Study Of Airport Image Feature Extraction For UAV's Visual Landing Based On DSP

With the gradual progress and maturity of computer vision technology, visual navigation tech-nology in UAV autonomous landing area has been extensively studied. In this papers,by airports image taken from a UAV landing processas the background ,we designed a set of methods,which is used to quickly extract airport image feature information during UAV vision navigation landing process.As well,we made use of DSP hardware platform to test the algorithm software practically application performance in the semi-physical simulation environment.Fristly,in this papers we have designed a set of effective image processing algorithms to quickly extract this three straight line. A method of segmentation sky form ground was designed based on Ostu algorithm principle,which is used for dividing the original image into tow Sub-image areas of sky and ground.The original image resolution was reduced by use of Gaussian image pyramid method so to reduce the amount of image data processing. On the basis of the sobel Edge image, a straight line extraction method,which can make fully use of the gradient direction and amplitude information of the Image edge points,was designed by combining the Hough Transform and Phase Classification methods to fastly extract the horizon and runway two edge line from the two sub-images respectivly.Secondly,On the basis of image processing algorithm test completed in VC++6.0,we program all image processing algorithms software running on the hardware platform as the core of DM642 processor. In the CCS software development environment,11 image processing function modules were coded according to the function of each part of image algorithm, so it can achieve modular pro-gram management, which is beneficial to the algorithm program maintenance and further expansion. In addition, some matters needing attention,which exist in the programming process,were descriped detailedly.Finally, we tested the actuall performance of algorithm program running on the DM642 platform through the simulation runway scene. Experimental results show that this algorithm has a good anti-disturbance performance in semi-physical simulation environment.As well,its real-time per-formance has basically meet the performance requirements of this subject.