主要思想是先检测外边圆和圆心
然后再外圆内检测小圆，计算小圆圆心与外圆圆心的距离判断是不是有问题
或者可以计算两圆圆心的距离
# coding:utf-
import math
import cv
import numpy as np
import os
def findNeedlePoints(img):
gray_src= cv.cvtColor(img,cv2.COLOR_BGR2GRAY)
minThreshValue =
_, gray = cv.threshold(gray_src, minThreshValue, 255, cv2.THRESH_BINARY)
erosion_size =
# element = cv.getStructuringElement(cv2.MORPH_ELLIPSE, (2 * erosion_size + 1, 2 * erosion_size + 1),
# (erosion_size, erosion_size))
element = cv.getStructuringElement(cv2.MORPH_ERODE, (2 * erosion_size + 1, 2 * erosion_size + 1),
(erosion_size, erosion_size))
# MORPH_ELLIPSE 不同的测试一下
erosion_gray = cv.erode(gray, element, 3)
cv.imshow("erosion_gray", erosion_gray)
paramsIn = cv.SimpleBlobDetector_Params()
paramsIn.filterByArea = True
# 不同图片应该调节的参数
paramsIn.minArea =
paramsIn.maxArea =
paramsIn.minDistBetweenBlobs =
paramsIn.filterByColor = True
paramsIn.filterByConvexity = False
paramsIn.minThreshold =*2
paramsIn.maxThreshold =
# 图像取反
needleGray = - erosion_gray.copy()
# 中值滤波和腐蚀去噪
needleGray = cv.medianBlur(needleGray, 3)
# cv.imshow('needleGray', needleGray)
erosion_size =
element = cv.getStructuringElement(cv2.MORPH_RECT, (2 * erosion_size + 1, 2 * erosion_size + 1),
(erosion_size, erosion_size))
needlePoints = cv.erode(needleGray, element, 1)
cv.imshow('needle=Points', needlePoints)
detector = cv2.SimpleBlobDetector_create(paramsIn)
needleKeypoints = detector.detect(needlePoints)
# opencv
needle_keypoints = cv.drawKeypoints(needlePoints, needleKeypoints, np.array([]), (255, 0, 0),
cv.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS)
allNeedlePoints = []
if needleKeypoints is not None:
for i in range(len(needleKeypoints)):
allNeedlePoints.append(needleKeypoints[i].pt)
color_img = cv.cvtColor(needle_keypoints, cv2.COLOR_BGR2RGB)
# needle_img = cv.cvtColor(im_with_keypoints, cv2.COLOR_BGR2RGB)
cv.imshow('holeShow', color_img)
# cv.imshow('needleShow', needle_img)
cv.waitKey()
def innerHoughCicle(hsv_image, src_image, rect):
# 霍夫变换圆检测
gray_src = cv.cvtColor(hsv_image, cv2.COLOR_HSV2RGB)
gray_src = cv.cvtColor(gray_src, cv2.COLOR_RGB2GRAY)
minThreshValue =
_, gray = cv.threshold(gray_src, minThreshValue, 255, cv2.THRESH_BINARY)
kernel = np.ones((3, 3), dtype=np.uint8)
kernel = np.ones((3, 3), dtype=np.uint8)
gray = cv.erode(gray, kernel2, 2)
gray = cv.dilate(gray, kernel1, 2) # 1:迭代次数，也就是执行几次膨胀操作
# cv.namedWindow("gray", 2)
# cv.imshow("gray", gray)
# cv.waitKey()
circles = cv.HoughCircles(gray, cv2.HOUGH_GRADIENT, 2, 100, param1=100, param2=60, minRadius=10, maxRadius=100)
