低时延检测技术分析

说明：  机器学习专门研究计算机怎样模拟或实现人类的学习行为，以获取新的知识或技 能，为包括人脸识别领域的人工智能的发展做出了极大的贡献。本文探索如何应用机器学习中的一些技术，使计算机更好地完成人脸识别领域中的人脸检测和人脸验证。 在人脸检测方面，针对如何快速、准确地检出人脸的问题，基于DLIB中特征模型，可以快速提取检测出人脸，并且提取出人脸特征点,主要作用就是快速检测定位人脸。在训练过程中，引入了ResNet机器学习算法，该算法采用多层卷积神经网络结构，对人脸进行多层特征提取和描述，得到人脸特征描述符。通过测试可以证实了上述方法能够检测和识别人脸。(In terms of face detection, for the problem of how to detect faces quickly and accurately, based on the feature model in DLIB, you can quickly extract and detect faces, and extract face feature points, the main role is to quickly detect and locate faces. In the training process, the VGG-FACE machine learning algorithm is introduced. This algorithm uses a multi-layer convolutional neural network structure to perform multi-layer feature extraction and description on the face to obtain the face feature descriptor. Tests can confirm that the above method can detect and recognize human faces.)

说明：  平面拟合，RANSAC拟合方法，关于平面重构，画示意图，计算法向量(Plane Fitting and RANSAC Fitting Method)

图像颜色聚类分割，实现了图形分割，基于RGB特征并显示出来。(Color image segmentation clustering, to achieve the graphics division, based on the characteristics of RGB and displayed.)

说明：  采用模糊C均值聚类算法求到欧式距离和方差，对数据集（特征+标签）进行分类并得到分类准确率(The Euclidean distance and variance are calculated by fuzzy c-means clustering algorithm, and the data set (feature + label) is classified and the classification accuracy is obtained)

非压缩视频图像帧内水印的密写与复密写研究，源码和论文材料(Video Watermarking Steganography)

说明：  词： 适用于医学图像配准的程序，可以设置不同参数配准(The program for medical image registration can set different parameters for registration)

红外图像细节增强仿真，其中使用双边滤波器进行图像分层。带测试源数据。(Infrared image detail enhancement, with test source data)

本例功能是显示一个由反走样处理的线段组成的八面体(functions of the cases showed an anti-aliasing by the Line of octahedral)

适用于任意图像中轮廓的直线段检测与描绘，对于公路车道直线检测具有良好效果。(Linear segment detection and description for contours in arbitrary Images,it has good effect for straight line detection of highway lanes.)

FPGA SDI 输入，HDMI输出例程(FPGA SDI_IN,HDMI_OUT)