1.6 尺度不变特征提取

说明：  该设计使用8个AXI视频直接存储器访问（AXI VDMA）引擎同时移动16个流（8个传输视频流和8个接收视频流），每个流以1920 x 1080像素格式以60赫兹刷新率移动，每个像素24个数据位。此设计还具有额外的视频等效AXI流量，该流量由为1080p视频模式配置的四个LogiCORE AXI流量发生器（ATG）核心生成。ATG核心根据其配置生成连续的AXI流量。在本设计中，ATG被配置成以1080p模式生成AXI4视频流量。这使得系统吞吐量需求达到DDR的80%左右带宽。每个AXI VDMA由LogiCORE IP测试模式生成器（AXI TPG）核心驱动。AXI VDMA配置为在自由运行模式下运行。每个AXI VDMA读取的数据被发送到能够将多个视频流多路复用或叠加到单个输出视频流的通用视频屏幕显示（AXI OSD）核心。AXI OSD核心的输出驱动板载高清媒体接口（HDMI技术）视频显示接口通过RGB到YCrCb颜色空间转换器核心和逻辑核心IP色度重采集器核心。LogiCore视频定时控制器（AXI VTC）生成所需的定时信号。(The design uses eight AXI video direct memory access (AXI VDMA) engines to simultaneously move 16 streams (eight transmit video streams and eight receive video streams), each in 1920 x 1080 pixel format at 60 Hz refresh rate, and 24 data bits per pixel. This design also has additional video equivalent AXI traffic generated from four LogiCORE AXI Traffic Generator(ATG) cores configured for 1080p video mode. The ATG core generates continuous AXI traffic based on its configuration. In this design, ATG is configured to generate AXI4 video traffic in 1080p mode. This pushes the system throughput requirement to approximately 80% of DDR bandwidth. Each AXI VDMA is driven from a LogiCORE IP Test Pattern Generator (AXI TPG)core. AXI VDMA is configured to operate in free running mode. Data read by each AXI VDMA is sent to a common Video On-Screen Display (AXI OSD) core capable of multiplexing or overlaying multiple video streams to a single output video stream.)

说明：  这是一个用matlab实现对图片的任意形状的剪切的程序，具有极强的实用性(This is a matlab implementation of the image with arbitrary shape cut in the process, with a strong practical)

说明：  利用PCA对人脸图像进行降维，然后训练神经网络分类器的Matlab程序(the Matlab in face recoginization using PCA algorithm)

一个用IDL编写的GUI视图工具 详细源代码与工程文件 包含加载影像 RGB波段合成列表选择 放缩与全图漫游 鹰眼图 矢量数据空间叠加等视图处理功能 按照说明书操作即可！(IDL prepared using a GUI tool to view a detailed source code and project file contains the load image list to select RGB-Band Synthetic zoom map with roaming space Hawkeye map overlay vector data such as view processing functions can be operated in accordance with the Manual!)

说明：  我们提出了一个简单、灵活和通用的对象实例分割框架。我们的方法能有效检测图像中的对象，同时为每个实例生成高质量的 segmentation mask。这种被称为 Mask R-CNN 的方法通过添加用于预测 object mask 的分支来扩展 Faster R-CNN，该分支与用于边界框识别的现有分支并行。Mask R-CNN 训练简单，只需在以 5fps 运行的 Faster R-CNN 之上增加一个较小的 overhead。此外，Mask R-CNN 很容易推广到其他任务，例如它可以允许同一个框架中进行姿态估计。我们在 COCO 系列挑战的三个轨道任务中均取得了最佳成果，包括实例分割、边界对象检测和人关键点检测。没有任何 tricks，Mask R-CNN 的表现优于所有现有的单一模型取得的成绩，包括 COCO 2016 挑战赛的冠军。(Mask R-CNN code by HeKaiming)

UCLA一个博士生Andrea Vedaldi编写的SIFT算法(A doctoral student named Andrea Vedaldi of UCLA wrote this SIFT algorithm code)

汽车识别，计数源代码 通过opencv色差提取，对移动的车辆进行识别 当已经识别的车辆通过定义直线时候，进行计数(Vehicle identification, counting opencv source code by extracting the color of a moving vehicle to identify when a vehicle has been identified by defining a straight line when counting)

说明：  这是一个矢量控制成功的双馈异步风机的mdl模型(This is an mdl model of a doubly-fed asynchronous fan with successful vector control)

图像分解，基于能量泛函的，将图像分解成结构和纹理(Image decomposition, based on energy functional, there will be broken down into the structure and texture image)

自己编写的C++小程序，十字算法，用到KD树，能够有效进行点云去噪，压缩数据(I have written a small program C++, cross algorithm used in KD tree, the point cloud can be effectively de-noising, compression of data)