sy456's profile - activity

Hi Harris,

This is very simple image from Caltech Lanes Dataset. However, I don't work on single images but image sequences under changing lighting conditions.

Hi,

I use hough transform for straight lane detection but I got lots of false positives even in clear lane markings. My approach is like this:

1. Select ROI for near field
2. Eliminate noise (various options like gaussian, median filter...)
3. Detect edges and binarize image
4. Detect Hough lines
5. Eliminate lines considering the angle of the lines (different angles for left and right side of image)
6. Group Hough lines and draw bounding box around them
7. Track bounding boxes with Kalman filter

I am stuck with 5th and 6th step. Some of the lines, which are very close to each other, are drawn on the lane markings. How can I distinguish these lines and the others that are not on the lane markings? I also wonder if hough transform is a really good idea for straight lines. Maybe I need to use RANSAC for fitting. Any suggestions?

This is very simple image from Caltech Lanes Dataset. However, I don't work on single images but image sequences under changing lighting conditions.

Regards

I didn't get it, can you specify little bit?

Hi,

I have searched geometric models (e.g.: straight line, circular arc, polinomals, splines etc.) to make some assumptions about the road’s structure. After I define model, I will fit the extracted features with the geometric model by using different methods like Least Squares Method, Least Median Squares, RANSAC etc.

As a start, I will use straight line for highway scenario and then continue with others (e.g.: spline models for curves). I understand the general idea. I am looking for a clue about how to implement geometric lane models in OpenCV or C++. Any suggestions?

Regards,

Suleyman

Hi everybody,

I would like to learn if it possible to use MATLAB functions in OpenCV. I want to use KITTI raw data development kit that contains Matlab demonstration code with C++ wrappers [please see the picture below]. It has everything I need, reading annotations/ground truth froim data, visualization...However, I am using OpenCV C++ environment. My aim is to compare my object detection with the ground truth data that stored in tracket_labels.xml file.

I already found wrappers that enable to use OpenCV in Matlab but I couldn't find the other way around. The worst case option is to translate Matlab code to C++ manually. I wanted to ask before I start. Any suggestions?

Regards

Actually, I used cars3.xml that @albertofernandez mentioned without using negative.txt It is trained with old haar training. I think i need to use train cascade with new negative images.

@StevenPuttemans, I checked Beyond Hard Negative Mining (http://www2.isr.uc.pt/~henriques/beyond/index.html). Afaik, it improves the training part. Have you evert tried it?

actually, i already achieved that part:) thanks anyway

OpenTLD is used to track single objects, also you need to define bounding box by hand. However,I need to detect multiple objects automatically.

Hi,

I am trying to detect and track vehicles by using Haar Feature-based Cascade Classifier in OpenCV. However, I got lots of false positives. Is there any way to eliminate false positives?

Suleyman

thank you for the answer and the link. what do you mean by texture-less objects?

Hi,

I need to make a decision between key-point based detection vs HoG/Haar training. Sorry for long question in advance but I am really stuck!!

So far, I have been trying to use SIFT, SURF and other key-point based feature extraction methods to detect and track vehicles, pedestrians, traffic signs and lanes. I have to detect these objects at the same time with moving camera. I used that approach for two consecutive frames to analyse the movement of key-points:

detect features --> describe features--> match features between frames --> filter matches

After that I want to group the features onto onto the cars, pedestrians, traffic signs and lanes. I think there should be same way to achieve this. I need to make a data reduction inside the camera because HD cameras produces large data streams. I thought that using these approach I can create a cheap vision pipeline without using any trained data.

However, when I read research paers and talk to experts, i see that if you want to draw a bounding box on the object, you mostly need a trained data. Most of the people uses HoG/Haar training and feed a classifier (SVM/Cascade) for specific objects. Why HoG and Haar is mostly preferred by the community rather than using SIFT or SURF? I cannot convince myself to switch to HoG of Haar!

Also, people use different detectors specific for the object. For instance; HoG + SVM for pedestrian detection; Haar+Cascade for vehicle detection; edge detection+ hough transform+ line fitting for lanes etc.. What I want to do is to find the commonalities and variabilities of these different pipelines and (if possible) come up with a pipeline as generic as possible.

Any advice or pointer to resources??

Regards

Hi Vinay, did you use opencv_haartraining (obsolete version) or opencv_traincascade?

Hi,

I have two frame sequences; C:\fakepath\image1.jpg and C:\fakepath\image2.jpg. When I use drawmatches function in opencv, it shows a window with two consecutive frames next to each other for matching.

However, I want to show feature points of frame 1(image1.jpg) plotted in red, the feature points of frame 2 (image1.jpg) plotted in blue (also plotted in the image1.jpg) and draw a line between matched points.

How can I do it in OpenCV?

Regards