# 如果没检测到会报错
# 这种判断方式过于简单
if circles is None:
print("没有检测到连接器外圆")
else:
for circle in circles[]:
# 圆的基本信息
# print(circle[])
# 坐标行列－圆心坐标
out_x = int(circle[])
out_y = int(circle[])
# 半径
r = int(circle[])
# # 在原图用指定颜色标记出圆的边界
cv.circle(hsv_image, (out_x, out_y), r, (0, 0, 255), 2)
# # 画出圆的圆心
cv.circle(hsv_image, (out_x, out_y), 3, (0, 0, 255), -1)
cv.namedWindow("hsv_circle", 2)
cv.imshow("hsv_circle",hsv_image)
cv.waitKey()
def outHoughCicle(hsv_image, src_image, rect):
# 霍夫变换圆检测
gray_src = cv.cvtColor(hsv_image, cv2.COLOR_HSV2RGB)
gray_src = cv.cvtColor(gray_src, cv2.COLOR_RGB2GRAY)
minThreshValue =
_, gray = cv.threshold(gray_src, minThreshValue, 255, cv2.THRESH_BINARY)
kernel = np.ones((3, 3), dtype=np.uint8)
kernel = np.ones((3, 3), dtype=np.uint8)
gray = cv.erode(gray, kernel2, 2)
gray = cv.dilate(gray, kernel1, 2) # 1:迭代次数，也就是执行几次膨胀操作
cv.namedWindow("gray", 2)
cv.imshow("gray", gray)
cv.waitKey()
circles = cv.HoughCircles(gray, cv2.HOUGH_GRADIENT, 1, 10e10, param1=100, param2=60, minRadius=500, maxRadius=10000)
# 如果没检测到会报错
# 这种判断方式过于简单
if circles is None:
print("没有检测到连接器外圆")
else:
for circle in circles[]:
# 圆的基本信息
# print(circle[])
# 坐标行列－圆心坐标
out_x = int(circle[])
out_y = int(circle[])
# 半径
r = int(circle[])
# # 在原图用指定颜色标记出圆的边界
cv.circle(hsv_image, (out_x, out_y), r, (0, 0, 255), 2)
# # 画出圆的圆心
cv.circle(hsv_image, (out_x, out_y), 3, (0, 0, 255), -1)
# 画在原图上
cv.circle(src_image, (out_x + rect[0], out_y + rect[1]), r, (0, 0, 255), 2)
# # 画出圆的圆心
cv.circle(src_image, (out_x + rect[0], out_y+ rect[1]), 3, (0, 0, 255), -1)
cv.namedWindow("hsv_circle", 2)
cv.imshow("hsv_circle",hsv_image)
cv.namedWindow("src_image", 2)
cv.imshow("src_image",src_image)
cv.waitKey()
# 检测针脚位置
def needelCenter_detect(img):
params = cv.SimpleBlobDetector_Params()
# Setup SimpleBlobDetector parameters.
# print('params')
# print(params)
# print(type(params))
# Filter by Area.
params.filterByArea = True
params.minArea =
params.maxArea =e3
params.minDistBetweenBlobs =
# params.filterByColor = True
params.filterByConvexity = False
# tweak these as you see fit
# Filter by Circularity
params.filterByCircularity = False
params.minCircularity =.2
# params.blobColor =
# # # Filter by Convexity
# params.filterByConvexity = True
# params.minConvexity =.87
# Filter by Inertia
# params.filterByInertia = True
# params.filterByInertia = False
# params.minInertiaRatio =.01
gray = cv.cvtColor(img, cv2.COLOR_BGR2GRAY)
