├── 40252bbc6e20560076b9c4e223dd38af.png ├── 569e12a5e669a595f1b0eab12fa976df.png ├── 6124b1f84adf69ccc173a39e28b67384.png ├── 8f4c0aaca2845c770ec874df00913160.png ├── 93ed7db9f0643e083032fbc84d8d8de1.png ├── README.md ├── abfa5c7db024793c18bb165d09b887e5.png ├── b8e7e07c4c9ee5bcd4f124367568f477.png ├── cv2_fillPoly_mask_pt.py ├── cv2_line_img_tuple_l.py ├── db848810a9f691a8c45479a480cb6541.png ├── edge_img_cv2_Canny_g.py ├── gaussian_cv2_Gaussia.py ├── gray_img_cv2_cvtColo.py ├── idx_np_argmax_diff_.py ├── img_mask_cv2_bitwise.py ├── lines_cv2_HoughLines.py ├── mask_np_zeros_like_g.py └── poly_pts_numpy_array.py /40252bbc6e20560076b9c4e223dd38af.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/qunshansj/opencv-python-video-lane-line-detection/94a6be491f80c3258c3015b510a92284c17b5190/40252bbc6e20560076b9c4e223dd38af.png -------------------------------------------------------------------------------- /569e12a5e669a595f1b0eab12fa976df.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/qunshansj/opencv-python-video-lane-line-detection/94a6be491f80c3258c3015b510a92284c17b5190/569e12a5e669a595f1b0eab12fa976df.png -------------------------------------------------------------------------------- /6124b1f84adf69ccc173a39e28b67384.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/qunshansj/opencv-python-video-lane-line-detection/94a6be491f80c3258c3015b510a92284c17b5190/6124b1f84adf69ccc173a39e28b67384.png -------------------------------------------------------------------------------- /8f4c0aaca2845c770ec874df00913160.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/qunshansj/opencv-python-video-lane-line-detection/94a6be491f80c3258c3015b510a92284c17b5190/8f4c0aaca2845c770ec874df00913160.png -------------------------------------------------------------------------------- /93ed7db9f0643e083032fbc84d8d8de1.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/qunshansj/opencv-python-video-lane-line-detection/94a6be491f80c3258c3015b510a92284c17b5190/93ed7db9f0643e083032fbc84d8d8de1.png -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # 1.图片演示 2 | ![2.png](93ed7db9f0643e083032fbc84d8d8de1.png) 3 | 4 | ![3.png](8f4c0aaca2845c770ec874df00913160.png) 5 | 6 | 7 | 8 | ![5.png](6124b1f84adf69ccc173a39e28b67384.png) 9 | 10 | 11 | 12 | # 2.视频演示 13 | [Python基于OpenCV的视频车道线检测系统[源码＆部署教程]](https://www.bilibili.com/video/BV1ZP41137Ny?spm_id_from=333.999.0.0&vd_source=bc9aec86d164b67a7004b996143742dc) 14 | 15 | # 3.边缘检测 16 | [参考该博客方案](https://mbd.pub/o/bread/Y5WamZtp)，其算法思想如下： 17 | 1. 使用高斯模糊，去除噪音点（cv2.GaussianBlur） 18 | 2. 灰度转换（cv2.cvtColor） 19 | 3. 使用sobel算子，计算出每个点的梯度大小和梯度方向 20 | 4. 使用非极大值抑制(只有最大的保留)，消除边缘检测带来的杂散效应 21 | 5. 应用双阈值，来确定真实和潜在的边缘 22 | 6. 通过抑制弱边缘来完成最终的边缘检测 23 | #### 1.高斯滤波 24 | ``` 25 | gaussian = cv2.GaussianBlur(color_img, (gaussian_ksize,gaussian_ksize), gaussian_sigmax) 26 | ``` 27 | color_img 输入图片 28 | gaussian_ksize 高斯核大小，可以为方形矩阵，也可以为矩形 29 | gaussian_sigmax X方向上的高斯核标准偏差 30 | 31 | #### 2.