深度学习和目标检测系列教程 12-尊龙游戏旗舰厅官网

@author：runsen

由于cv需要熟练使用opencv，因此总结了opencv常见的api和用法。

opencv（opensourcecomputervision）于1999年正式推出，它来自英特尔的一项倡议。

• opencv的核心是用c 编写的。在python中，我们只使用一个包装器，它在python内部执行c 代码。

• 它对于几乎所有的计算机视觉应用程序都非常有用，并且在windows、linux、macos、android、ios上受支持，并绑定到python、java和matlab。

锐化

usm锐化的全称是：unsharp mask，译为「模糊掩盖锐化处理」，是一种胶片时代处理图片锐度的手法，延续到数码时代的产物。在胶片时代，我们通过将模糊的负片与正片叠加可产生边缘锐化的效果。

对，锐化的效果离不开模糊，甚至可以说，锐化的效果就是来源于模糊。usm的锐化实际上就是利用原图和模糊图产生的反差，来实现锐化图片的效果。

公式：（源图像– w*高斯模糊）/（1-w）；其中w表示权重（0.1～0.9）。

我感觉我喜欢上，毕业前在学校自拍的照片

import numpy as np import matplotlib.pyplot as plt import cv2image = cv2.imread('demo.jpg') image = cv2.cvtcolor(image, cv2.color_bgr2rgb)plt.figure(figsize=(20, 20)) plt.subplot(1, 2, 1) plt.title("original") plt.imshow(image)# create our shapening kernel # the values in the matrix sum to 1 kernel_sharpening = np.array([[-1,-1,-1], [-1,9,-1], [-1,-1,-1]])# 对输入图像应用不同的内核 sharpened = cv2.filter2d(image, -1, kernel_sharpening)plt.subplot(1, 2, 2) plt.title("image sharpening") plt.imshow(sharpened)plt.show()

阈值化、二值化

image = cv2.imread('demo.jpg', 0)plt.figure(figsize=(30, 30)) plt.subplot(3, 2, 1) plt.title("original") plt.imshow(image)# 小于127的值变为0（黑色，大于等于255（白色） ret,thresh1 = cv2.threshold(image, 127, 255, cv2.thresh_binary)plt.subplot(3, 2, 2) plt.title("threshold binary") plt.imshow(thresh1)# 模糊图像，消除噪音 image = cv2.gaussianblur(image, (3, 3), 0)# adaptivethreshold thresh = cv2.adaptivethreshold(image, 255, cv2.adaptive_thresh_mean_c, cv2.thresh_binary, 3, 5) plt.subplot(3, 2, 3) plt.title("adaptive mean thresholding") plt.imshow(thresh)_, th2 = cv2.threshold(image, 0, 255, cv2.thresh_binary cv2.thresh_otsu)plt.subplot(3, 2, 4) plt.title("otsu's thresholding") plt.imshow(th2)plt.subplot(3, 2, 5) # 高斯滤波后的大津阈值法 blur = cv2.gaussianblur(image, (5,5), 0) _, th3 = cv2.threshold(blur, 0, 255, cv2.thresh_binary cv2.thresh_otsu) plt.title("guassian otsu's thresholding") plt.imshow(th3) plt.show()

降噪

image = cv2.imread('demo.jpg') image = cv2.cvtcolor(image, cv2.color_bgr2rgb)plt.figure(figsize=(20, 20)) plt.subplot(3, 2, 1) plt.title("original") plt.imshow(image)# let's define our kernel size kernel = np.ones((5,5), np.uint8)# now we erode erosion = cv2.erode(image, kernel, iterations = 1)plt.subplot(3, 2, 2) plt.title("erosion") plt.imshow(erosion)dilation = cv2.dilate(image, kernel, iterations = 1) plt.subplot(3, 2, 3) plt.title("dilation") plt.imshow(dilation)# opening - good for removing noise opening = cv2.morphologyex(image, cv2.morph_open, kernel) plt.subplot(3, 2, 4) plt.title("opening") plt.imshow(opening)# closing - good for removing noise closing = cv2.morphologyex(image, cv2.morph_close, kernel) plt.subplot(3, 2, 5) plt.title("closing") plt.imshow(closing)

边缘检测与图像梯度

image = cv2.imread('demo.jpg') image = cv2.cvtcolor(image, cv2.color_bgr2rgb)height, width,_ = image.shape# extract sobel edges sobel_x = cv2.sobel(image, cv2.cv_64f, 0, 1, ksize=5) sobel_y = cv2.sobel(image, cv2.cv_64f, 1, 0, ksize=5)plt.figure(figsize=(20, 20))plt.subplot(3, 2, 1) plt.title("original") plt.imshow(image)plt.subplot(3, 2, 2) plt.title("sobel x") plt.imshow(sobel_x)plt.subplot(3, 2, 3) plt.title("sobel y") plt.imshow(sobel_y)sobel_or = cv2.bitwise_or(sobel_x, sobel_y)plt.subplot(3, 2, 4) plt.title("sobel_or") plt.imshow(sobel_or)laplacian = cv2.laplacian(image, cv2.cv_64f)plt.subplot(3, 2, 5) plt.title("laplacian") plt.imshow(laplacian)## 提供两个值：threshold1和threshold2。任何大于threshold2的梯度值。低于threshold1的任何值都不被视为边。 # threshold1和threshold2之间的值可以根据其大小分类为边或非边 # 在这种情况下，低于60的任何渐变值都被视为非边 # 而大于120的任何值都被视为边。 # the first threshold gradient canny = cv2.canny(image, 50, 120)plt.subplot(3, 2, 6) plt.title("canny") plt.imshow(canny)