# Detect blobs.
minThreshValue =
_, gray = cv.threshold(gray, minThreshValue, 255, cv2.THRESH_BINARY)
erosion_size =
# element = cv.getStructuringElement(cv2.MORPH_ELLIPSE, (2 * erosion_size + 1, 2 * erosion_size + 1),
# (erosion_size, erosion_size))
element = cv.getStructuringElement(cv2.MORPH_ERODE, (2 * erosion_size + 1, 2 * erosion_size + 1),
(erosion_size, erosion_size))
dilate_gray = cv.dilate(gray, element, 1)
# cv.namedWindow("gray", 2)
# cv.imshow("gray",dilate_gray)
# cv.waitKey()
detector = cv.SimpleBlobDetector_create(params)
keypoints = detector.detect(dilate_gray)
# print(len(keypoints))
# print(keypoints[].pt[0])
# 如果这儿没检测到可能会出错
if len(keypoints) ==:
print("没有检测到针角坐标，可能需要调整针角斑点检测参数")
print(keypoints)
return keypoints
else:
print("检测到孔的数量", len(keypoints))
# im_with_keypoints = cv.drawKeypoints(img, keypoints, np.array([]), (255, 0, 0),
# cv.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS)
#
# color_img = cv.cvtColor(im_with_keypoints, cv2.COLOR_BGR2RGB)
# 画出圆的圆心
# for kp in keypoints:
# cv.circle(img, (int(kp.pt[0]), int(kp.pt[1])), 3, (0, 0, 255), -1)
#
# cv.namedWindow("color_img", 2)
# cv.imshow("color_img",img)
# # cv.waitKey()
return keypoints
# 检测外部区域针或孔的位置
def out_circle_detect(rect_hole_info, src):
# 灰度化
circle_img = rect_hole_info
gray = cv.cvtColor(circle_img, cv2.COLOR_HSV2RGB)
gray = cv.cvtColor(gray, cv2.COLOR_RGB2GRAY)
# 输出图像大小，方便根据图像大小调节minRadius和maxRadius
# print(image.shape)
# 进行中值滤波
img = cv.medianBlur(gray, 3)
erosion_size =
# element = cv.getStructuringElement(cv2.MORPH_ELLIPSE, (2 * erosion_size + 1, 2 * erosion_size + 1),
# (erosion_size, erosion_size))
element = cv.getStructuringElement(cv2.MORPH_ERODE, (2 * erosion_size + 1, 2 * erosion_size + 1),
(erosion_size, erosion_size))
dilate_gray = cv.dilate(img, element, 1)
# cv.namedWindow("dilate_gray", 2)
# cv.imshow("dilate_gray", dilate_gray)
# cv.waitKey()
# 针角圆心坐标
out_x, out_y, r =, 0, 0
# 霍夫变换检测最大圆
circles = cv.HoughCircles(dilate_gray, cv2.HOUGH_GRADIENT, 1, 1000, param1=100, param2=30, minRadius=500, maxRadius=1000)
# 如果没检测到会报错
# 这种判断方式过于简单
if circles is None:
print("没有检测到连接器外圆")
return, 0, 0
else:
for circle in circles[]:
# 圆的基本信息
# print(circle[])
# 坐标行列－圆心坐标
out_x = int(circle[])
out_y = int(circle[])
# 将检测到的坐标保存
# 半径
r = int(circle[])
# print(r)
# # # 在原图用指定颜色标记出圆的边界
cv.circle(circle_img, (out_x, out_y), r, (0, 0, 255), 2)
# # 画出圆的圆心
cv.circle(circle_img, (out_x, out_y), 5, (0, 0, 255), -1)
cv.namedWindow("circle_imgs", 2)
cv.imshow("circle_imgs", circle_img)
cv.waitKey()
return out_x, out_y, r
# 检测内部区域针或孔的位置
def inner_circle_detect(rect_hole_info, src):
# 灰度化
circle_img = rect_hole_info
gray = cv.cvtColor(circle_img, cv2.COLOR_HSV2RGB)
gray = cv.cvtColor(gray, cv2.COLOR_RGB2GRAY)
# 输出图像大小，方便根据图像大小调节minRadius和maxRadius
# print(image.shape)
# 进行中值滤波
img = cv.medianBlur(gray, 3)
