Motion Detection and Optical Flow

In this paper Ransac algorithm is used to detect the motion is happened or not and this is by finding out the matching features between the two images. And then finding out the motion using optical flow algorithm. A common problem of optical flow judgment is fine motion structures cannot always be correctly estimated, especially for regions with significant and abrupt displacement variation. To address this issue introduce Lucas Kanade’s optical flow method. This method is very fast and easy calculation. And it is also very fast method. Keywords—Ransac,Motion,LucasKanade’s, Optical flow, matching features INTRODUCTION Motion analysis and estimation is one of the most challenging tasks in digital video processing and computer vision [8]-[12]. Optical flow presents an apparent change of a moving object’s location or deformation between frames. Optical flow estimation yields a two-dimensional vector field, i.e., motion field, that represents velocities and directions of each point of an image sequence [8]. As it is an illposed problem, so far a wide variety of constraints between frames have been introduced in optical-flow modeling. Such constraints are based on image brightness and velocity. In particular, assumption of image brightness constancy between frames is one of the mostly used constraints. Optical flow is an approximation of the local image motion based upon local derivatives in a given sequence of images. That is, in 2D it specifies how much each image pixel moves between adjacent images while in 3D in specifies how much each volume voxel moves between adjacent volumes. The 2D image sequences used here are formed under perspective projection via the relative motion of a camera and scene objects. The 3D volume sequences used here were formed under orthographic projection for a stationary sensor and a moving/deforming object. In both cases, the moving patterns cause temporal varieties of the image brightness. It is assumed that all temporal intensity changes are due to motion only. The computation of differential optical flow is, essentially, a two-step procedure: 1. Measure the spatio-temporal intensity derivatives (which is equivalent to measuring the velocities normal to the local intensity structures) and 2. Integrate normal velocities into full velocities, for example, either locally via a least squares calculation [1, 3] or globally via a regularization [2, 3]. In general, the use of optical flow in a generic machine vision system will probably require a sophisticated analysis of image content and motion in order to determine that all of the algorithmic assumptions are likely to be met. Quantitative use of the data will also require quantitative predictions of accuracy. In optical flow, a motion vector of each pixel is computed and entire image could be imagined as a vector field. The motion vector of each pixel represents the brightness of the pixel. The region of the image where brightness change is observed is considered as a candidate for moving object. The method based on optical flow is complex, but it can detect the motion accurately even without knowing the background. This approach results in good performance, however this algorithm need one more than image to be stored, thus resulting in higher memory requirements, in-turn resulting in high cost. RELATED WORK In [4] optical flow can benefit from sparse point correspondences from descriptor matching. The local optimization involved in optical flow methods fails to capture large motions even with coarse-to-fine strategies if small subparts move considerably faster than their surroundings. Point correspondences obtained from global nearest neighbor matching using strong descriptors can guide the local optimization to the correct large displacement. Conversely, also shown that weakly descriptive information, as is thrown away when selecting keypoints, contains valuable information and should not be ignored. The flow field obtained by exploiting all image information is much more accurate than the interpolated point correspondences. Moreover, outliers can be avoided by integrating multiple hypotheses into the variational approach and making use of the smoothness prior to select the most consistent one. This work extends the applicability of optical flow to fields with larger displacements, particularly to tasks where large displacements are due to object rather than camera motion. Here expect good results in action recognition when using the dense flow as a dynamic orientation feature correspondingly to orientation histograms in static image recognition. However, with larger displacements there also appear new challenges such as occlusions. In [5] presented a new optical flow estimation framework to reduce the reliance on the coarse level estimation in the variational setting for small-size salient motion estimation. Differing from previous efforts mainly to improve the model, instead revise flow initialization in Deepa Raju et al, / (IJCSIT) International Journal of Computer Science and Information Technologies, Vol. 5 (4) , 2014, 5716-5719