Research On Detection Algorithm Of Airport Birds Based On Machine Vision

With the development of China’s economy,the air transport industry is advancing to a new level.More attentions are being paid to the development of so called "smart airport".More scientific methods are used to improve the airport safety.Bird strike is one of the most important factors affecting air transportation safety,and it significantly affects the healthy development of air transportation industry.Damage from Bird striking can be greatly reduced if birds and their flight path can be detected,and necessary methods are used to repel birds.A number of tracking algorithms have been proposed by researchers.Most focused on video surveillance tracking.Very few studies have been done,and the algorithms were not optimized on repelling birds.Most did not work very well under different lighting conditions,and when there was only small color difference between objects and background.To counter those problems,this paper studied an algorithm to differentiate birds from aircraft by using particle filtering,and by integrating color and shape characteristics.1、The paper first introduces the current states of most object tracking algorithms and recent research development.It also emphasizes the importance of bird repelling.It explains how Bayesian filtering and Monte Carlo integration work.It also outlines the process of applying standard particle filtering,and steps of implementation.Actual Test also proves the advantage of the algorithm under complex conditions.2、The accuracy of bird and aircraft tracking is significantly affected by complex conditions and the variation of objects being tracked.Smog and the tracking objects being temporarily blocked will reduce the accuracy of tracking,and they may even cause failure to track.This paper proposed an improved algorithm by integrating color and shape characteristics.By using particle filtering methods,the matrix of the object characteristics is decomposed,analyzed and optimized.Characteristics are ranked according to their effects on tracking.Only important characteristics are kept.The test results also show that tracking accuracy is improved by 10%using the proposed algorithm comparing with other algorithms which only use color characteristics.3、To distinguish birds from aircraft,this paper proposed an algorithm which takes into account the flying characteristics of birds and aircraft.The algorithm first uses the difference of color and shape characteristics to differentiate the birds from aircraft.Then,algorithm advances one more step by considering the different flying path and speed characteristics.For objects which cannot be identified using above methods,an identification module is used to distinguish the two.This improves the speed and avoids lengthy object recognition.4、A testing environment was established using type Exynos4412 development board to demonstrate the tracking capabilities.This paper also introduces the sub-systems of the hardware platform.The proposed algorithm was implemented using OpenCV of the Android platform.The code was tested for its feasibility,stability and robustness.A final test was conducted in the development board to verify the effectiveness of the hardware and related software.