Chunhui Zhao

A vision based sense and avoid system for small unmanned helicopter

Sense and Avoid (SAA) has been identified as one of the crucial technologies for integrating UAV into civil airspace, and vision based SAA is regarded as a promising solution. In this paper we described our vision based SAA system consisted of onboard vision detection system and Groud Control Station (GCS) to fulfill the intruder detection and avoid maneuver control tasks. To realize the small intruder detection in collision scenario, we adopt a CMO filter based method for pre-processing as well as a HMM filter for target detection. The adopted algorithms are modified and immigrated on a high performance DM8168 onboard computer for efficiency consideration. Due to the depth missing problem of vision sensor, we developed a relative angle based avoidance strategy to perform the collision avoid of intruders. In the end, the proposed system is tested on real flight scenario to demonstrate its capability for small intruder collision avoid.

A UAV sense and avoid system design method based on software simulation

Sense and Avoid (SAA) has been identified as one of the crucial technologies for UAV safety flight. The SAA systematic design is to develop a system with sensor configuration and SAA algorithms. Normally, the SAA system integration and test can be both expensive and dangerous without carefully system and algorithms development. In this paper, we proposed a software based SAA simulation framework for SAA system configuration and algorithms design. First, the software structure and functional submodules are described. Then the SAA function is modelled and optimization based algorithms integrating environment sensing, collision evaluation and avoidance. Simulation results and analysis are provided to validate the simulation software and algorithms. By configuring the sensors and adjusting the algorithms parameters, the software based simulation framework can act as a guidance to the development of practical SAA systems.

SLAM-based cooperative calibration for optical sensors array with GPS/IMU aided

Loading wide baseline multi-arrays optical sensor system enables a UAV to obtain large field of view and multi-view information, and maximize its detection performance. An accurate calibration is a necessary prerequisite. Since the key problem on the calibration of multi-array optical system is how to find the accurate extrinsic parameters of single array, we propose a cooperative calibration with GPS/IMU aided for the single array optical system to improve the accuracy of calibration. Our method does not require an initial guess of any extrinsic parameter. Through extensive field tests on ground prototype experimental system, we demonstrate that our algorithm has more accuracy compared to multi-camera calibration without GPS/IMU aided as well as robustness for calibration of single array optical system.

Simultaneously multi-UAV mapping and control with visual servoing

In this article, we proposed a framework to realize vision based multi-UAV cooperative mapping and control with image based visual servoing (IBVS). Equipped with a camera on each UAV platform respecitively, the multi-UAV system with overlapping camera field of view (FOV) can achieve ground mapping and formation cooperatively by image information interaction. Initially, we introduce the Multi-UAV system and the related geometry frames. Then the system functions are described in two parallel aspects. On the formation control aspect, we introduce the visual servoing feature with SIFT moment. A visual servoing controller is utilized to guide SUAV to track the desired feature given by MUAV so as to realize the formation flying. On the mapping aspect, two mapping modes, maximum combined filed of view (CFOV) and maximum Overlapping field of view (OFOV) are designed and different desired feature generation methods are proposed. In the end, simulations are carried out on both mapping mode to demonstrate the feasibility of proposed formation control and mapping method.

An application of panoramic mosaic in UAV aerial image

Panoramic mosaic plays an important role in the field of computer vision, robot navigation and virtual reality. This paper summarizes the specific process of panoramic stitching and proposes a coordinate transformation stitching method based on down-sampling. In order to reduce the processing time, images are compressed by down-sampling before processing, and spliced in the corresponding points of original images. Especially, the middle image is chosen as the reference image and the others are directly spliced onto it by transformation matrix. Considering the illumination of UAV aerial images, just one overlapping region of the adjacent images is remained when doing the fusion. In the experiment, images with different resolutions are tested. The results show the performance of good efficiency and little time consuming.

Motion Deblur Based on the Distributed Optical- Arrays System

In order to alleviate the blur of target caused by the local motion, we have constructed a distributed opticalarrays system with six cameras that consist of three distributed arrays. The target association algorithm using Gaussian Mixture Model (GMM) and template matching is proposed to recognize the same target within inner-array. The motion of target is triangulated after the 3D scene flow is estimated that was used for describe the three-dimensional motion of moving target in the view. Finally, based on linear interpolation, the point spread function (PSF) was estimated to deblur the fuzzy targets. The experiments demonstrate the feasibility and effectiveness of the developed system.

UAV Sense and Avoid with ADS-B, Simulation and Experiment

ADS-B is regarded as a promising fashion on UAV SAA application. In this article, we proposed simulation and HIL platforms to test ADS-B based SAA functions, trials are carried out with real on-the-air ADS-B data. In the simulation, we build the simulation scenario with ArcGIS, and we use a small ADS-B receiver to collect airspace traffic status to generate a real UAV flying environment. In the experiment setup, we use both ADS-B In and ADS-B Out to carry out the SAA function with two small quadrotors. Both the simulation and experiment has shown the advantages of information sensing and collision resolution towards SAA scnarios, with higer precision, wider range as well as better stability.

Vision-based Sense and Avoid of UAV with 2D Dynamic Safety Envelop

Sense and Avoid (SAA) has been identified as one of the most significant challenge for integrating UAV into civil airspace, and vision system has demonstrated its capacity to SAA. To deal with the depth-independent vision based sense and avoid problem, an image-based envelop formulation method is proposed and single point feature is found. Then a angle only visual servoing controller is designed to track the selected point at the same time the motion of the feature point is compensated. In the end, simulation represent different collision scenarios are carried out to illustrate the feasibility of the proposed method, the simulation results demonstrate that the method in this article can achieve SAA in without range information.