The Research And Development Of Target Tracking Based On Airbone Photoelectric Platform

With the development of technology,all countries in the world are carrying out revolution in military informatization.Various flying weapons with strong mobility and excellent endurance have mushroomed.Measurement and monitoring technologies for new types of weapons have also drawn widespread attention.The airbone photoelectric platform is valued by its flexible high-altitude viewing angle.The target tracking technology based on computer vision compensates for the shortcomings of traditional radar monitoring by optical means.In this paper,a target tracking system based on the airbone photoelectric platform is proposed.By focusing on the fields of image processing and pattern recognition,the paper studies the aircraft tracking technology under the sky-air background and focuses on three key aspects including target detection,target tracking and trajectory prediction.The main work and innovations are presented as follows:On the aspect of target detection,on the basis of the research of the existing target detection algorithm,the target detection scheme based on the manual calibration is selected according to the project background.To cope with the shortcoming of artificial background in the target area,a series of preprocessing operations are used to finely extract the target.The key steps are as follows: First,the framed image is transformed into a gray-level image and the image is processed by homomorphic filtering to enhance the image;then the preliminary image segmentation is performed by the dynamic binarization method based on Canny operator,and the interference is removed by the morphology technology;Finally,using the mask principle,the target image in the the framed image is extracted to generate a precise target template image.On the aspect of target tracking,due to the fact that the Camshift algorithm is characterized by a single feature and the limited target template,a dynamic updated Camshift tracking algorithm based on multi-feature fusion is proposed.In the aspect of feature extraction,the histogram probability distribution of the target template is extracted by using the parallel structure of H channel,V channel and LBP texture features.After the probability density distribution of the image of the current frame is generated,the edge features extracted by Sobel operator are used to filter it.As for the target template,the parallel feature histogram distribution of the tracking target in the video frame image is extracted under the condition of stable tracking at the interval of 30 frames,and the corresponding distribution of the original target template is dynamically updated.Compared with the original algorithm,the proposed algorithm has been greatly improved in accuracy,the adaptability to the characteristics of large-scale flight target changes,fast movement,easy to flip,and the resistance of birds and the uneven il umination.In the aspect of trajectory prediction,the deficiency of the traditional Kalman filter model is analyzed and a relative Kalman filter model is proposed.In view of the complexity of calculating the similarity measure,this paper proposes a twice judgment method from coarse to fine.The prediction of flight target trajectory is converted into the prediction of the background center point of the video image and the difference between the coordinates of the flight target and the center point.The reference frame is restored from the image origin to the the real scene,avoiding camera shake or interference caused by self-motion.The algorithm further improves the accuracy of trajectory prediction under occlusion.The first judgment use the comparement of the forecast and tracking results,which has grate gap will get into secondary judgments.To prevent misjudgement,the second judgment calculates the accurate occlusion judgment of the H channel histogram of the tracking targets in both the target template and the current frame.The twice occlusion judgment algorithm reduces the redundant calculation work and improves the real-time performance of the system.In order to verify the feasibility and practicability of the target tracking algorithm based on the airborne optoelectronic platform,this paper develops a video capture and target tracking software system based on VS2013 and OpenCV.The paper made a joint environment combined with airborne optical platform to carry out functional,robustness,accuracy and real-time testing.The test results show that the software platform has perfect functions and stable operation,the accuracy and speed of the proposed target tracking algorithm meet the predetermined requirements.