erosion_size =
# element = cv.getStructuringElement(cv2.MORPH_ELLIPSE, (2 * erosion_size + 1, 2 * erosion_size + 1),
# (erosion_size, erosion_size))
element = cv.getStructuringElement(cv2.MORPH_ERODE, (2 * erosion_size + 1, 2 * erosion_size + 1),
(erosion_size, erosion_size))
dilate_gray = cv.dilate(img, element, 1)
# cv.namedWindow("dilate_gray", 2)
# cv.imshow("dilate_gray", dilate_gray)
# cv.waitKey()
# 针角圆心坐标
out_x_p = []
out_y_p = []
rudis = []
# 霍夫变换检测最大圆
circles = cv.HoughCircles(dilate_gray, cv2.HOUGH_GRADIENT, 1, 100, param1=100, param2=30, minRadius=20, maxRadius=100)
# 如果没检测到会报错
# 这种判断方式过于简单
if circles is None:
print("没有检测到连接器外圆")
return out_x_p, out_y_p
else:
for circle in circles[]:
# 圆的基本信息
# print(circle[])
# 坐标行列－圆心坐标
out_x = int(circle[])
out_y = int(circle[])
# 将检测到的坐标保存
out_x_p.append(out_x)
out_y_p.append(out_y)
# 半径
r = int(circle[])
rudis.append(r)
# print(r)
# # # 在原图用指定颜色标记出圆的边界
cv.circle(circle_img, (out_x, out_y), r, (0, 0, 255), 2)
# # 画出圆的圆心
cv.circle(circle_img, (out_x, out_y), 5, (0, 0, 255), -1)
cv.namedWindow("circle_img", 2)
cv.imshow("circle_img",circle_img)
cv.waitKey()
# 记录外圆坐标
out_xpoints = out_x_p.copy()
out_ypoints = out_y_p.copy()
out_rudis = rudis.copy()
# print("out_xpoints",out_xpoints)
# print("out_ypoints",out_ypoints)
# 只框出单个针角的位置区域
step_center =
step_rect =
# 遍历所有的孔的位置
# 记录孔的位置
in_x_p = []
in_y_p = []
for i in range(, len(out_xpoints)):
out_x_begin = out_xpoints[i] - step_center
out_y_begin = out_ypoints[i] - step_center
needleRect = circle_img[out_y_begin: out_y_begin + step_rect, out_x_begin: out_x_begin + step_rect]
# cv.namedWindow("needleRect", 2)
# cv.imshow("needleRect", needleRect)
# cv.waitKey()
# 根据检测到的圆形连接器中心找针角位置
centerPoint = needelCenter_detect(needleRect)
# print(len(centerPoint))
if len(centerPoint) ==:
out_x_p.remove(out_xpoints[i])
out_y_p.remove(out_ypoints[i])
rudis.remove(out_rudis[i])
print("调整位置")
else:
for cp in centerPoint:
# 将针角的坐标原还至原图
in_x = int(cp.pt[])
in_y = int(cp.pt[])
in_x += out_x_begin
in_y += out_y_begin
in_x_p.append(in_x)
in_y_p.append(in_y)
# # # 画出中心孔的圆心
# cv.circle(circle_img, (in_x, in_y), 4, (0, 255, 0), -1)
# # 画出外孔的圆心
# cv.circle(circle_img, (out_xpoints[i], out_ypoints[i]), 4, (0, 0, 255), -1)
# # 计算两者的距离
# # 假设通过标定其一个像素代表.0056mm
# DPI =.0198
# dis = math.sqrt(math.pow(out_xpoints[i] - in_x,) + math.pow(out_ypoints[i] - in_y,2))
# print("两者相互之间的距离为（mm）:", dis*DPI)
return in_x_p,in_y_p
# cv.namedWindow("image", 2)
# cv.imshow("image",circle_img)
# cv.waitKey()
# if len(out_x_p) ==:
# print("没检测到，需要调整位置")
# else:
# for j in range(,len(out_x_p)):
# # 画出外孔的圆心
# cv.circle(circle_img, (out_x_p[j], out_y_p[j]), rudis[j], (0, 0, 255), 3)
# cv.circle(circle_img, (out_x_p[j], out_y_p[j]), 3, (0, 0, 255), -1)
#
# # cv.circle(circle_img, (in_x_p[j], in_y_p[j]), 3, (0, 255, 0), -1)
#
# cv.namedWindow("image", 2)
# cv.imshow("image",circle_img)
# cv.waitKey()
def j_4_holes_dectWX(imagePath, templatePath):
