海洋图像水平线检测：Hough变换容差设置及检测失效解决方案咨询

Hey there! Let's work through your ocean horizon detection problem—those tricky missing horizontal lines are definitely solvable with some parameter tweaks and preprocessing adjustments. Here's a breakdown of how to address your questions:

Absolutely, angle tolerance is a key part of the solution, but you’ll also need to pair it with adjustments for line gaps (since that’s one of your root causes). Let’s dive into how to set this up properly with OpenCV’s Hough transforms.

For Standard Hough Transform (cv2.HoughLines()) #

This method returns lines in polar coordinates (rho, theta). You can narrow down to horizontal lines by filtering for angles close to 0 or π radians (0° or 180°) with a small tolerance:

import cv2
import numpy as np

# After getting Canny edges
lines = cv2.HoughLines(edges, rho=1, theta=np.pi/1800, threshold=50)
if lines is not None:
    for line in lines:
        rho, theta = line[0]
        # Keep only lines within ±0.1 radians of horizontal (adjust tolerance as needed)
        if abs(theta - 0) < 0.1 or abs(theta - np.pi) < 0.1:
            # Calculate endpoints to draw the line
            a = np.cos(theta)
            b = np.sin(theta)
            x0 = a * rho
            y0 = b * rho
            x1 = int(x0 + 1000 * (-b))
            y1 = int(y0 + 1000 * (a))
            x2 = int(x0 - 1000 * (-b))
            y2 = int(y0 - 1000 * (a))
            cv2.line(your_image, (x1, y1), (x2, y2), (0, 0, 255), 2)

• The theta parameter sets angle resolution (smaller = more precise). Using np.pi/1800 gives you 0.1° increments, which helps catch lines slightly off perfect horizontal.

For Probabilistic Hough Transform (cv2.HoughLinesP()) #

This method returns line segments directly, so you’ll calculate each segment’s angle and filter for horizontal ones. Plus, use the maxLineGap parameter to bridge small gaps in Canny edges—this is critical for your gap issue!

lines = cv2.HoughLinesP(
    edges,
    rho=1,
    theta=np.pi/180,
    threshold=30,
    minLineLength=50,  # Minimum length of a valid line
    maxLineGap=50      # Allow gaps up to 50px to be connected into one line
)
if lines is not None:
    for line in lines:
        x1, y1, x2, y2 = line[0]
        # Calculate the line's angle in radians
        angle = np.arctan2(y2 - y1, x2 - x1)
        # Filter for near-horizontal lines
        if abs(angle) < 0.1 or abs(angle - np.pi) < 0.1:
            cv2.line(your_image, (x1, y1), (x2, y2), (0, 255, 0), 2)

• Tweak maxLineGap based on your images: if gaps are larger, increase this value (e.g., to 70 or 100).

Your current preprocessing (only erosion) might be making gaps worse—here’s how to adjust:

• Replace erosion with closing operation: Closing (dilate then erode) fills small gaps in edges while preserving overall shape.

  kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (3, 3))
  closed_edges = cv2.morphologyEx(canny_edges, cv2.MORPH_CLOSE, kernel)
  

• Tweak Canny parameters: If your edges are too broken, lower the lower threshold or add a Gaussian blur before Canny to reduce noise:

  blurred = cv2.GaussianBlur(gray_image, (5, 5), 0)
  canny_edges = cv2.Canny(blurred, lower_threshold=40, upper_threshold=120)
  

• Fit a single line to multiple segments: If the horizon isn’t perfectly straight, collect all near-horizontal segments and fit a single line using least squares:

  points = []
  if lines is not None:
      for line in lines:
          x1, y1, x2, y2 = line[0]
          angle = np.arctan2(y2 - y1, x2 - x1)
          if abs(angle) < 0.1 or abs(angle - np.pi) < 0.1:
              points.append((x1, y1))
              points.append((x2, y2))
  # Fit a line to all collected points
  if points:
      points = np.array(points)
      [vx, vy, x0, y0] = cv2.fitLine(points, cv2.DIST_L2, 0, 0.01, 0.01)
      # Calculate endpoints for drawing
      lefty = int((-x0 * vy / vx) + y0)
      righty = int(((your_image.shape[1] - x0) * vy / vx) + y0)
      cv2.line(your_image, (your_image.shape[1]-1, righty), (0, lefty), (255, 0, 0), 2)
  

内容的提问来源于stack exchange，提问作者latida