The Design And Implementation Of Zynq Based Image Matching Coprocessing Unit

Image matching is a key technology in digital image processing,and it is also the basis of many computer vision applications.Image matching is the key step in the application of 3D reconstruction,image mosaic,object recognition,weapon guidance,automatic driving and so on.The software implementation of the image matching processing an image costs a few seconds.With the growing of demand for any real-time system,more and more applications need image can be calculated in real time.Therefore,the hardware based image matching system has become a hot research topic in recent years.In this paper,we design and implement the image matching coprocessor based on the Zynq-7000 based intelligent camera platform project.In order to improve the portability,the coprocessor can be transplanted to other programmable logic hardware platform by using the common video interface.The main contents of this paper are as follows: Firstly,this paper introduces the background and significance of the current image matching algorithm and its hardware acceleration scheme,and introduces the theory and algorithm of the two algorithms.Secondly,aiming at the problem of serious performance degradation due to the optimization of the speed of the algorithm,this paper proposes a method to increase the number of matching points by using the edge detail image.The method does not affect the results of the original algorithm,the use of the original image to extract the details of the image,can be parallel processing with the original algorithm.Through the software test,our algorithm can perform well.Third,combining the advantages of SIFT algorithm and ORB algorithm,the key point detection contains FAST points and extreme points,and then use the method to enhance the image matching points are proposed in this paper,design a new improved algorithm hardware structure.Through the software test,the algorithm can perform well on the basis of hardware implementation.Fourthly,the hardware structure of the improved hardware algorithm is designed and implemented.The calculation process of algorithm using a lot of windows,so the hardware need to consume a lot of storage resources.Combining with functional modules can be calculated at the same time in the module can improve the utilization efficiency of storage space.To implement FAST corner detection,we design a new hardware pipeline structure using the method of grouping comparison.To implement the descriptor,the direction of the feature points and the descriptors are generated,and the parallel computation is carried out to improve the operation speed.In this paper,we design and implement the hardware structure of image matching processing unit.From the point of view of the improved algorithm,an improved algorithm which is suitable for hardware implementation is designed according to the characteristics of hardware devices.Compared with the current mainstream implementation scheme,the coprocessing unit designed in this paper makes reasonable use of the advantages of the two algorithms,and uses the method of raising the number of matching points proposed in this paper to achieve better matching results.