检测和分割目标图像实例

代码示例：

import cv2 as cv
import numpy as np


# 读取图片
img_path = r"D:workplacedataopencvfootball.jpg"
img = cv.imread(img_path)
# 转灰度
gray = cv.cvtColor(img,cv.COLOR_BGR2GRAY)
# 对灰度图，高斯去噪
blurred = cv.GaussianBlur(gray, (9, 9), 0)  # todo: 如何去噪
cv.imshow('blurred_img', blurred)

# 提取图像梯度，以sobel算子，计算x，y方向上的梯度
grad_x = cv.Sobel(blurred, ddepth=cv.CV_32F, dx=1, dy=0) # 或者对blurred,gray
grad_y = cv.Sobel(blurred, ddepth=cv.CV_32F, dx=0, dy=1)

gradient = cv.subtract(grad_x, grad_y)
gradient = cv.convertScaleAbs(gradient)
cv.imshow('gradient_img', gradient)

# 去噪
blurred = cv.GaussianBlur(gradient, (9,9),0)
(_, thresh) = cv.threshold(blurred, 90,255,cv.THRESH_BINARY)

# 图像形态学
# 建立一个椭圆函数
kernel = cv.getStructuringElement(cv.MORPH_ELLIPSE, (25,25))
closed = cv.morphologyEx(thresh, cv.MORPH_CLOSE, kernel)
# 细节刻画
closed = cv.erode(closed, None, iterations=4)
closed = cv.dilate(closed, None, iterations=4)


# 找出区域轮廓
(_,cnts, _) = cv.findContours(closed.copy(),cv.RETR_EXTERNAL, cv.CHAIN_APPROX_SIMPLE)
# 画出轮廓
c = sorted(cnts, key=cv.contourArea, reverse=True)[0]
rect = cv.minAreaRect(c)
box = np.int0(cv.boxPoints(rect))

draw_img = cv.drawContours(img.copy(), [box], -1, (0,0,255), 3)

# 剪切出对区域
xs = [i[0] for i in box]
ys = [i[1] for i in box]
x1 = abs(min(xs))
x2 = abs(max(xs))
y1 = abs(min(ys))
y2 = abs(max(ys))

height = y2 - y1
width = x2 - x1
crop_img = img[y1:y1+height, x1:x1+width]


if __name__ == '__main__':
    # 显示
    # cv.imshow('origin_img', origin_img)
    # cv.imshow('blurred_img', blurred)
    # cv.imshow('grad_x_img', grad_x)
    # cv.imshow('grad_y_img', grad_y)
    # cv.imshow('gradient_img', gradient)
    # cv.imshow('thresh_img', thresh)
    # cv.imshow('closed_img', closed)

    cv.imshow('draw_img', draw_img)
    cv.imshow('select_img', crop_img)
    cv.waitKey(0)
    cv.destroyAllWindows()

结果展示：

opencv api函数调用

# 用来转化图像格式的
img = cv2.cvtColor(src, 
    COLOR_BGR2HSV # BGR---->HSV
    COLOR_HSV2BGR # HSV---->BGR
    ...)
# For HSV, Hue range is [0,179], Saturation range is [0,255] and Value range is [0,255]


# 返回一个阈值，和二值化图像，第一个阈值是用来otsu方法时候用的
# 不过现在不用了，因为可以通过mahotas直接实现
T = ret = mahotas.threshold(blurred)
ret, thresh_img = cv2.threshold(src, # 一般是灰度图像
    num1, # 图像阈值
    num2, # 如果大于或者num1, 像素值将会变成 num2
# 最后一个二值化参数
    cv2.THRESH_BINARY      # 将大于阈值的灰度值设为最大灰度值，小于阈值的值设为0
    cv2.THRESH_BINARY_INV  # 将大于阈值的灰度值设为0，大于阈值的值设为最大灰度值
    cv2.THRESH_TRUNC       # 将大于阈值的灰度值设为阈值，小于阈值的值保持不变
    cv2.THRESH_TOZERO      # 将小于阈值的灰度值设为0，大于阈值的值保持不变
    cv2.THRESH_TOZERO_INV  # 将大于阈值的灰度值设为0，小于阈值的值保持不变
)
thresh = cv2.AdaptiveThreshold(src, 
    dst, 
    maxValue, 
    # adaptive_method 
    ADAPTIVE_THRESH_MEAN_C,      
    ADAPTIVE_THRESH_GAUSSIAN_C,      
    # thresholdType
    THRESH_BINARY, 
    THRESH_BINARY_INV, 
    blockSize=3,
    param1=5
)


# 一般是在黑色背景中找白色物体，所以原始图像背景最好是黑色
# 在执行找边缘的时候，一般是threshold 或者是canny 边缘检测后进行的。
# warning:此函数会修改原始图像、
# 返回：坐标位置（x,y）, 
(_, cnts, _) = cv2.findContours(mask.copy(), 
    # cv2.RETR_EXTERNAL,             #表示只检测外轮廓
    # cv2.RETR_CCOMP,                #建立两个等级的轮廓,上一层是边界
    cv2.RETR_LIST,                 #检测的轮廓不建立等级关系
    # cv2.RETR_TREE,                   #建立一个等级树结构的轮廓
    # cv2.CHAIN_APPROX_NONE,           #存储所有的轮廓点，相邻的两个点的像素位置差不超过1
    cv2.CHAIN_APPROX_SIMPLE,       #例如一个矩形轮廓只需4个点来保存轮廓信息
    # cv2.CHAIN_APPROX_TC89_L1,
    # cv2.CHAIN_APPROX_TC89_KCOS
   )
img = cv2.drawContours(src, cnts, whichToDraw(-1), color, line)


img = cv2.imwrite(filename, dst,  # 文件路径，和目标图像文件矩阵
    
    # 对于JPEG，其表示的是图像的质量，用0-100的整数表示，默认为95
    # 注意，cv2.IMWRITE_JPEG_QUALITY类型为Long，必须转换成int
    [int(cv2.IMWRITE_JPEG_QUALITY), 5] 
    [int(cv2.IMWRITE_JPEG_QUALITY), 95]
    # 从0到9,压缩级别越高，图像尺寸越小。默认级别为3
    [int(cv2.IMWRITE_PNG_COMPRESSION), 5])
    [int(cv2.IMWRITE_PNG_COMPRESSION), 9])

# 寻找某个函数或者变量 
events = [i for i in dir(cv2) if 'PNG' in i]
print( events )

寻找某个变量开头的flags
flags = [i for i in dir(cv2) if i.startswith('COLOR_')]
print flags

批量读取文件名字
import os
filename_rgb = r'C:Users...'
for filename in os.listdir(filename_rgb):              #listdir的参数是文件夹的路径
    print (filename)