基于相机模拟信号生成的图像数据集的椭圆拟合技术咨询

Hey there! Let's work through a practical, easy-to-implement solution for fitting an ellipse to the image generated by your camera signal simulation code. The image you're dealing with is a scaled circle (which behaves like an ellipse), so we can use computer vision tools to extract and fit it accurately.

Step 1: Preprocess the Image #

First, let's tweak your existing code to get a binary image that works well with contour detection. Your current data array has non-zero values where the ellipse line is—we'll convert this to a clear binary format (black background, white ellipse):

import cairo
import numpy
import matplotlib.pyplot as plt
import cv2

# Your original image generation code
data = numpy.zeros((352, 352), dtype = numpy.uint8)
surface = cairo.ImageSurface.create_for_data( data, cairo.FORMAT_A8, 352, 352)
cr = cairo.Context(surface)
cr.scale(0.85, 1.15)
cr.arc(200, 150, 100, 0, 2. * numpy.pi)
cr.set_line_width(15)
cr.stroke()

# Convert to binary image (ellipse = white, background = black)
binary_img = numpy.where(data > 0, 255, 0).astype(numpy.uint8)

Step 2: Extract the Ellipse Contour #

Next, we'll use OpenCV to find the outline of the ellipse. Since there's only one distinct shape in the image, this is straightforward:

# Find external contours in the binary image
contours, _ = cv2.findContours(binary_img, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

# Grab the only contour (our ellipse)
ellipse_contour = contours[0]

Step 3: Fit the Ellipse #

OpenCV's cv2.fitEllipse() function handles all the heavy lifting here. It takes the contour points and returns three key parameters:

• Center: (x, y) coordinates of the ellipse's center
• Axes: (major axis length, minor axis length)
• Angle: Rotation angle of the ellipse (in degrees)

# Fit the ellipse to the contour
ellipse = cv2.fitEllipse(ellipse_contour)
center, axes, angle = ellipse

# Print parameters for verification
print(f"Ellipse Center: {center}")
print(f"Major/Minor Axes: {axes}")
print(f"Rotation Angle: {angle} degrees")

Step 4: Visualize the Fitted Result #

Let's overlay the fitted ellipse on the original image to confirm it's a perfect match:

# Convert binary image to color for drawing
color_img = cv2.cvtColor(binary_img, cv2.COLOR_GRAY2BGR)

# Draw the fitted ellipse (green, thickness 2)
cv2.ellipse(color_img, ellipse, (0, 255, 0), 2)

# Display the result
plt.imshow(cv2.cvtColor(color_img, cv2.COLOR_BGR2RGB))
plt.title("Original Ellipse with Fitted Overlay")
plt.axis('off')
plt.show()

Alternative: Manual Least Squares Fitting #

If you want a more hands-on approach without OpenCV, you can implement a least squares method to fit the general ellipse equation (Ax^2 + Bxy + Cy^2 + Dx + Ey + F = 0) to the contour points. Here's a simplified snippet:

def fit_ellipse_least_squares(points):
    # Extract x and y coordinates from contour
    x = points[:, 0, 0]
    y = points[:, 0, 1]
    
    # Build design matrix for the ellipse equation
    X = numpy.column_stack((x**2, x*y, y**2, x, y, numpy.ones_like(x)))
    
    # Solve using SVD to get coefficients
    _, _, V = numpy.linalg.svd(X)
    coeffs = V[-1, :]
    
    # Convert coefficients to standard ellipse parameters (requires additional math)
    A, B, C, D, E, F = coeffs
    # Insert code here to compute center, axes, and angle from coefficients
    return center, axes, angle

# Usage
center, axes, angle = fit_ellipse_least_squares(ellipse_contour)

Note: The manual method requires extra math to convert the general coefficients to readable parameters, so OpenCV's method is recommended for most real-world use cases.

内容的提问来源于stack exchange，提问作者The Dude