图片转换 32 | ``` 33 | gray_img = cv2.cvtColor(input_image, flag) 34 | ``` 35 | 用于颜色空间转换。input_image为需要转换的图片，flag为转换的类型，返回值为颜色空间转换后的图片矩阵。flag对应： 36 | cv2.COLOR_BGR2GRAY BGR -> Gray 37 | cv2.COLOR_BGR2RGB BGR -> RGB 38 | cv2.COLOR_BGR2HSV BGR -> HSV 39 | 可以了解一下该函数和cv2.imread()的区别 40 | 41 | #### 3.边缘检测 42 | ``` 43 | edge_img = cv2.Canny(gray_img,canny_threshold1,canny_threshold2) 44 | ``` 45 | imag为所操作的图片，threshold1为下阈值，threshold2为上阈值，返回值为边缘图。 46 | ![6.png](40252bbc6e20560076b9c4e223dd38af.png) 47 | 48 | # 4.ROI and mask 49 | 利用数组选取ROI(region of interest,感性趣的区域，然后与原图进行布尔运算(与运算)。 50 | ![7.png](db848810a9f691a8c45479a480cb6541.png) 51 | ``` 52 | poly_pts = numpy.array([[[0,368],[300,210],[340,210],[640,368]]]) 53 | ``` 54 | 四个数组创建后保存在变量poly_pts中。 55 | ``` 56 | mask = np.zeros_like(gray_img) 57 | ``` 58 | 构建一个与gray_img同维度的数组，并初始化所有变量为零。 59 | ``` 60 | cv2.fillPoly(mask, pts, color) 61 | ``` 62 | 绘制多边形函数。mask为绘制对像，pts为绘制范围，color为绘制颜色。 63 | ``` 64 | img_mask = cv2.bitwise_and(gray_img, mask) 65 | ``` 66 | 与运算 67 | ![8.png](569e12a5e669a595f1b0eab12fa976df.png) 68 | 69 | # 5.霍夫变换 70 | ``` 71 | lines = cv2.HoughLinesP(edge_img, 1, np.pi / 180, 15, minLineLength=40, maxLineGap=20) 72 | ``` 73 | edge_img: 要检测的图片矩阵 74 | 参数2: 距离r的精度,值越大,考虑越多的线 75 | 参数3: 距离theta的精度,值越大,考虑越多的线 76 | 参数4: 累加数阈值,值越小,考虑越多的线 77 | minLineLength: 最短长度阈值,短于这个长度的线会被排除 78 | maxLineGap:同一直线两点之间的最大距离 79 | 返回值:: 80 | [np.array([[x_1,y_1, x_2, y_2]]), 81 | np.array([[x_1, y_1, x_2, y_2]]), 82 | …, 83 | np.array([[x_1, y_1, x_2, y_2]])] 84 | 85 | # 6.离群值过滤 86 | [受该博客思路启发，循环查找超出设定范围的斜率，并去除](https://afdian.net/item?plan_id=2f4d3d58625111ed88d152540025c377)。 87 | ``` 88 | idx = np.argmax(diff) 89 | ``` 90 | 当diff中存放的是一维时，找到diff中最大的数值，并返回该值所在位置。 91 | 从列表lines中把第idx数去除 92 | 93 | # 7.车道线标注 94 | ``` 95 | cv2.line(img, tuple(line[0]), tuple(line[1]), color,thickness) 96 | ``` 97 | 在图片上画线，img为所标注的目标图片，ine[0]为起点坐标，line[1]为终点坐标，thickness为线宽。 98 | 注意两个坐标需要为tuple元组格式，即tuple(line[0]) 99 | ![5.png](abfa5c7db024793c18bb165d09b887e5.png) 100 | 101 | # 8.系统整合 102 | 下图[完整源码＆环境部署视频教程](https://s.xiaocichang.com/s/4149b5) 103 | ![1.png](b8e7e07c4c9ee5bcd4f124367568f477.png) 104 | 参考博客[《Python基于OpenCV的视频车道线检测系统\[源码＆部署教程\]》](https://mbd.pub/o/qunma/work) 105 | 106 | 107 | 108 | # 9.参考文献 109 | * *[1]*[基于改进Hough变换的车道线识别](https://kns.cnki.net/kcms/detail/detail.aspx?filename=SXDS202106012&dbcode=CJFD&dbname=CJFD2021&v=hJej5pxuFEzHMB9I1foLJySODrysESKlYX7GPFWGeFM0eQK6ydtbVt-sjHQusG3p)[J]. 辛敏,罗山. 山西电子技术. 2021(06) 110 | * *[2]*[基于改进Hough变换的车道线检测](https://kns.cnki.net/kcms/detail/detail.aspx?filename=SXQC202107006&dbcode=CJFD&dbname=CJFD2021&v=CRV_0b2A8SFBvI527gOI77pgQH5AFi6jp_NZQMaFr9oukez6f2ScZwN4wmUowpzl)[J]. 李后隆,马骊溟,钟林伟. 汽车实用技术. 