如何在OpenCV Python中实现基于FLANN的特征匹配？

 我们使用尺度不变特征变换（ SIFT ）和Fast Library for Approx Nearest Neighbor（ FLANN ）来实现两个图像之间的特征匹配。使用SIFT来找到特征关键点和描述符。使用基于FLANN的KNN匹配器来匹配两个图像中的描述符。我们使用 cv2.FlannBasedMatcher（） 作为基于FLANN的匹配器。

步骤

要使用尺度不变特征变换（ SIFT ）特征检测器和基于FLANN的匹配器实现两个图像之间的特征匹配，您可以按照以下步骤操作：

• 导入所需库 OpenCV、Matplotlib 和 NumPy 。确保您已经安装了这些库。

• 使用 cv2.imread（） 方法读取两个灰度图像。指定图像的完整路径。

• 使用默认值初始化SIFT对象， sift=cv2.SIFT_create（） 。

• 使用 sift.detectAndCompute（） 在两个输入图像中检测并计算关键点' kp1 '和' kp2 '以及描述符' des1 '和' des2 '。

• 创建基于FLANN的匹配器对象， flann = cv2.FlannBasedMatcher（） 并使用 flann.knnMatch（des1，des2，k=2） 匹配描述符。它返回匹配项。对匹配项进行比率测试以获得最佳匹配项。使用 cv2.drawMatchesKnn（） 绘制匹配项。

• 可视化关键点匹配。

让我们来看一些使用尺度不变特征变换（ SIFT ）特征检测器和基于FLANN的匹配器匹配两个图像的关键点的示例。

输入图像

我们在下面的示例中使用以下图像作为输入文件。

示例

在本例中，我们使用SIFT算法检测两个输入图像的关键点和描述符，并使用基于FLANN的匹配器和KNN匹配算法匹配描述符，同时应用比率测试以仅查找好的匹配项。我们还绘制关键点和匹配项。

<span style="color:#212121"><span style="background-color:#ffffff"><span style="background-color:#272822"><span style="color:#f8f8d4"><code class="language-python"><span style="color:#999999"><span style="color:#93a1a1"># 导入所需库</span></span>
<span style="color:#cc99cd"><span style="color:#f92659">import</span></span><span style="color:#66d9ef"> numpy </span><span style="color:#cc99cd"><span style="color:#f92659">as</span></span><span style="color:#66d9ef"> np
</span><span style="color:#cc99cd"><span style="color:#f92659">import</span></span><span style="color:#66d9ef"> cv2
</span><span style="color:#cc99cd"><span style="color:#f92659">from</span></span><span style="color:#66d9ef"> matplotlib </span><span style="color:#cc99cd"><span style="color:#f92659">import</span></span><span style="color:#66d9ef"> pyplot </span><span style="color:#cc99cd"><span style="color:#f92659">as</span></span><span style="color:#66d9ef"> plt

</span><span style="color:#999999"><span style="color:#93a1a1"># 读取两张输入图像</span></span><span style="color:#66d9ef">
img1 </span><span style="color:#67cdcc"><span style="color:#f8f8f2">=</span></span><span style="color:#66d9ef"> cv2</span><span style="color:#cccccc"><span style="color:#f8f8f2">.</span></span><span style="color:#66d9ef">imread</span><span style="color:#cccccc"><span style="color:#f8f8f2">(</span></span><span style="color:#7ec699"><span style="color:#e6db74">'car.jpg'</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#f08d49"><span style="color:#ae81ff">0</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">)</span></span><span style="color:#66d9ef">
img2 </span><span style="color:#67cdcc"><span style="color:#f8f8f2">=</span></span><span style="color:#66d9ef"> cv2</span><span style="color:#cccccc"><span style="color:#f8f8f2">.</span></span><span style="color:#66d9ef">imread</span><span style="color:#cccccc"><span style="color:#f8f8f2">(</span></span><span style="color:#7ec699"><span style="color:#e6db74">'car-rotated-crop.jpg'</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#f08d49"><span style="color:#ae81ff">0</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">)</span></span>

<span style="color:#999999"><span style="color:#93a1a1"># 初始化SIFT检测器</span></span><span style="color:#66d9ef">
sift </span><span style="color:#67cdcc"><span style="color:#f8f8f2">=</span></span><span style="color:#66d9ef"> cv2</span><span style="color:#cccccc"><span style="color:#f8f8f2">.</span></span><span style="color:#66d9ef">SIFT_create</span><span style="color:#cccccc"><span style="color:#f8f8f2">(</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">)</span></span>

<span style="color:#999999"><span style="color:#93a1a1"># 找到关键点和描述符</span></span><span style="color:#66d9ef">
kp1</span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#66d9ef"> des1 </span><span style="color:#67cdcc"><span style="color:#f8f8f2">=</span></span><span style="color:#66d9ef"> sift</span><span style="color:#cccccc"><span style="color:#f8f8f2">.</span></span><span style="color:#66d9ef">detectAndCompute</span><span style="color:#cccccc"><span style="color:#f8f8f2">(</span></span><span style="color:#66d9ef">img1</span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#f08d49"><span style="color:#f92659">None</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">)</span></span><span style="color:#66d9ef">
kp2</span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#66d9ef"> des2 </span><span style="color:#67cdcc"><span style="color:#f8f8f2">=</span></span><span style="color:#66d9ef"> sift</span><span style="color:#cccccc"><span style="color:#f8f8f2">.</span></span><span style="color:#66d9ef">detectAndCompute</span><span style="color:#cccccc"><span style="color:#f8f8f2">(</span></span><span style="color:#66d9ef">img2</span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#f08d49"><span style="color:#f92659">None</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">)</span></span>

