如何在 HoughLinesP 之后合并行?

问题描述

我的任务是找到线(startX、startY、endX、endY)和矩形(4 线)的坐标.这是输入文件:

我使用下一个代码:

img = cv2.imread(image_src)灰色 = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)ret, thresh1 = cv2.threshold(灰色,127,255,cv2.THRESH_BINARY)边缘 = cv2.Canny(thresh1,50,150,apertureSize = 3)minLineLength = 100最大线间隙 = 10线 = cv2.HoughLinesP(edges,1,np.pi/180,10,minLineLength,maxLineGap)打印(长度(行))对于行中的行:cv2.line(img,(line[0][0],line[0][1]),(line[0][2],line[0][3]),(0,0,255),6)

我得到下一个结果:

从最后一张图片中，您可以看到大量的小红线.

问题:

1. 合并小线条的最佳方法是什么?
2. 为什么有很多HoughLinesP 检测不到的小部分?

我终于完成了流水线:

1. 修正了不正确的参数(Dan 建议)
2. 开发了我自己的合并线段"算法.

   还有 572 行.在我的合并线段"之后，我们只有 89 行

   My task is to find coordinates of lines (startX, startY, endX, endY) and rectangles (4 lines). Here is input file:

   I use the next code:

   img = cv2.imread(image_src)
   gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
   ret, thresh1 = cv2.threshold(gray,127,255,cv2.THRESH_BINARY)
   
   edges = cv2.Canny(thresh1,50,150,apertureSize = 3)
   
   minLineLength = 100
   maxLineGap = 10
   lines = cv2.HoughLinesP(edges,1,np.pi/180,10,minLineLength,maxLineGap)
   print(len(lines))
   for line in lines:
       cv2.line(img,(line[0][0],line[0][1]),(line[0][2],line[0][3]),(0,0,255),6)
   

   I get the next results:

   From the last image you can see big amount of small red lines.

   Questions:

   1. What is the best way to merge small lines?
   2. Why there are a lot of small portions that are not detected by HoughLinesP?
   解决方案

   I have finally completed the pipeline:

   1. fixed incorrect parameters (as were suggested by Dan)
   2. developed my own 'merging line segments' algorithm. I had bad results when I implemented TAVARES and PADILHA algorithm (as were suggested by Andrew).
   3. I have skipped Canny and got better results (as were suggested by Alexander)

   Please find the code and results:

   def get_lines(lines_in):
       if cv2.__version__ < '3.0':
           return lines_in[0]
       return [l[0] for l in lines_in]
   
   
   def process_lines(image_src):
       img = mpimg.imread(image_src)
       gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
   
       ret, thresh1 = cv2.threshold(gray,127,255,cv2.THRESH_BINARY)
   
       thresh1 = cv2.bitwise_not(thresh1)
   
       edges = cv2.Canny(thresh1, threshold1=50, threshold2=200, apertureSize = 3)
   
       lines = cv2.HoughLinesP(thresh1, rho=1, theta=np.pi/180, threshold=50,
                               minLineLength=50, maxLineGap=30)
   
       # l[0] - line; l[1] - angle
       for line in get_lines(lines):
           leftx, boty, rightx, topy = line
           cv2.line(img, (leftx, boty), (rightx,topy), (0,0,255), 6) 
   
       # merge lines
   
       #------------------
       # prepare
       _lines = []
       for _line in get_lines(lines):
           _lines.append([(_line[0], _line[1]),(_line[2], _line[3])])
   
       # sort
       _lines_x = []
       _lines_y = []
       for line_i in _lines:
           orientation_i = math.atan2((line_i[0][1]-line_i[1][1]),(line_i[0][0]-line_i[1][0]))
           if (abs(math.degrees(orientation_i)) > 45) and abs(math.degrees(orientation_i)) < (90+45):
               _lines_y.append(line_i)
           else:
               _lines_x.append(line_i)
   
       _lines_x = sorted(_lines_x, key=lambda _line: _line[0][0])
       _lines_y = sorted(_lines_y, key=lambda _line: _line[0][1])
   
       merged_lines_x = merge_lines_pipeline_2(_lines_x)
       merged_lines_y = merge_lines_pipeline_2(_lines_y)
   
       merged_lines_all = []
       merged_lines_all.extend(merged_lines_x)
       merged_lines_all.extend(merged_lines_y)
       print("process groups lines", len(_lines), len(merged_lines_all))
       img_merged_lines = mpimg.imread(image_src)
       for line in merged_lines_all:
           cv2.line(img_merged_lines, (line[0][0], line[0][1]), (line[1][0],line[1][1]), (0,0,255), 6)
   
   
       cv2.imwrite('prediction/lines_gray.jpg',gray)
       cv2.imwrite('prediction/lines_thresh.jpg',thresh1)
       cv2.imwrite('prediction/lines_edges.jpg',edges)
       cv2.imwrite('prediction/lines_lines.jpg',img)
       cv2.imwrite('prediction/merged_lines.jpg',img_merged_lines)
   
       return merged_lines_all
   
   def merge_lines_pipeline_2(lines):
       super_lines_final = []
       super_lines = []
       min_distance_to_merge = 30
       min_angle_to_merge = 30
   
       for line in lines:
           create_new_group = True
           group_updated = False
   
           for group in super_lines:
               for line2 in group:
                   if get_distance(line2, line) < min_distance_to_merge:
                       # check the angle between lines       
                       orientation_i = math.atan2((line[0][1]-line[1][1]),(line[0][0]-line[1][0]))
                       orientation_j = math.atan2((line2[0][1]-line2[1][1]),(line2[0][0]-line2[1][0]))
   
                       if int(abs(abs(math.degrees(orientation_i)) - abs(math.degrees(orientation_j)))) < min_angle_to_merge: 
                           #print("angles", orientation_i, orientation_j)
                           #print(int(abs(orientation_i - orientation_j)))
                           group.append(line)
   
