Support vector machines based stereo matching method for advanced driver assistance systems
- PDF / 4,756,809 Bytes
- 17 Pages / 439.642 x 666.49 pts Page_size
- 36 Downloads / 232 Views
Support vector machines based stereo matching method for advanced driver assistance systems Zakaria Kerkaou1 · Mohamed El Ansari1 Received: 6 August 2019 / Revised: 13 May 2020 / Accepted: 24 June 2020 / © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract Stereo vision is a measurement method for finding correspondence between two or more input images in order to obtain a detailed 3D representation of a scene. This paper presents an approach for matching stereo sequences acquired by a stereo sensor mounted on an intelligent vehicle. The approach uses machine learning alongside spatio-temporal information to predict the matching results. This means that the obtained matching results during the previous frame are used as a training samples for a support vector machine classifier, as well as to derive disparity ranges for each scan-line, which are then used to predict the matching of the current frame. The distance to the hyper-plan computed by the SVM is used as a cost function to fill a 2D search space. Then, the dynamic programming algorithm is performed for matching edge points in the stereo pair. Experiments on both virtual and real stereo image sequences have been conducted, demonstrating satisfactory performance. Keywords Stereo vision · Support vector machine · Spatio-temporal matching · ADAS
1 Introduction Recent development in computer vision and image processing fields led to the inevitable employment of the stereo vision methods in different subjects such as 3D navigation, obstacle detection, 3D reconstruction and so on. One of the prime applications of stereo vision is the Advanced Driver Assistance Systems (ADAS) [13, 21, 29, 30, 53], which are developed to minimise the human error and improve vehicle systems for safety and better driving. Stereo correspondence is an inherently complicated problem. The key problem is finding correspondence between pixels of stereo images taken from different viewpoints. Stereo matching methods can be classified into two main categories: global methods and local methods. Global methods present the matching task as an energy minimisation problem and Zakaria Kerkaou
[email protected] Mohamed El Ansari [email protected] 1
LabSIV, Department of Computer Science, Faculty of Science, Ibn Zohr University, BP 8106, 80000, Agadir, Morocco
Multimedia Tools and Applications
suggest to solve it using algorithms such as graph-cut (GC) [3, 26], dynamic programming (DP) [1, 10, 22, 23, 41, 60] or belief propagation (BP) [24, 25, 47]. These methods have become very popular due to their accurate results. Unfortunately, they come with low efficiency and complicated parameters, and are difficult to apply into real-time applications. On the other hand local methods that are generally based on block matching or pixel correlation [19, 43, 50], consider only local information, which means that a disparity of a pixel depends only on neighbouring pixels. These local methods are also popular due to their simplicity. However, since the used information
Data Loading...
