RANSAC(Randon Sample Consensus) 예시 코드

import numpy as np
import matplotlib.pyplot as plt

def generate_data(num_points=100, outlier_ratio=0.2, seed=42):
    np.random.seed(seed)
    x = np.random.rand(num_points) * 10
    y = 2 * x + 1 + np.random.randn(num_points)
    
    # Add outliers
    num_outliers = int(outlier_ratio * num_points)
    outliers_index = np.random.choice(num_points, num_outliers, replace=False)
    y[outliers_index] += 10 * np.random.randn(num_outliers)
    
    return x, y

def fit_line_ransac(x, y, num_iterations=100, threshold=1.0):
    best_model = None
    best_inliers = None
    
    for _ in range(num_iterations):
        # Randomly sample two points
        sample_indices = np.random.choice(len(x), size=2, replace=False)
        x_sample = x[sample_indices]
        y_sample = y[sample_indices]
        
        # Fit a line to the sampled points
        model = np.polyfit(x_sample, y_sample, deg=1)
        
        # Calculate distances to the line for all points
        distances = np.abs(y - (model[0] * x + model[1]))
        
        # Count inliers (points within the threshold)
        inliers = np.where(distances < threshold)[0]
        
        # Update the best model if the current model has more inliers
        if len(inliers) > len(best_inliers):
            best_model = model
            best_inliers = inliers
    
    return best_model, best_inliers

def plot_ransac_result(x, y, model, inliers):
    plt.scatter(x, y, label='Data', color='blue')
    plt.scatter(x[inliers], y[inliers], label='Inliers', color='green')
    
    if model is not None:
        x_line = np.linspace(min(x), max(x), 100)
        y_line = model[0] * x_line + model[1]
        plt.plot(x_line, y_line, label='RANSAC Model', color='red')
    
    plt.legend()
    plt.show()

# Generate synthetic data with outliers
x_data, y_data = generate_data()

# Apply RANSAC algorithm to fit a line
best_fit, inliers_indices = fit_line_ransac(x_data, y_data)

# Plot the original data, inliers, and the RANSAC model
plot_ransac_result(x_data, y_data, best_fit, inliers_indices)