Design And Implementation Of Wearable Visual And Infrared Fusion Imaging Surveillance System

With the rapid development of sensor technology,a variety of image sensors are widely used in various fields.Image fusion technology,which relies on redundant and complementary information from multi cameras,is wildly used to obtain richer,higher-layered and more completed information than single camera gets in the same scene.This paper mainly uses image fusion technology to design and implement wearable visual-infrared fusion imaging surveillance system,including wearable separation design of hardware structure,research and improvement of heterogeneous image fusion algorithm and optimize deployment on the embedded platform of ARM-DSP.Then,experiments are carried out and achieve good results.First,the wearable structure is determined and the overall program design of the system is completed.The system is divided into a data acquisition terminal and an interactive display terminal to implement a wearable separate structural design which can reduce the size of the device and improve the human-computer interaction experience.The main function of data acquisition terminal is to realize the acquisition of visible and infrared images,and the interactive display terminal is to complete the processing and display.The hardware of the system is based on the ARM-DSP processor,and the multi-threaded,buffer line structure is used on the software to ensure the efficiency and real-time processing.Data exchange is performed between subsystems using a 2.4G Wi Fi network.Secondly,the characteristics of infrared and visible images are analyzed,and the visible and infrared camera calibration and image preprocessing are carried out.The camera calibration is done using a custom LED array that is recognizable in both infrared and visible light sensors.Image pre-processing process includes image filtering and image enhancement,in which the image filter is used to filter noise,and the gray scale transformation,histogram equalization and other methods are used to improve the image quality.Then,the registration of heterogeneous images is studied and implemented.Using the affine model as the transformation model between images,the method of using the semi-automated registration method based on artificial aid and the improved SURF registration method using the Canny edge diagram using image stability is proposed.To ensure that the system can accurately match the image,laying the foundation for image fusion.Next,pixel-level fusion of heterogeneous images is performed.In this paper,a simple and fast method based on pixel weighted average and a more prominent fusion method based on RPCA and NSCT transformation are proposed.Based on the pixel average weighting method,the infrared gray image and the visible gray image or the luminance component transformed by the color space are fused by weighted average.The method based on RPCA and NSCT transform uses multi-scale NSCT transform to decompose the image into low-frequency information containing background and high-frequency information containing edge details.At the same time,the sparse matrix of the two images is extracted through RPCA to guide the fusion of low-frequency information and high-frequency information.Transform domain fusion,and finally use NSCT inverse transform to reconstruct the fused high and low frequency information into a fused image.In this paper,the evaluation index of fusion images is given,and the results show that the fusion algorithm of both methods can be applied to the system in real-time and fusion effect.Finally,the testing and analysis of the complete system is carried out.Test whether each function module of the system can function properly,then construct the test set of this paper with the public data set and the real-shot data set,combine the image fusion indicators and operating parameters,verify the effect and performance of the algorithm in this system.Finally,run the whole system for test.The results of the test show that the proposed two algorithms have outstanding fusion effects,and have the conditions for application under the embedded system.The system is stable and reliable.