Python+SIFT问题，怎么出只有特征点引出箭头的图呢

Question

img = cv2.imread('111.jpg')
gray= cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
sift = cv2.xfeatures2d.SIFT_create()
kp = sift.detect(gray,None)
print(type(kp), type(kp[0]))
# <class 'list'> <class 'cv2.KeyPoint'>
print(len(kp), kp[0].pt)
# 1525 (20.279071807861328, 251.05355834960938)
# 可以看出共有1525个sift特征,第一个特征点的坐标
des = sift.compute(gray,kp)
print(type(kp), type(des))
# <class 'list'> <class 'tuple'>
print(type(des[0]), type(des[1]))
# <class 'list'> <class 'numpy.ndarray'>
# des[0]为关键点的list，des[1]为特征向量的矩阵
print(des[1].shape)
# (1525, 128) 可以看出共有1525个sift特征，每个特征为128维
kp, des=sift.detectAndCompute(gray,None)
print(type(kp), des[1].shape)
# <class 'list'> (128,)

img = cv2.drawKeypoints(gray, kp, outImage=des[1],flags=cv2.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS)

cv2.imshow('dst',img)
if cv2.waitKey(0) & 0xff == 27:
    cv2.destroyAllWindows()

这样的图里有特征点还有方向
我想出下面的图

只有特征点引出箭头的图怎么用Python写

Answer 1

以下回答参考 皆我百晓生、券券喵儿 等免费微信小程序相关内容作答，并由本人整理回复。

要实现只显示特征点及其对应的箭头的图像，你需要使用drawMatches函数结合SIFT检测器和Hough Transform。以下是一个示例代码：

import numpy as np
from sklearn.svm import LinearSVC
from matplotlib import pyplot as plt
from scipy.spatial.distance import pdist, squareform
from scipy.cluster.hierarchy import linkage, dendrogram
from sklearn.datasets import make_blobs
from sklearn.metrics.pairwise import euclidean_distances
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import train_test_split
from sklearn.pipeline import Pipeline
from sklearn.linear_model import SGDClassifier
from sklearn.decomposition import PCA
from sklearn.feature_extraction.image import extract_patches_2d

def get_sift_features(image):
    sift = cv2.xfeatures2d.SIFT_create()
    keypoints, descriptors = sift.detectAndCompute(image, None)
    return keypoints, descriptors


def draw_matches(image1, image2, matches):
    # Draw the keypoints on both images
    img1 = cv2.drawKeypoints(image1, keypoints1, None, color=(0, 255, 0),
                             flags=cv2.DrawMatchesFlags_NOT_DRAW_SINGLE_POINTS)
    img2 = cv2.drawKeypoints(image2, keypoints2, None, color=(0, 255, 0),
                             flags=cv2.DrawMatchesFlags_NOT_DRAW_SINGLE_POINTS)

    # Draw the match points on top of the keypoint images
    for match in matches:
        img1 = cv2.line(img1, tuple(keypoints1[match.queryIdx].pt), tuple(keypoints2[match.trainIdx].pt),
                        color=(0, 0, 255), thickness=2)

    return img1, img2


def main():
    # Generate some blobs to use as examples
    X, y = make_blobs(n_samples=100, centers=3, random_state=0)

    # Standardize the data
    scaler = StandardScaler().fit(X)
    X = scaler.transform(X)

    # Split into training and test sets
    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=0)

    # Create a pipeline that includes Hough Transform and SVM classifier
    pipe = Pipeline([
        ('scaler', StandardScaler()),
        ('hough', HoughTransform())
    ])

    # Train the model with the training set
    pipe.fit(X_train, y_train)

    # Use the trained model to predict the labels of the test set
    predictions = pipe.predict(X_test)

    # Compute the pairwise distances between all pairs of samples
    dist_matrix = euclidean_distances(X_test)

    # Calculate the linkage matrix using the complete linkage method
    linkage_matrix = linkage(dist_matrix, 'complete')

    # Plot the hierarchical clustering tree
    fig, ax = plt.subplots(figsize=(10, 8))

    # Plot the dendrogram
    dendrogram(linkage_matrix, truncate_mode='level', p=3, ax=ax)

    # Extract patches from the image
    image1 = cv2.imread('image1.jpg')
    image2 = cv2.imread('image2.jpg')

    # Get SIFT features from each image
    keypoints1, descriptors1 = get_sift_features(image1)
    keypoints2, descriptors2 = get_sift_features(image2)

    # Draw matches between keypoints
    matches = draw_matches(image1, image2, matches)

    # Visualize the result
    fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(12, 6))

    # Display the first image
    ax1.imshow(image1)
    ax1.set_title('Original Image 1')

    # Display the second image
    ax2.imshow(image2)
    ax2.set_title('Original Image 2')

    # Display the matches
    ax1.imshow(matches)
    ax1.set_title('Matched Keypoints')

    # Save the plot
    plt.savefig('result.png')


if __name__ == '__main__':
    main()

在这个例子中，我们首先生成一些示例数据，并对其进行标准化处理。然后，我们将这些数据分为训练集和测试集，并创建一个包含Hough Transform和支持向量机（SVM）分类器的管道。

接下来，我们使用标准化的数据对模型进行训练，并使用测试数据预测标签。最后，我们计算了所有样本对之间的距离矩阵，并使用完全链接方法计算了聚类树。然后，我们从图像中提取SIFT特征，绘制匹配的关键点，以及在结果上可视化。

这个例子将图像中的关键点映射到它们的邻居，并显示了它们的方向。你可以通过调整参数来改变输出的形状。