2021(07) 111 | * *[3]*[基于改进Hough变换与双点去除R-最小二乘法的车道线检测优化算法](https://kns.cnki.net/kcms/detail/detail.aspx?filename=KXJS202005054&dbcode=CJFD&dbname=CJFD2020&v=7ufNVsW8Ho_uLhyvsFSw5-mf7iGOO_vXnlTzbG16VHfqrtrQg1rYbL4J_HDwpgy5)[J]. 姜立标,李静轩. 科学技术与工程. 2020(05) 112 | * *[4]*[基于改进Hough变换的结构化道路车道线识别](https://kns.cnki.net/kcms/detail/detail.aspx?filename=KXJS202026041&dbcode=CJFD&dbname=CJFD2020&v=7ufNVsW8Ho9RWCMJdJaSDgPuVVh7ggc2C2ILoObrln4nQR45Sj15riiRslo1j8Q1)[J]. 陈政宏,李爱娟,王希波,葛庆英,韩文尧,刘刚. 科学技术与工程. 2020(26) 113 | * *[5]*[基于叠加约束的Hough变换车道线检测研究](https://kns.cnki.net/kcms/detail/detail.aspx?filename=YXDJ201903010&dbcode=CJFD&dbname=CJFD2019&v=DPo6jhEIGONyMtd8rnHM9rBNICTYwQyBZCztSp2pr9i_pme5RdeEr8a44gliPvER)[J]. 伍玲玲,林志贤,郭太良. 有线电视技术. 2019(03) 114 | * *[6]*[基于Hough变换的车道线检测算法设计](https://kns.cnki.net/kcms/detail/detail.aspx?filename=SDLG201911022&dbcode=CJFD&dbname=CJFD2019&v=Vwalx3XAwkDr43eOtyamaxiSN_u4WLX0t6XkDxeoYNsxZZLRLpdn4cCPy_mHTs-5)[J]. 何旭光,江磊,罗一平,张伟伟. 农业装备与车辆工程. 2019(11) 115 | * *[7]*[基于Hough变换的车道线检测](https://kns.cnki.net/kcms/detail/detail.aspx?filename=SDGJ201710236&dbcode=CJFD&dbname=CJFD2017&v=RI4_BvRfjVhEGSp-9kOzxxa_V2E5MdVjA5I10eLIgESNg9QMCUiELIMT_xr5iZxB)[J]. 钱怡. 山东工业技术. 2017(10) 116 | * *[8]*[Hough变换的车道线并行检测](https://kns.cnki.net/kcms/detail/detail.aspx?filename=DZKK201512028&dbcode=CJFD&dbname=CJFD2015&v=iN1md_QPjdN5J-W8ZYWixoTeIAG70FQwzF5bDXgeU9oBwPxbWA31P-IZ9WqfsTd2)[J]. 王赛,徐伯庆,苏栋骐. 电子科技. 2015(12) 117 | * *[9]*[一种改进Hough变换的车道线检测算法](https://kns.cnki.net/kcms/detail/detail.aspx?filename=JSSG201411040&dbcode=CJFD&dbname=CJFD2014&v=V_nRtn35IOkNRGJyeD0anQkAMB8Ltk6pPFBMa5kEN-4qbJX1wAXd0ZQkX3ZUJb2K)[J]. 王全. 计算机与数字工程. 2014(11) 118 | * *[10]*[一种基于 Hough 变换的车道线检测新方法](https://kns.cnki.net/kcms/detail/detail.aspx?filename=OXZG201304001&dbcode=CJFD&dbname=CJFD2013&v=AKKUylPNebSJwYMBkybvm0GWlbo9Yqq9-AYxLg4BlNvANjGTrVWFbGcFyuGYic0A)[J]. 许宏科,秦严严. 徐州工程学院学报(自然科学版). 2013(04) 119 | 120 | 121 | --- 122 | #### 如果您需要更详细的【源码和环境部署教程】，除了通过【系统整合】小节的链接获取之外，还可以通过邮箱以下途径获取: 123 | #### 1.请先在GitHub上为该项目点赞（Star），编辑一封邮件，附上点赞的截图、项目的中文描述概述（About）以及您的用途需求，发送到我们的邮箱 124 | #### sharecode@yeah.net 125 | #### 2.我们收到邮件后会定期根据邮件的接收顺序将【完整源码和环境部署教程】发送到您的邮箱。 126 | #### 【免责声明】本文来源于用户投稿，如果侵犯任何第三方的合法权益，可通过邮箱联系删除。 -------------------------------------------------------------------------------- /abfa5c7db024793c18bb165d09b887e5.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/qunshansj/opencv-python-video-lane-line-detection/94a6be491f80c3258c3015b510a92284c17b5190/abfa5c7db024793c18bb165d09b887e5.png -------------------------------------------------------------------------------- /b8e7e07c4c9ee5bcd4f124367568f477.