Research And Implementation Of Quadrotor Indoor Navigation Using Monocular Vision

With the rapid development of science and technology, aircraft has played a more and more important role in the fields of daily life, adventure and disaster relief, security inspections, etc. Compared to the inertial navigation, visual navigation has some advantages, such as wide detection zone and large amount of accessible information, etc., and it is becoming an important navigation technology of aircraft. For the ground robots, using the mileage meter and a complex vision sensor of SLAM (simultaneous localization and mapping) has been proved can build a very successful visual navigation system, and there exist many visual positioning and navigation algorithms for the ground robots environment. However, using these algorithms on the aircraft in order to get the same work needs equipped with more sensors, and also will be limited to these characteristics, such as aircraft’s load, on-board computing power, inertial positioning error, and more quick motion control response to the request and so on. For these reasons, some algorithms are not apply in the aircraft shall.For the above problem, this paper designed and realized an indoor visual navigation system based on AR.Drone quadcopter. Equipped with a low-cost camera and without prior knowledge of the indoor environment, and also do not need a complex3D model, this visual navigation system can achieve to allow the aircraft to fly autonomously in corridors, and can also be able to detect and avoid smart many obstacles at any flight time, such as the front pedestrian or a blank wall. Related technologies are as follows:(1) In the obstacle avoidance, this paper put forward a comprehensive obstacle avoidance strategy which combines Image entropy technology and image entropy technology. For the general indoor environment stream obtained by the camera video, we did two-step processing. In the first step, we do image entropy calculation and analysis for each frame image to achieve the walls or other low texture barriers close detection, and then make a smart retreat or steering. The second step is to do optical flow detection for the image frames by zoning, the specific operation including pre-treatment, the characteristic angle point selection and feature points of the LK optical flow tracking. Then, based on the optical flow balance avoidance strategy, testing the pedestrians or other general obstacles in the corridors and make escape.(2) In the navigation, we designed and realized a fast vanishing point navigation algorithm. For the corridor picture obtained by camera, do preprocessing, edge detection, random Hough transform linear extraction and VQME fast vanishing point estimate, then to obtain the point to the vanishing point of the corridor’s end, and then computing the navigation error to control the aircraft navigation.Experimental results show that this visual navigation system is feasible and effective for general obstacle avoidance and corridor navigation, and has a definite theoretical value and a strong practical.