一、通过训练的模型可以将织物的起球区域进行识别区分

原图：
模型识别：

二、对比度的计算

为了对织物起球等级进行评定，需要这边不同的参数特征来构建模型的评级系统
通过查阅相关文献，确定最终的特征参数为：织物起球个数、织物起球最大面积、织物起球平均面积、对比度、光学体积、密度分布
织物起球个数、织物起球最大面积、织物起球平均面积这些通过轮廓即可实现统计，详情可参考博文：十七、轮廓发现

文献中给出的定义如下：

①毛球区域平均灰度值

织物原图为：E:/Jupyter_notebook/pytorch/opencv/2_origin.jpg
模型预测的图片为：E:/Jupyter_notebook/pytorch/opencv/2_predict.jpg

此时只需要对模型预测的图片2_predict.jpg转灰度图
统计灰度图中灰度值总和、非0像素值的像素点个数，然后相除即可得到毛球区域平均灰度值

img = cv2.imread(r"E:\\Jupyter_notebook\\pytorch\\opencv\\2_predict.jpg")，读取图片
grayImage = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)，转灰度图
x = cv2.countNonZero(grayImage)，非0值的像素点个数
pill_gray = np.sum(grayImage)，像素点之和，因为黑色像素点值为0，没有影响
aver_pill_gray = pill_gray/x，得到毛球区域平均灰度值
最终得到毛球区域平均灰度值aver_pill_gray

import cv2
import numpy as np
from matplotlib import pyplot as pltdef show_photo(name,picture):#图像显示函数cv2.imshow(name,picture)cv2.waitKey(0)cv2.destroyAllWindows()img = cv2.imread(r"E:\\Jupyter_notebook\\pytorch\\opencv\\2_predict.jpg")
grayImage = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)show_photo("origin",img)
show_photo("gray",grayImage)x = cv2.countNonZero(grayImage) #非0值的像素点个数
pill_gray = np.sum(grayImage) #像素点之和 aver_pill_gray = pill_gray/xprint(pill_gray)
print(x)
print(aver_pill_gray)
"""
160314
8964
17.884203480589022
"""

起球区域：
起球区域灰度图：

②非毛球区域平均灰度值

原图和模型预测的图片做差即可得到非毛球区域，然后转换为灰度图，求非0像素点总个数、求总像素点之和，就可以计算出非毛球区域平均灰度值

img_origin = cv2.imread(r"E:/Jupyter_notebook/pytorch/opencv/2_origin.jpg")，读取原图
img_predict = cv2.imread(r"E:/Jupyter_notebook/pytorch/opencv/2_predict.jpg")，读取模型预测之后的只含有起球区域的图像
img = cv2.subtract(img_origin, img_predict)，原图和起球区域做差，得到非起球区域图像
grayImage = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)，非起球区域图像灰度图
x = cv2.countNonZero(grayImage)，计算非起球区域图像灰度图中非0值的像素点个数
nonpill_gray = np.sum(grayImage)，计算非起球区域图像灰度图像素点之和，因为黑色像素点值为0，没有影响
aver_nonpill_gray = nonpill_gray/x，计算非起球区域图像灰度图的平均值灰度值
最终得到非毛球区域平均灰度值aver_nonpill_gray

import cv2
import numpy as np
from matplotlib import pyplot as pltdef show_photo(name,picture):#图像显示函数cv2.imshow(name,picture)cv2.waitKey(0)cv2.destroyAllWindows()img_origin = cv2.imread(r"E:/Jupyter_notebook/pytorch/opencv/2_origin.jpg")
img_predict = cv2.imread(r"E:/Jupyter_notebook/pytorch/opencv/2_predict.jpg")img = cv2.subtract(img_origin, img_predict)
grayImage = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
x = cv2.countNonZero(grayImage) #非0值的像素点个数
nonpill_gray = np.sum(grayImage)#像素点之和 
aver_nonpill_gray = nonpill_gray/xshow_photo("origin",img_origin)
show_photo("pilling",img_predict)
show_photo("nonpilling",img)
show_photo("gray",grayImage)print(nonpill_gray)
print(x)
print(aver_nonpill_gray)
"""
6103678
50237
121.49766108645022
"""

原图：
起球区域：
非起球区域：
非起球区域灰度图：

③对比度

根据论文中给出的公式，对比度就是aver_pill_gray和aver_nonpill_gray做差
C = abs(aver_pill_gray - aver_nonpill_gray)

三、完整代码

import cv2
import numpy as np
from matplotlib import pyplot as pltdef show_photo(name,picture):#图像显示函数cv2.imshow(name,picture)cv2.waitKey(0)cv2.destroyAllWindows()img_origin = cv2.imread(r"E:/Jupyter_notebook/pytorch/opencv/2_origin.jpg")
img_predict = cv2.imread(r"E:/Jupyter_notebook/pytorch/opencv/2_predict.jpg")
img_nonpill = cv2.subtract(img_origin, img_predict)grayImage_pill = cv2.cvtColor(img_predict,cv2.COLOR_BGR2GRAY)
grayImage_nonpill = cv2.cvtColor(img_nonpill,cv2.COLOR_BGR2GRAY)pill_number = cv2.countNonZero(grayImage_pill) #非0值的像素点个数
nonpill_number = cv2.countNonZero(grayImage_nonpill) #非0值的像素点个数pill_gray = np.sum(grayImage_pill) #像素点之和 
nonpill_gray = np.sum(grayImage_nonpill) #像素点之和 aver_pill_gray = pill_gray/pill_number
aver_nonpill_gray = nonpill_gray/nonpill_numberC = abs(aver_pill_gray - aver_nonpill_gray)print(aver_pill_gray)
print(aver_nonpill_gray)
print(C)
"""
17.884203480589022
121.49766108645022
103.6134576058612
"""