Abstract
The road marking recognition is an important part of road scene understanding based on computer vision. In road marking recognition, the contour-based algorithm such as generalized Hough transform (GHT) is more effective than the texture-based algorithm for no texture noise effect. However, there are a lot of redundant calculations in template matching which reduces the system efficiency. This paper presents a voting based matching pursuit (VMP) algorithm for locating reference points automatically in solving the sparse optimization problem, which achieves automatic alignment of samples in recognition. Then a dictionary learning method based on VMP algorithm is proposed, which uses a simple strategy to update the dictionary elements. The experimental results from two data sets have shown that the system efficiency is significant improved while ensuring the accuracy rate by the proposed method.
Highlights
In recent years, contour image recognition is more and more widely used
The generalized Hough transform (GHT) is a typical contour-based algorithm, which just use the contour information in recognition, so it effectively avoid the influence of texture noise
V-sparse representation based classification (SRC) This paper presents a modified matching pursuit algorithm based on voting which called voting based matching pursuit(VMP)
Summary
Taking the recognition of road markings as an example, the main information of the object is concentrated on the contour, so that the texture-based method is vulnerable to the influence of noise. This paper focuses on sparse dictionary learning for contour image recognition. Fang: Contour Sparse Representation Using VMP for Road Marking Recognition representation with the l2,1-norm minimization [14], [15], sparse representation with the l2-norm minimization[16],[17] Another important way to obtain an approximate sparse representation solution is the greedy strategy. The methods based on sparse representation mentioned above have high speed and accuracy in dealing with aligned object. Wright et al proposed the sparse representation based classification (SRC) method for robust face recognition [1]. SRC utilizes the representation residual ei associated with each class to do classification
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