<span style="color:#999999"><span style="color:#93a1a1"># FLANN参数</span></span><span style="color:#66d9ef">
FLANN_INDEX_KDTREE </span><span style="color:#67cdcc"><span style="color:#f8f8f2">=</span></span> <span style="color:#f08d49"><span style="color:#ae81ff">0</span></span><span style="color:#66d9ef">
index_params </span><span style="color:#67cdcc"><span style="color:#f8f8f2">=</span></span> <span style="color:#cc99cd"><span style="color:#66d9ef">dict</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">(</span></span><span style="color:#66d9ef">algorithm </span><span style="color:#67cdcc"><span style="color:#f8f8f2">=</span></span><span style="color:#66d9ef"> FLANN_INDEX_KDTREE</span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#66d9ef"> trees </span><span style="color:#67cdcc"><span style="color:#f8f8f2">=</span></span> <span style="color:#f08d49"><span style="color:#ae81ff">5</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">)</span></span><span style="color:#66d9ef">
search_params </span><span style="color:#67cdcc"><span style="color:#f8f8f2">=</span></span> <span style="color:#cc99cd"><span style="color:#66d9ef">dict</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">(</span></span><span style="color:#66d9ef">checks</span><span style="color:#67cdcc"><span style="color:#f8f8f2">=</span></span><span style="color:#f08d49"><span style="color:#ae81ff">50</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">)</span></span> <span style="color:#999999"><span style="color:#93a1a1"># 或传入空字典</span></span>

<span style="color:#999999"><span style="color:#93a1a1"># 应用基于FLANN的knn匹配器</span></span><span style="color:#66d9ef">
flann </span><span style="color:#67cdcc"><span style="color:#f8f8f2">=</span></span><span style="color:#66d9ef"> cv2</span><span style="color:#cccccc"><span style="color:#f8f8f2">.</span></span><span style="color:#a6e22e">FlannBasedMatcher</span><span style="color:#cccccc"><span style="color:#f8f8f2">(</span></span><span style="color:#66d9ef">index_params</span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#66d9ef">search_params</span><span style="color:#cccccc"><span style="color:#f8f8f2">)</span></span><span style="color:#66d9ef">
matches </span><span style="color:#67cdcc"><span style="color:#f8f8f2">=</span></span><span style="color:#66d9ef"> flann</span><span style="color:#cccccc"><span style="color:#f8f8f2">.</span></span><span style="color:#66d9ef">knnMatch</span><span style="color:#cccccc"><span style="color:#f8f8f2">(</span></span><span style="color:#66d9ef">des1</span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#66d9ef">des2</span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#66d9ef">k</span><span style="color:#67cdcc"><span style="color:#f8f8f2">=</span></span><span style="color:#f08d49"><span style="color:#ae81ff">2</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">)</span></span>

<span style="color:#999999"><span style="color:#93a1a1"># 只需绘制好的匹配项，因此创建掩码</span></span><span style="color:#66d9ef">
matchesMask </span><span style="color:#67cdcc"><span style="color:#f8f8f2">=</span></span> <span style="color:#cccccc"><span style="color:#f8f8f2">[</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">[</span></span><span style="color:#f08d49"><span style="color:#ae81ff">0</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#f08d49"><span style="color:#ae81ff">0</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">]</span></span> <span style="color:#cc99cd"><span style="color:#f92659">for</span></span><span style="color:#66d9ef"> i </span><span style="color:#cc99cd"><span style="color:#f92659">in</span></span> <span style="color:#cc99cd"><span style="color:#66d9ef">range</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">(</span></span><span style="color:#cc99cd"><span style="color:#66d9ef">len</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">(</span></span><span style="color:#66d9ef">matches</span><span style="color:#cccccc"><span style="color:#f8f8f2">)</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">)</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">]</span></span>

