不变性，可在图像中检测出关键点，是一种局部特征描述子。[1] 该方法于1999年由David Lowe[2] 首先发表于计算机视觉国际会议（International Conference on Computer Vision，ICCV），2004年再次经David Lowe整理完善后发表于International journal of computer vision（IJCV）。截止2014年8月，该论文单篇被引次数达25000余次。(SIFT, namely Scale Invariant Feature Transform (Scale-invariant feature transform, SIFT), it is a description for image processing. This description has scale invariance, can detect critical point in the image, it is a local feature descriptor. [1] The method in 1999 by David Lowe [2] First published in the International Conference on Computer Vision (International Conference on Computer Vision, ICCV), 2004 was again perfect finishing after David Lowe published in the International journal of computer vision (IJCV) . As of August 2014, the paper cited the number of times a single article of more than 25,000 times.)

Gaussian Kernel Density Estimation with Bounded Support

用matlab实现不规则光照下自适应阈值的前景背景分离算法。文件里面附测试图像。(Matlab implementation to use irregular light of the prospects for adaptive threshold background separation algorithm. Test image files attached inside.)

在高斯混合噪声背景下实现SLIM谱估计算法和l1范数SLIM谱估计法，并在-5到15的信噪比条件与CRLB对比均方频率误差(the paper achieves the SLIM spectral estimation algorithm and l1-SLIM spectral estimation method under the Gaussian mixture background noise,then compares the mean square frequency error with CRLB under SNR from-5 to 15 )