Moving Target Detection And Tracking Based On Improved Optical Flow Method

The detection and tracking of moving targets is a research and application field that has been popular in recent years.The technology can recognize moving targets from the complex background accurately and automatically,and can continuously obtain the position of the target and then track it.Currently it has been used in daily life and military fields widely.At present,the detection and tracking algorithms that are widely used in the field of computer vision are optical flow methods.The L-K optical flow method involved in this paper is ineffective for large inter-frame motions.In order to overcome the defects of the L-K optical flow method,we added a pyramid image to downgrade the original image and collect them continuously.Then we combine it with the Meanshift algorithm to remove the interference of the feature points in the background and speed up the tracking.This article mainly contains the following work:(1)The article introduces the image processing technology and how to collect images and preprocess the images.In the article we introduce the common algorithms for detecting and tracking moving targets and compare the results of them.(2)In the article we analyze the advantages and the disadvantages of L-K optical flow method,explore the solution for correction.We research the principle,advantages and disadvantages of the Meanshift algorithm and improve it after combining with the L-K optical flow method.Then wu use image pyramids and the ability to find extreme values in the density distribution of the Meanshift algorithm to enhance the accuracy of the L-K optical flow method for tracking moving targets.The improved algorithm is compared with the tracking effect of the particle filter algorithm.(3)We use the Samsung Cortex-a9 quad-core Exynos4412 development board to detect and track moving targets real time.First,We introduce the hardware components of the development platform.Next,the Linux kernel is ported to the ARM development board.Then we introduce the QT migration,CCD camera video acquisition,OPENCV library and header file migration.After the configuration is completed,the detection and tracking algorithms mentioned in the text are tested,and we finally realize the result an display it on the LCD screen.