detect

图像处理中的深度学习，主要是原理的讲解，比较易懂,对于。(Image processing in deep learning, mainly to explain the principle, relatively easy to understand.)

数字图像处理中的散度特征空间中的无监督的图像纹理分割(Digital image processing in the feature space of divergence Unsupervised texture segmentation of images)

数字图像处理（MATLAB版）（第二版）的工具箱，m文件(Digital image processing (MATLAB version) (Second Edition) toolbox, m file)

说明：  这是一个基于活动轮廓模型的快速全局最小算法的matlab实现程序，对学习现代图像处理有一定帮助。(This is an active contour model based on the rapid global minimum algorithm matlab implementation process of learning in modern image processing have some help.)

基于灰度图像进行傅立叶变换后的二值化处理(Binarization based on the Fourier transform after the gray image processing)

1．选择一幅图像，对该图像进行DCT块操作（将图像分成8×8小块，对每个小块进行DCT变换）。 2.利用屏蔽矩阵对DCT系数进行区域选择，然后进行逆变换（IDCT），合成原图像，计算合成图像的PSNR。 3.选择一小波函数，对一幅图像进行分解，显示各变换系数，观察其规律，并进行小波逆变换，合成原图像。 (1. Select an image, the image DCT block operations (the image is divided into 8 × 8 small pieces of each small piece for DCT transform). 2. The use of shielding on the DCT coefficient matrix of regional choices, and then inverse transform (IDCT), synthesis of the original image, calculating the synthetic image PSNR. 3. To select a wavelet function, decomposition of an image to show the transform coefficients, to observe its laws, and inverse wavelet transform, the original synthetic image.)

课题是用粒子滤波进行目标跟踪，对经典粒子滤波与其它粒子滤波进行了详细的讲解.(For target tracking particle filter algorithm)

说明：  使用vivado HLS工具实现的直方图均衡化(EqualizeHist with vivado HLS)

实现一个高精度指针仪表读数自动识别系统．用数码摄像头采集仪表的图像，然后利用数字图像处理技术识别刻度和指针，根据指针和刻度的相对位置计算指针读数值．使用条件霍夫变换（Constrained HOUGH Transfer）和中心投影分析法实现刻度的全自动识别，采用快速中心投影法进行指针检测与识别，识别速度达到68ms．实验结果表明该方法比传统方法速度快、精度高，系统读数平均误差仅为0．016％(To achieve a high precision automatic identification system pointer meter reading. Instrument with a digital camera capture images, then using digital image processing technology to identify scale and pointer, pointer and scale according to the relative position of the pointer readings calculated. Use Hough transform (Constrained HOUGH Transfer) and center projection analysis method to achieve scale automatic identification method using fast center projection target detection and recognition, recognition speed of 68ms. Experimental results show that faster than traditional methods, high accuracy, the system average error of only 0.016 of reading)

说明：  基于灰度矩的亚像素边缘检测，零件边缘提取(Subpixel edge contour extraction)