CS 7495 – Computer Vision Support Vector Tracking
نویسنده
چکیده
The paper selected for implementation was Support Vector Tracking by Shai Avidan. The paper introduces a simple but novel idea of replacing the standard optical flow algorithm equation with equations that use the SVM to calculate the derivatives. The standard way of calculating optical flow is by solving the following matrix: A 11 A 12 u b 1 =-(1) A 21 A 22 v b 2 where (using Lucas Kanade optical flow algorithm), A 11 = (I x 2) A 12 = A 21 = (I x *I y) A 22 = (I y 2) b 1 = I*I x b 2 = I*I y u = optical flow in x direction v = optical flow in y direction I = change in from one frame to another I x = derivative of I in x-direction I y = derivative of I in y-direction Hence the final position, I final = I prev + uI x + vI y-(2) This means that the values of u and v are maximized over several iterations until either the changes become stagnant or the threshold for maximum number of loops is reached. The author suggests that an svm classifier trained to detect vehicles can be used to track the vehicle directly. This means that the very same svm classifier that is used to start the tracking by identifying the vehicle, can be used to track the vehicle over time. If an svm with quadratic polynomial kernel is used, then the output generated is calculated by the svm as follows: I final = (yj j *I*x j) – (3) where
منابع مشابه
Visual Tracking using Learning Histogram of Oriented Gradients by SVM on Mobile Robot
The intelligence of a mobile robot is highly dependent on its vision. The main objective of an intelligent mobile robot is in its ability to the online image processing, object detection, and especially visual tracking which is a complex task in stochastic environments. Tracking algorithms suffer from sequence challenges such as illumination variation, occlusion, and background clutter, so an a...
متن کاملCovariance Analysis of a vector tracking GPS receiver based on MMSE multiuser Detection
In high dynamic conditions, using vector tracking loops instead of scalar tracking loops in GPS receivers is proved as an efficient method to compensate the performance. The Minimum Mean Squared Error detector as a multiuser detector is applied in the vector tracking loop for more reliability and efficiency. The Kalman filter does the two tasks of tracking and extracting the navigation data aft...
متن کاملOnline multiple people tracking-by-detection in crowded scenes
Multiple people detection and tracking is a challenging task in real-world crowded scenes. In this paper, we have presented an online multiple people tracking-by-detection approach with a single camera. We have detected objects with deformable part models and a visual background extractor. In the tracking phase we have used a combination of support vector machine (SVM) person-specific classifie...
متن کاملRobust Facial Expression Recognition via Compressive Sensing
Recently, compressive sensing (CS) has attracted increasing attention in the areas of signal processing, computer vision and pattern recognition. In this paper, a new method based on the CS theory is presented for robust facial expression recognition. The CS theory is used to construct a sparse representation classifier (SRC). The effectiveness and robustness of the SRC method is investigated o...
متن کامل