Research On The Expansion Of Imaging Field Of View For Intelligent Microscope Based On Image Stitching

Microscopes have an irreplaceable importance in fields such as biology,medicine,and materials science.They help us observe the microscopic world,revealing the essence and inherent laws of things.With the continuous development of computer technology and artificial intelligence,microscopes are also evolving towards intelligence.The field of view of traditional microscopes decreases as the magnification increases,which brings many inconveniences to scientific research and production.Due to the limitations of the structure and physical principles of traditional optical microscopes,it is difficult or too costly to extend the field of view through optical mechanisms.Therefore,image stitching technology is a worthwhile research problem for expanding imaging fields.Specifically,this technology is based on the matching and fusion of image features,stitching multiple images with overlapping areas into a larger field of view image.Essentially,this is a method of expanding microscope functions through imaging control and image processing.A microscopic image registration algorithm based on motion information of the stage is proposed to address the problem of a large number of false matches produced by traditional image stitching algorithms when dealing with micro-images with similar structures.The algorithm predicts the position of the feature points to be matched based on the motion information of the stage,and selects the feature point with the smallest Euclidean distance within the predicted region as the actual matching point,reducing the interference of similar feature points outside the region.Coarse screening is performed by pairing the feature points based on their slopes,and precise matching is achieved using the random sample consensus algorithm to calculate the homography matrix for image registration.Experimental results show that the proposed algorithm is superior to traditional BF and FLANN matching algorithms in terms of accuracy and speed.A nonlinear weighting image fusion algorithm is proposed to address the trade-off between speed and quality in handling seams of image fusion algorithms.The linear weighting in weighted fusion algorithms is replaced by a nonlinear weighting to achieve smoother seam edges.Experimental results show that compared with weighted average fusion,wavelet fusion,and best seam line fusion algorithms,this algorithm is more suitable for real-time stitching systems of microscopic images.The language follows the standard expression for a scientific dissertationAiming at the problem of numerous feature points and long registration time in microscopic image registration,a CPU multi-threaded parallel processing strategy for image registration is proposed.The images to be stitched are divided into multiple groups of sub-block images,which are assigned to multiple threads for calculating homography matrices,and finally multiple homography matrices are merged into one matrix.The experimental results show that this method significantly shortens the image registration time while maintaining high registration accuracy.A frame-to-frame multi-image stitching method is proposed to address the problem of local image mismatch that can cause global stitching failure in the "capture first,stitch later" approach.Firstly,the microscope scanning step size is controlled to determine the image overlap region,and feature points are extracted only in the overlap region.Then,a frame with overlap region from the previous panorama result image is selected as the reference image,and only these two frames are stitched each time to effectively reduce the system computation load.The image registration and fusion evaluation metrics are used to determine whether to re-capture the images,achieving automation of the stitching system.Experimental results show that this method outperforms Auto Stitch software in terms of stitching time and quality.Finally,the above image stitching algorithm was applied to the imaging field expansion of an intelligent microscope system under development..The experiments demonstrate that this method outperforms Auto Stitch software and Image J software in terms of stitching time.The stitching effect is superior to that of Auto Stitch and comparable to that of Image J...