How to make this work for custom videos

[abhinavsagar]

No description provided.

[pch9520]

Recently I have a try.
Firstly,I calibrate my camera parameter with the calibrate_camera.py;
secondly,I calibrate meters to pixel of my picture taken by my camera;
Thirdly,I amend the parameters in line_fit.py perspective_transform.py and line_fit_video.py
Now,I can use this code to detect lines in my video but in some moment the detection is not accurate

[StephanieCoding]

### When I use my own videos to test, this error often occurs. Anyone knows which parameters should I amend?

C:\Users\moon5\Anaconda3\python.exe D:/Documents/Desktop/智能驾驶/advanced_lane_detection-master/advanced_lane_detection-master/line_fit_video.py
Traceback (most recent call last):
File "D:/Documents/Desktop/智能驾驶/advanced_lane_detection-master/advanced_lane_detection-master/line_fit_video.py", line 104, in
annotate_video('video_1.mp4', 'out.mp4')
File "D:/Documents/Desktop/智能驾驶/advanced_lane_detection-master/advanced_lane_detection-master/line_fit_video.py", line 98, in annotate_video
annotated_video = video.fl_image(annotate_image)
File "C:\Users\moon5\Anaconda3\lib\site-packages\moviepy\video\VideoClip.py", line 514, in fl_image
return self.fl(lambda gf, t: image_func(gf(t)), apply_to)
File "C:\Users\moon5\Anaconda3\lib\site-packages\moviepy\Clip.py", line 137, in fl
newclip = self.set_make_frame(lambda t: fun(self.get_frame, t))
File "", line 2, in set_make_frame
File "C:\Users\moon5\Anaconda3\lib\site-packages\moviepy\decorators.py", line 14, in outplace
f(newclip, *a, **k)
File "C:\Users\moon5\Anaconda3\lib\site-packages\moviepy\video\VideoClip.py", line 669, in set_make_frame
self.size = self.get_frame(0).shape[:2][::-1]
File "", line 2, in get_frame
File "C:\Users\moon5\Anaconda3\lib\site-packages\moviepy\decorators.py", line 89, in wrapper
return f(*new_a, **new_kw)
File "C:\Users\moon5\Anaconda3\lib\site-packages\moviepy\Clip.py", line 94, in get_frame
return self.make_frame(t)
File "C:\Users\moon5\Anaconda3\lib\site-packages\moviepy\Clip.py", line 137, in
newclip = self.set_make_frame(lambda t: fun(self.get_frame, t))
File "C:\Users\moon5\Anaconda3\lib\site-packages\moviepy\video\VideoClip.py", line 514, in
return self.fl(lambda gf, t: image_func(gf(t)), apply_to)
File "D:/Documents/Desktop/智能驾驶/advanced_lane_detection-master/advanced_lane_detection-master/line_fit_video.py", line 43, in annotate_image
ret = line_fit(binary_warped)
File "D:\Documents\Desktop\智能驾驶\advanced_lane_detection-master\advanced_lane_detection-master\line_fit.py", line 79, in line_fit
left_fit = np.polyfit(lefty, leftx, 2)
File "C:\Users\moon5\Anaconda3\lib\site-packages\numpy\lib\polynomial.py", line 550, in polyfit
raise TypeError("expected non-empty vector for x")
TypeError: expected non-empty vector for x

[pch9520]

Did you feed your video straightly without amending any parameter?
If so ,it's normal to encounter issues.
@StephanieCoding

[neishka]

I am facing the same issue, @pch9520 what parameters are you referring to.

[pch9520]

I am facing the same issue, @pch9520 what parameters are you referring to.

If you want to feed your own video,I think the following parameters you should amend:
1.the size of each frame picture;
2.some parameters in perspective_transform.py to fit your scene
src = np.float32(
[[200, 720],
[1100, 720],
[595, 450],
[685, 450]])
dst = np.float32(
[[300, 720],
[980, 720],
[300, 0],
[980, 0]])
@neishka

[Phillweston]

I am facing the same issue, @pch9520 what parameters are you referring to.

If you want to feed your own video,I think the following parameters you should amend:
1.the size of each frame picture;
2.some parameters in perspective_transform.py to fit your scene
src = np.float32(
[[200, 720],
[1100, 720],
[595, 450],
[685, 450]])
dst = np.float32(
[[300, 720],
[980, 720],
[300, 0],
[980, 0]])
@neishka

What do you mean by the parameters? If I use my own video, how can I adjust the above parameters to get the best recognition effect?

[pch9520]

Hello, @Phillweston
你可以看看这篇解析（网上还有很多其他的）：
https://zhuanlan.zhihu.com/p/46146266
如果你要在自己的视频上用，就要先调整联合阈值的参数，把车道线部分过滤出来，然后再调整鸟瞰图参数，设置你的感兴趣区域并将前视图映射到俯视图，就可以了
如果你还要得出具体的车辆曲率和车辆偏离车道中心的距离，还需要修改距离标定参数（横纵向一个像素分别映射到现实世界是多少距离（m），在俯视图上进行标定）