Research On Vision-based Line Tracking Algorithms For Unmanned Aerial Vehicles

In recent years,computer vision has made tremendous development in many fields,such as unmanned autonomous navigation system,intelligent transportation,security monitoring and intelligent city.At the same time,Unmanned Aerial Vehicle(UAV)has been applied to civil area due to its popularity increased.UAV combined with computer vision has been applied to various fields,such as power cruise,highway cruise,pipeline cruise,coastline cruise,etc.These tasks are accomplished based on line tracking.Path planning and avigation of UAV in low-altitude flight can be improved by taking advantage of UAV and computer vision.This dissertation summaries the development and application of line tracking technology.Next,this dissertation studies the line tracking algorithms which are applied to UAV platform to recognize static lane line and dynamic coastline by using computer vision.The main research contents in this dissertation are as follows:UAV platform for the experiments of the following chapters is designed firstly,then the line tracking algorithms will be transplanted to it.We further use camera and Raspberry Pi 2 to capture and process images in real time.We focus on the recognition of static lane line.Firstly,the lane line image is pre-processed.Specifically,gray conversion,morphological filtering and edge detection are involved.In addition,two general lane models: linear model and curve model are introduced.In general,linear model is utilized to identify straight line instead of curve.By contrast,the curve model has advantages in recognizing curve.However,calculation and time cost are remarkably increased,due to the complexity of the curve model increased.Given that the above considerations,this inspires us to propose a new mixed scanning model based on edge detection which can well identify both straight line and curve.Then the algorithms are transplanted to UAV platform to verify the availabitily of the algorithms during the day and night.The recognition of dynamic coastline is further investigated.In this context,taking the color feature of sea water and beach int account,the images are analyzed in the color space.Given that HSV space is accorded with human visual perception of color and H channel has little effect on light illumination,H channel is choosed to realize K means clustering segmentation,then particulate parts are removed by morphology filtering.With the above development,contour extraction algorithm is employed to extract the contour of coastline.After that the poposed mixed scanning model is applied to identify the contour of coastline.Likewise,the UAV platform is utilized to demonstrate the effectiveness of the recognition.