# templatePath需要用户手动框获取ROI
img = cv.imread(imagePath)
img_roi = cv.imread(templatePath)
if img_roi is None:
print("no image")
# HSV二值化
img_roi = cv.medianBlur(img_roi, 5) # 中值滤波
outx, outy, outR = out_circle_detect(img_roi, img)
print(outx, outy, outR )
inx, iny = inner_circle_detect(img_roi, img)
if len(inx) == or outx == 0:
print("没检测到位置")
return "没检测到对象", -
else:
cv.circle(img_roi, (outx, outy), outR, (0, 0, 255), 3)
is_ok = []
for k in range(, len(inx)):
# 计算两者的距离
# 假设通过标定其一个像素代表.0056mm
# 两者相互之间的距离为（mm）:.311053946788194
# 两者相互之间的距离为（mm）:.163550379629067
# 两者相互之间的距离为（mm）:.95984457900917
# 两者相互之间的距离为（mm）:.977940966613671
# 平均值为.103 所以其阈值为9.103 + 0.5
DPI =.0198
dis = math.sqrt(math.pow(outx - inx[k],) + math.pow(outy - iny[k], 2))
dis *= DPI
# print("两者相互之间的距离为（mm）:", dis)
if dis <.603:
cv.circle(img_roi, (inx[k], iny[k]), 8, (0, 255, 0), -1)
# print("没有插针歪斜，产品合格")
is_ok.append()
else:
cv.circle(img_roi, (inx[k], iny[k]), 20, (0, 0, 255), 3)
# print("有插针歪斜，不合格")
is_ok.append()
# cv.namedWindow("image", 2)
# cv.imshow("image",img_roi)
# cv.waitKey()
isExists = os.path.exists("./runs/J/")
if not isExists:
os.makedirs("./runs/J/")
cv.imwrite("./runs/J599/result.jpg", img_roi)
if in is_ok:
print("有插针歪斜，不合格")
return "有插针歪斜，不合格"
else:
print("没有插针歪斜，产品合格")
return "没有插针歪斜，产品合格"
if __name__ == "__main__":
reslut = j_4_holes_dectWX("images/Final/E_0_8.jpg","J599-4holes_template.jpg")
print(reslut)
#
# # # #holes
# img = cv.imread("images/Final/E_0_8.jpg", 1)
# # img_roi = img[:2027, 1713:2751]
# # img_roi = img[:2224, 1515:2940]
# img_roi = img[:2234, 1480:2950]
# cv.imwrite("J599-4holes_template.jpg",img_roi)
#
# # cv.namedWindow("img_roi",2)
# # cv.imshow("img_roi", img_roi)
# # cv.waitKey()
# if img_roi is None:
# print("no image")
# else:
# # HSV二值化
# img_roi = cv.medianBlur(img_roi, 5) # 中值滤波
# outx, outy, outR = out_circle_detect(img_roi, img)
# print(outx, outy, outR )
# inx, iny = inner_circle_detect(img_roi, img)
# if len(inx) == or outx == 0:
# print("没检测到位置")
# else:
# cv.circle(img_roi, (outx, outy), outR, (0, 0, 255), 3)
#
# for k in range(, len(inx)):
# # 计算两者的距离
# # 假设通过标定其一个像素代表.0056mm
# # 两者相互之间的距离为（mm）:.311053946788194
# # 两者相互之间的距离为（mm）:.163550379629067
# # 两者相互之间的距离为（mm）:.95984457900917
# # 两者相互之间的距离为（mm）:.977940966613671
# # 平均值为.103 所以其阈值为9.103 + 0.5
# DPI =.0198
# dis = math.sqrt(math.pow(outx - inx[k],) + math.pow(outy - iny[k], 2))
# dis *= DPI
# # print("两者相互之间的距离为（mm）:", dis)
# if dis >.603:
# cv.circle(img_roi, (inx[k], iny[k]), 20, (0, 0, 255), 3)
# print("有插针歪斜，不合格")
# else:
# cv.circle(img_roi, (inx[k], iny[k]), 8, (0, 255, 0), -1)
# print("没有插针歪斜，产品合格")
#
# cv.namedWindow("image", 2)
# cv.imshow("image",img_roi)
# cv.waitKey()