透视变换

image = cv2.imread('scan.jpg') image = cv2.cvtcolor(image, cv2.color_bgr2rgb)plt.figure(figsize=(20, 20))plt.subplot(1, 2, 1) plt.title("original") plt.imshow(image)# 原始图像四个角的坐标 points_a = np.float32([[320,15], [700,215], [85,610], [530,780]])# 所需输出的4个角的坐标 # 使用a4纸的比例是1:1.41 points_b = np.float32([[0,0], [420,0], [0,594], [420,594]])# 使用两组四个点进行计算 # 透视变换矩阵，m m = cv2.getperspectivetransform(points_a, points_b)warped = cv2.warpperspective(image, m, (420,594))plt.subplot(1, 2, 2) plt.title("warpperspective") plt.imshow(warped)

缩放、重新调整大小和插值

使用cv2.resize函数可以很容易地重新调整大小，它的参数有：cv2.resize（image，dsize（output image size），x scale，y scale，interpolation）

image = cv2.imread('demo.jpg') image = cv2.cvtcolor(image, cv2.color_bgr2rgb)plt.figure(figsize=(20, 20))plt.subplot(2, 2, 1) plt.title("original") plt.imshow(image)# let's make our image 3/4 of it's original size image_scaled = cv2.resize(image, none, fx=0.75, fy=0.75)plt.subplot(2, 2, 2) plt.title("scaling - linear interpolation") plt.imshow(image_scaled)# let's double the size of our image img_scaled = cv2.resize(image, none, fx=2, fy=2, interpolation = cv2.inter_cubic)plt.subplot(2, 2, 3) plt.title("scaling - cubic interpolation") plt.imshow(img_scaled)# let's skew the re-sizing by setting exact dimensions img_scaled = cv2.resize(image, (900, 400), interpolation = cv2.inter_area)plt.subplot(2, 2, 4) plt.title("scaling - skewed size") plt.imshow(img_scaled)

影像金字塔

在目标检测中缩放图像时非常有用。

image = cv2.imread('demo.jpg') image = cv2.cvtcolor(image, cv2.color_bgr2rgb)plt.figure(figsize=(20, 20))plt.subplot(2, 2, 1) plt.title("original") plt.imshow(image)smaller = cv2.pyrdown(image) larger = cv2.pyrup(image)plt.subplot(2, 2, 2) plt.title("smaller") plt.imshow(smaller)plt.subplot(2, 2, 3) plt.title("larger") plt.imshow(larger)

裁剪

image = cv2.imread('demo.jpg') image = cv2.cvtcolor(image, cv2.color_bgr2rgb)plt.figure(figsize=(20, 20))plt.subplot(2, 2, 1) plt.title("original") plt.imshow(image)height, width = image.shape[:2]# let's get the starting pixel coordiantes (top left of cropping rectangle) start_row, start_col = int(height * .25), int(width * .25)# let's get the ending pixel coordinates (bottom right) end_row, end_col = int(height * .75), int(width * .75)# simply use indexing to crop out the rectangle we desire cropped = image[start_row:end_row , start_col:end_col]plt.subplot(2, 2, 2) plt.title("cropped") plt.imshow(cropped)

模糊

image = cv2.imread('demo.jpg') image = cv2.cvtcolor(image, cv2.color_bgr2rgb)plt.figure(figsize=(20, 20))plt.subplot(2, 2, 1) plt.title("original") plt.imshow(image)# creating our 3 x 3 kernel kernel_3x3 = np.ones((3, 3), np.float32) / 9# we use the cv2.fitler2d to conovlve the kernal with an image blurred = cv2.filter2d(image, -1, kernel_3x3)plt.subplot(2, 2, 2) plt.title("3x3 kernel blurring") plt.imshow(blurred)# creating our 7 x 7 kernel kernel_7x7 = np.ones((7, 7), np.float32) / 49blurred2 = cv2.filter2d(image, -1, kernel_7x7)plt.subplot(2, 2, 3) plt.title("7x7 kernel blurring") plt.imshow(blurred2)

contours

# let's load a simple image with 3 black squares image = cv2.imread('demo.jpg') image = cv2.cvtcolor(image, cv2.color_bgr2rgb)plt.figure(figsize=(20, 20))plt.subplot(2, 2, 1) plt.title("original") plt.imshow(image)# grayscale gray = cv2.cvtcolor(image,cv2.color_bgr2gray)# find canny edges edged = cv2.canny(gray, 30, 200)plt.subplot(2, 2, 2) plt.title("canny edges") plt.imshow(edged)# finding contours # use a copy of your image e.g. edged.copy(), since findcontours alters the image contours, hierarchy = cv2.findcontours(edged, cv2.retr_external, cv2.chain_approx_none)plt.subplot(2, 2, 3) plt.title("canny edges after contouring") plt.imshow(edged)print("number of contours found = " str(len(contours)))# draw all contours # use '-1' as the 3rd parameter to draw all cv2.drawcontours(image, contours, -1, (0,255,0), 3)plt.subplot(2, 2, 4) plt.title("contours") plt.imshow(image)

总结

以上是尊龙游戏旗舰厅官网为你收集整理的深度学习和目标检测系列教程 12-300：常见的opencv的api和用法总结的全部内容，希望文章能够帮你解决所遇到的问题。

如果觉得尊龙游戏旗舰厅官网网站内容还不错，欢迎将尊龙游戏旗舰厅官网推荐给好友。