                           create_new_group = False
                           group_updated = True
                           break
   
               if group_updated:
                   break
   
           if (create_new_group):
               new_group = []
               new_group.append(line)
   
               for idx, line2 in enumerate(lines):
                   # check the distance between lines
                   if get_distance(line2, line) < min_distance_to_merge:
                       # check the angle between lines       
                       orientation_i = math.atan2((line[0][1]-line[1][1]),(line[0][0]-line[1][0]))
                       orientation_j = math.atan2((line2[0][1]-line2[1][1]),(line2[0][0]-line2[1][0]))
   
                       if int(abs(abs(math.degrees(orientation_i)) - abs(math.degrees(orientation_j)))) < min_angle_to_merge: 
                           #print("angles", orientation_i, orientation_j)
                           #print(int(abs(orientation_i - orientation_j)))
   
                           new_group.append(line2)
   
                           # remove line from lines list
                           #lines[idx] = False
               # append new group
               super_lines.append(new_group)
   
   
       for group in super_lines:
           super_lines_final.append(merge_lines_segments1(group))
   
       return super_lines_final
   
   def merge_lines_segments1(lines, use_log=False):
       if(len(lines) == 1):
           return lines[0]
   
       line_i = lines[0]
   
       # orientation
       orientation_i = math.atan2((line_i[0][1]-line_i[1][1]),(line_i[0][0]-line_i[1][0]))
   
       points = []
       for line in lines:
           points.append(line[0])
           points.append(line[1])
   
       if (abs(math.degrees(orientation_i)) > 45) and abs(math.degrees(orientation_i)) < (90+45):
   
           #sort by y
           points = sorted(points, key=lambda point: point[1])
   
           if use_log:
               print("use y")
       else:
   
           #sort by x
           points = sorted(points, key=lambda point: point[0])
   
           if use_log:
               print("use x")
   
       return [points[0], points[len(points)-1]]
   
   # https://docs.scipy.org/doc/scipy/reference/generated/scipy.spatial.distance.cdist.html
   # https://stackoverflow.com/questions/32702075/what-would-be-the-fastest-way-to-find-the-maximum-of-all-possible-distances-betw
   def lines_close(line1, line2):
       dist1 = math.hypot(line1[0][0] - line2[0][0], line1[0][0] - line2[0][1])
       dist2 = math.hypot(line1[0][2] - line2[0][0], line1[0][3] - line2[0][1])
       dist3 = math.hypot(line1[0][0] - line2[0][2], line1[0][0] - line2[0][3])
       dist4 = math.hypot(line1[0][2] - line2[0][2], line1[0][3] - line2[0][3])
   
       if (min(dist1,dist2,dist3,dist4) < 100):
           return True
       else:
           return False
   
   def lineMagnitude (x1, y1, x2, y2):
       lineMagnitude = math.sqrt(math.pow((x2 - x1), 2)+ math.pow((y2 - y1), 2))
       return lineMagnitude
   
   #Calc minimum distance from a point and a line segment (i.e. consecutive vertices in a polyline).
   # https://nodedangles.wordpress.com/2010/05/16/measuring-distance-from-a-point-to-a-line-segment/
   # http://paulbourke.net/geometry/pointlineplane/
   def DistancePointLine(px, py, x1, y1, x2, y2):
       #http://local.wasp.uwa.edu.au/~pbourke/geometry/pointline/source.vba
       LineMag = lineMagnitude(x1, y1, x2, y2)
   
       if LineMag < 0.00000001:
           DistancePointLine = 9999
           return DistancePointLine
   
       u1 = (((px - x1) * (x2 - x1)) + ((py - y1) * (y2 - y1)))
       u = u1 / (LineMag * LineMag)
   
       if (u < 0.00001) or (u > 1):
           #// closest point does not fall within the line segment, take the shorter distance
           #// to an endpoint
           ix = lineMagnitude(px, py, x1, y1)
           iy = lineMagnitude(px, py, x2, y2)
           if ix > iy:
               DistancePointLine = iy
           else:
               DistancePointLine = ix
       else:
           # Intersecting point is on the line, use the formula
           ix = x1 + u * (x2 - x1)
           iy = y1 + u * (y2 - y1)
           DistancePointLine = lineMagnitude(px, py, ix, iy)
   
       return DistancePointLine
   
   def get_distance(line1, line2):
       dist1 = DistancePointLine(line1[0][0], line1[0][1], 
                                 line2[0][0], line2[0][1], line2[1][0], line2[1][1])
       dist2 = DistancePointLine(line1[1][0], line1[1][1], 
                                 line2[0][0], line2[0][1], line2[1][0], line2[1][1])
       dist3 = DistancePointLine(line2[0][0], line2[0][1], 
                                 line1[0][0], line1[0][1], line1[1][0], line1[1][1])
       dist4 = DistancePointLine(line2[1][0], line2[1][1], 
                                 line1[0][0], line1[0][1], line1[1][0], line1[1][1])
   
   
       return min(dist1,dist2,dist3,dist4)
   

   There are still 572 lines. After my "merging line segments" we have only 89 lines

   这篇关于如何在 HoughLinesP 之后合并行?的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持跟版网！