Design And Implementation Of Image Mosaic Based On Embedded System

Image mosaicing technique match and align two or more images that are collected from different time, different imaging device or under different conditions (weather, lighting, camera angle and position, etc.) into a panoramic image. It has been widely used in areas such as video surveillance, military defense, medical imaging and so on. Past research of image matching algorithm is also limited to the computer simulation environment. However, compared to the computer system, embedded system has the advantages such as the convenience for carrying, better real-time performance, specific applications and long life cycle. Therefore, scholars have paid attention to the research of embedded system. And they do all sorts of application development based on the embedded platform.This paper mainly studies image matching algorithm is transplanted to embedded systems. Specifically we use TMS320DM642 produced by the TI company as the core of hardware platform to realize the image mosaicing. By using DSP embedded system and image mosaicing technology to realize intelligent security system design and implementation. Firstly, this paper introduce how to build the hardware platform and function, characteristics and the internal structure of hardware platform; Secondly, we study the software development environment:CCS integrated development environment, DSP/BIOS system and FVID module driver design. Thirdly, in view of SURF algorithm has the advantages of fast speed, good performance and strong robustness, etc. This paper selects the SURF algorithm to extract feature. Fourthly, after the feature exaction, feature matching and feature purification is needed, this paper use BBF algorithm to match the feature and use RANSAC algorithm to eliminate false matching points. Fifthly, the image mosaicing algorithm is successfully transplanted into DM642, complete image mosaic on DSP. Experimental results show that the algorithm is transplanted into the system, has the good image mosaicing effect,In this paper, the SURF algorithm is transplanted in the embedded systems, and it has high value. At present there are some problems need to be improved, and improvement suggestions as follows:first:although SURF algorithm is very efficient, the selection of feature points and matching both are very large amount of calculation, it is unable to achieve the real-time on DSP. Suggest research more efficient methods of feature extraction algorithm to improve; Second:optimize the code of DSP, improve the efficiency of the execution of the program; Third:the image which has noise signal and the overlap between two images is small, the blurred image mosaic effect can’t achieve a satisfactory effect, needs further research.