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/qunshansj/opencv-python-video-lane-line-detection/94a6be491f80c3258c3015b510a92284c17b5190/b8e7e07c4c9ee5bcd4f124367568f477.png -------------------------------------------------------------------------------- /cv2_fillPoly_mask_pt.py: -------------------------------------------------------------------------------- 1 | 2 | cv2.fillPoly(mask, pts, color) 3 | -------------------------------------------------------------------------------- /cv2_line_img_tuple_l.py: -------------------------------------------------------------------------------- 1 | 2 | cv2.line(img, tuple(line[0]), tuple(line[1]), color,thickness) 3 | -------------------------------------------------------------------------------- /db848810a9f691a8c45479a480cb6541.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/qunshansj/opencv-python-video-lane-line-detection/94a6be491f80c3258c3015b510a92284c17b5190/db848810a9f691a8c45479a480cb6541.png -------------------------------------------------------------------------------- /edge_img_cv2_Canny_g.py: -------------------------------------------------------------------------------- 1 | 2 | edge_img = cv2.Canny(gray_img,canny_threshold1,canny_threshold2) 3 | -------------------------------------------------------------------------------- /gaussian_cv2_Gaussia.py: -------------------------------------------------------------------------------- 1 | 2 | gaussian = cv2.GaussianBlur(color_img, (gaussian_ksize,gaussian_ksize), gaussian_sigmax) 3 | -------------------------------------------------------------------------------- /gray_img_cv2_cvtColo.py: -------------------------------------------------------------------------------- 1 | 2 | gray_img = cv2.cvtColor(input_image, flag) 3 | -------------------------------------------------------------------------------- /idx_np_argmax_diff_.py: -------------------------------------------------------------------------------- 1 | 2 | idx = np.argmax(diff) 3 | -------------------------------------------------------------------------------- /img_mask_cv2_bitwise.py: -------------------------------------------------------------------------------- 1 | 2 | img_mask = cv2.bitwise_and(gray_img, mask) 3 | -------------------------------------------------------------------------------- /lines_cv2_HoughLines.py: -------------------------------------------------------------------------------- 1 | 2 | lines = cv2.HoughLinesP(edge_img, 1, np.pi / 180, 15, minLineLength=40, maxLineGap=20) 3 | -------------------------------------------------------------------------------- /mask_np_zeros_like_g.py: -------------------------------------------------------------------------------- 1 | 2 | mask = np.zeros_like(gray_img) 3 | -------------------------------------------------------------------------------- /poly_pts_numpy_array.py: -------------------------------------------------------------------------------- 1 | 2 | poly_pts = numpy.array([[[0,368],[300,210],[340,210],[640,368]]]) 3 | --------------------------------------------------------------------------------