Python初步实现车道线检测

车道线检测是一个常见的问题，本文主要介绍如何简单有效的用python实现这个功能

主要思路

1 读取图像并选择感兴趣的区域ROI进行下一步处理
2 对ROI区域进行预处理包括灰度化，膨胀和腐蚀
3 对预处理后的图像进行边缘检测（Canny算子）
4 对边缘检测后度图像进行概率霍夫变化得到多条直线
5 对直线进行分组，根据斜率分成左线和右线
6 选择左右线各一条，之后scale直线是其延伸至边界

To do things:
1在连续图片的车道线检测中会有水平的直线，需要处理
2某几帧没有检测结果用之前的结果代替
3弯道车道线检测

结论：
对简单单调的环境效果不错
依赖参数的设置

**

效果展示

原始图片

输出结果

---

代码展示

#created by BOXIAO  27/09/2017
#all rights reserved 
import os
import cv2
import sys
import numpy as np
import skimage.transform as st
import matplotlib.pyplot as plt
from skimage import feature 

#预处理图像膨胀和腐蚀（可选）
def preprocessimage(img):
    kernel=np.uint8(np.zeros((5,5)))
    for x in range(5):
        kernel[x,2]=1
        kernel[2,x]=1
    eroded=cv2.erode(img,kernel)
    dilated = cv2.dilate(img, kernel)
    result = cv2.absdiff(dilated, eroded)
    return result
#scale所得的车道线，使其充满屏幕
def scale_lines(chooses_lines):
    final_lines=[]
    for line in choose_lines:
        p0, p1 = line
        k = np.float32(((p0[1]-p1[1])*1.0)/(1.0*(p0[0]-p1[0])))
        b = np.float32(p1[1] - k*p1[0])
        y1=0
        x1=-b/k
        if k < 0:
            x2=0
            y2=b
        else:
            x2=640
            y2=640*k+b
        line=((x1,y1),(x2,y2))
        final_lines.append(line)
    return final_lines


if __name__ == '__main__':
    input_dir=sys.argv[1]
    output_dir=sys.argv[2]
    data_dir='/home/b_xi/codes/lanedetection/mymethod'
    frames_list = os.listdir(input_dir)
    frames_list.sort()
    for f, frame in enumerate(frames_list):
        img = cv2.imread(os.path.join(data_dir,input_dir,frame))
        ROI=[330,480,0,640]
        ROI_img=img[ROI[0]:ROI[1],ROI[2]:ROI[3]]
        image = cv2.cvtColor(ROI_img,cv2.COLOR_BGR2GRAY)
        edges = feature.canny(image, sigma=2, low_threshold=2, high_threshold=25)
        all_lines = st.probabilistic_hough_line(edges, threshold=10, line_length=70,line_gap=0)
        #根据斜率将直线分为两组：左直线和右直线
        left_lines = []
        right_lines = []
        for line in all_lines:
            p0, p1 = line
            theta = np.abs(np.arctan2((p0[0] - p1[0]), (p0[1] - p1[1])))
            if theta > 1:
                left_lines.append(line)
            else:
                right_lines.append(line)
        choose_lines = [left_lines[0], right_lines[0]]
       # scaling left and right lines
        lines=scale_lines(choose_lines)
        #绘图并且保存
        fig, (ax0,ax1,ax2,ax3) = plt.subplots(1, 4, figsize=(16, 6))
        plt.tight_layout()
        ax0.imshow(img)
        ax0.set_title('Input image')
        ax0.set_axis_off()
        ax1.imshow(edges,plt.cm.gray)
        ax1.set_title('Canny deges')
        ax1.set_axis_off()
        ax2.imshow(edges*0)
        for line in all_lines:
            p0, p1 = line
            ax2.plot((p0[0], p1[0]), (p0[1], p1[1]))
        row2, col2 = image.shape
        ax2.axis((0, col2, row2, 0))
        ax2.set_title('Probabilistic Hough')
        ax2.set_axis_off()
        ax3.imshow(image, plt.cm.gray)
        for line in lines:
            p0, p1 = line
            ax3.plot((p0[0], p1[0]), (p0[1], p1[1]),linewidth=10.0)
        ax3.set_title('Output image')
        ax3.set_axis_off()
        plt.savefig(os.path.join(data_dir,output_dir,frame))
        plt.show()

## 最后 ##
在学习代码的过程中不断度阅读他人的博客，给了我极大的帮助，希望之后自己也可以不断分享学习感悟，最后感谢开源。