Reseach On Airport Target Detetion Based On Linear Features Extraction

This thesis is mainly about research on the airport target detection, which can help guided weapons automatically identify enemy airports. The research plays a great role in making hitting targets more accurate and efficient. It collapses the enemy air force by destroying the enemy airport target in military area, and it can also improve the security of the civil aviation system and ensure the stability of landing aircraft by using aircraft autopilot in civil area. How to quickly and accurately extract the airport target from the aerial image and visualize the entire target recognition process is of great theoretical and practical significance in our research work.Since the airport runway is the most obvious and common facility which can reflect the position information of the airport and detecting the position of the airport runway means getting the airport location, this thesis focuses on the detection of the airport runway and studies the linear feature extraction algorithm in the airport image,from the image space to the parameter space of "one to many" mapping mode, the counting method of the accumulator, endpoint detection and other aspects. We point out the limitations of the traditional Hough Transform in airport detection, and use the prior knowledge-driven strategies of airport to propose the improved Hough Transform which improves the detection time and accuracy by 30% and 70% respectively, and introduce the international line segment detection algorithm of LSD which has the highest accuracy, assisted by the runway length of the airport in the threshold range,the restricted width of the runway, the parallel edge of runway and other characteristics, so as to detect the airport runway in the preserved line.Furthermore, in the actual detection of aerial image, accessing the airport position through more than one image needs multiple sequence image to find, validate and confirm. In this thesis, we study a reconnaissance flight along the straight section of this special case, consider the flight distance, delay and other aspects, as well as use the method of block to truncate processing image so as to maintain the original basis of the accuracy and reduce detection time to 22% of original.Finally, this thesis use Vega Prime, MapX and other development tools in VC++development platform, receive the spatial location of airport, missile, and targets through Socket, and visualize the process of patrol, finding the target in thethree-dimensional visual window and two-dimensional GIS window to show the ground observers a more realistic image.The results show that the improved Hough transform and LSD algorithm to detect the airport can meet practical application needs, and the visualized platform of ground can be used to observe the detecting process in real time.