<span style="color:#999999"><span style="color:#93a1a1"># Lowe论文中的比率测试</span></span>
<span style="color:#cc99cd"><span style="color:#f92659">for</span></span><span style="color:#66d9ef"> i</span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">(</span></span><span style="color:#66d9ef">m</span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#66d9ef">n</span><span style="color:#cccccc"><span style="color:#f8f8f2">)</span></span> <span style="color:#cc99cd"><span style="color:#f92659">in</span></span> <span style="color:#cc99cd"><span style="color:#66d9ef">enumerate</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">(</span></span><span style="color:#66d9ef">matches</span><span style="color:#cccccc"><span style="color:#f8f8f2">)</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">:</span></span>
   <span style="color:#cc99cd"><span style="color:#f92659">if</span></span><span style="color:#66d9ef"> m</span><span style="color:#cccccc"><span style="color:#f8f8f2">.</span></span><span style="color:#66d9ef">distance </span><span style="color:#67cdcc"><span style="color:#f8f8f2"><</span></span> <span style="color:#f08d49"><span style="color:#ae81ff">0.1</span></span><span style="color:#67cdcc"><span style="color:#f8f8f2">*</span></span><span style="color:#66d9ef">n</span><span style="color:#cccccc"><span style="color:#f8f8f2">.</span></span><span style="color:#66d9ef">distance</span><span style="color:#cccccc"><span style="color:#f8f8f2">:</span></span><span style="color:#66d9ef">
      matchesMask</span><span style="color:#cccccc"><span style="color:#f8f8f2">[</span></span><span style="color:#66d9ef">i</span><span style="color:#cccccc"><span style="color:#f8f8f2">]</span></span><span style="color:#67cdcc"><span style="color:#f8f8f2">=</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">[</span></span><span style="color:#f08d49"><span style="color:#ae81ff">1</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#f08d49"><span style="color:#ae81ff">0</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">]</span></span><span style="color:#66d9ef">
draw_params </span><span style="color:#67cdcc"><span style="color:#f8f8f2">=</span></span> <span style="color:#cc99cd"><span style="color:#66d9ef">dict</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">(</span></span><span style="color:#66d9ef">matchColor </span><span style="color:#67cdcc"><span style="color:#f8f8f2">=</span></span> <span style="color:#cccccc"><span style="color:#f8f8f2">(</span></span><span style="color:#f08d49"><span style="color:#ae81ff">0</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#f08d49"><span style="color:#ae81ff">255</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#f08d49"><span style="color:#ae81ff">0</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">)</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#66d9ef">singlePointColor </span><span style="color:#67cdcc"><span style="color:#f8f8f2">=</span></span> <span style="color:#cccccc"><span style="color:#f8f8f2">(</span></span><span style="color:#f08d49"><span style="color:#ae81ff">255</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#f08d49"><span style="color:#ae81ff">0</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#f08d49"><span style="color:#ae81ff">0</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">)</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#66d9ef">matchesMask </span><span style="color:#67cdcc"><span style="color:#f8f8f2">=</span></span><span style="color:#66d9ef"> matchesMask</span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#66d9ef">flags </span><span style="color:#67cdcc"><span style="color:#f8f8f2">=</span></span> <span style="color:#f08d49"><span style="color:#ae81ff">0</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">)</span></span><span style="color:#66d9ef">
img3 </span><span style="color:#67cdcc"><span style="color:#f8f8f2">=</span></span><span style="color:#66d9ef"> cv2</span><span style="color:#cccccc"><span style="color:#f8f8f2">.</span></span><span style="color:#66d9ef">drawMatchesKnn</span><span style="color:#cccccc"><span style="color:#f8f8f2">(</span></span><span style="color:#66d9ef">img1</span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#66d9ef">kp1</span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#66d9ef">img2</span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#66d9ef">kp2</span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#66d9ef">matches</span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#f08d49"><span style="color:#f92659">None</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#67cdcc"><span style="color:#f8f8f2">**</span></span><span style="color:#66d9ef">draw_params</span><span style="color:#cccccc"><span style="color:#f8f8f2">)</span></span><span style="color:#66d9ef">
plt</span><span style="color:#cccccc"><span style="color:#f8f8f2">.</span></span><span style="color:#66d9ef">imshow</span><span style="color:#cccccc"><span style="color:#f8f8f2">(</span></span><span style="color:#66d9ef">img3</span><span style="color:#cccccc"><span style="color:#f8f8f2">)</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">,</span></span><span style="color:#66d9ef">plt</span><span style="color:#cccccc"><span style="color:#f8f8f2">.</span></span><span style="color:#66d9ef">show</span><span style="color:#cccccc"><span style="color:#f8f8f2">(</span></span><span style="color:#cccccc"><span style="color:#f8f8f2">)</span></span>
</code></span></span></span></span>

输出结果

当你运行以上的Python程序时，它将生成如下输出窗口：

为帮助更多对人工智能感兴趣的小伙伴们能够有效的系统性的学习以及论文的研究，小编特意制作整理了一份人工智能学习资料给大家，整理了很久，非常全面。

大致内容包括一些人工智能基础入门视频和文档+AI常用框架实战视频、计算机视觉、机器学习、图像识别、NLP、OpenCV、YOLO、pytorch、深度学习与神经网络等学习资料、课件源码、国内外知名精华资源、以及AI热门论文等全套学习资料。

需要以上这些文中提到的资料，请点击此处→【人工智能全套完整VIP资料】即可免费获取。

看完三件事❤️

• 如果你觉得这篇内容对你还蛮有帮助，我想邀请你帮我三个小忙：
• 点赞，转发，有你们的 『点赞和评论』，才是我创造的动力。
• 关注作者公众号 『 python深度学习NLP 』，不定期分享原创知识。
• 关注后回复【666】扫码即可获取学习资料包。
• 同时可以期待后续文章ing🚀。