The Research And Application Of Mathematical Morphology In Image Edge Detection And Machine Vision

Mathematical morphology is a subject built on rigorous theoretical basis, effecting greatly on the theory and application of image processing, mathematical morphology has become a new kind of image processing method. On one hand, mathematical morphology is precise and complete in terms of theory, on the other hand, it is simple and graceful with regard to basic ideas, mathematical morphology is gathering momentum and is there to stay, which will surely usher in a new phase of development. Edge detection is one of the important problems and key technical challenges in image processing, having a very significant impact on higher level feature extraction, target identification and image analysis. Edge detection plays an important role in image segmentation, pattern recognition and machine vision.The dissertation regards mathematical morphology as the basic theory and image edge detection as the research object, the dissertation consists of five parts:image edge detection based on mathematical morphology, image edge detection based on mathematical morphology and wavelet transform, image edge detection based on mathematical morphology and fuzzy mathematics, the design and implementation of bar automatic counting system based on mathematical morphology, the design and implementation of signature identification system based on mathematical morphology.The first part, image edge detection based on mathematical morphology.The part summarizes the elementary knowledge of mathematical morphology and image edge detection, analyzes the ultimate principle of mathematical morphology used in image edge detection, studies the diverse function of different size, different shape, different direction structuring elements in mathematical morphology operation, proposes two mathematical morphology edge detection algorithm.In the process of edge detection based on mathematical morphology, the different structuring elements play different roles in filtering noise and keeping edge details intact, an adaptive edge detection algorithm based on multi-shape and multi-scale structuring elements is therefore proposed. The image edges are extracted using different direction and different size structuring elements. Then, the weight factors were determined adaptively by computing the information entropy so as to integrate the edges detected by multi-shape and multi-scale structuring elements. Experimental results showed the proposed algorithm suppresses the interference of noise more effectively, has higher detection accuracy, possesses better robust when applied to different images in comparison with several classical edge detection algorithm.Order morphology is the generalization of mathematical morphology, order morphology becomes general morphology by introducing sequential statistics based on classic morphology, which is an application of limited data ordering. A multiple order morphology edge detection algorithm is proposed based on the particular advantage of order morphology and used in body liver CT image, in which the influence of structuring elements and percentiles are discussed mainly, the algorithm provides important reference for multi-structuring elements and multi-percentiles order morphology transformation edge detection.The second part, image edge detection based on mathematical morphology and wavelet transform.Wavelet transform possesses good time-frequency localization characteristic and multi-scale analytical ability, mathematical morphology is a new subject based on set theory, which is very suitable for analyzing and describing geometrical feature of signal. Combining the advantages of both wavelet transform and mathematical morphology, a new edge detection algorithm is proposed. For edge detection based on mathematical morphology, We construct an anti-noise edge detection operator by improving existing operators and employs different direction linear structure elements; edge detection based on mathematical morphology can reserve details of edge effectively, ensure the continuity and integrity of edge detected. The greatest advantages of the proposed algorithm are good robustness and detection accuracy for noised images and different format images.The third part, image edge detection based on mathematical morphology and fuzzy mathematics.Fuzzy mathematics is widely used in image edge detection. In order to increase detection accuracy and save operation time, a new multiple order morphology edge detection algorithm based on partial fuzzy enhancement is proposed. We adopt fuzzy mathematics theory to enhance edge region which is located using two-dimensional histogram oblique segmentation, partial fuzzy enhancement not only emphasizes edge features, but also reduces computation, then promotes real-time; Different direction linear structuring elements and two percentiles are chosen to detect edge sub-images, the weight factors are determined adaptively by computing the information entropy so as to integrate the edge sub-images, final edge is obtained after thinning. Experimental results showed the edge detected is exquisite, continuous and intact, whose MSE and PSNR are superior to traditional methods; the algorithm possesses good robustness for noised images and different format images, cuts operation time by nearly half compared with global enhancement algorithm.The fourth part, the design and implementation of bar automatic counting system based on mathematical morphology.Bar counting is one of important bottlenecks to restrict the improvement of automation level for bar producing enterprises, so the development of high-accuracy, real-timely bar online automatic counting system is extremely urgent. A novel bar online automatic counting system is designed and realized in our laboratory, the system includes both hardware and software. Online bar image is captured、transmitted and stored by hardware with Spyder3 line CCD camera of Dalsa as major component; online bar image is pre-processed and Blob-analyzed using C++language by software combined with Sapera Processing. Mathematical morphology is introduced to remove false objects during the course of pre-procession, which establishes good foundation for Blob analysis and insures accuracy of counting. Experimental results showed the system can acquire online bar image quickly, calculate the number of bar exactly, reach the expected requirements, mathematical morphology plays an important role in the system.The fifth part, the design and implementation of signature identification system based on mathematical morphology.It is much more complex and difficult for off-line signature identification attributable to the limitation of available information, to solve the problem, a signature identification system based on mathematical morphology is proposed. The signature image is gray-scaled、two-valued、filtered、negated and normalized at the stage of preprocess; strokes which possess distinctive directional characteristics are extracted by using mathematical morphology and combining different direction and different size structuring elements; at last identification are maked for samples in accordance with length、thickness and distance of strokes. Experimental results showed the proposed system can enhance accurate rate effectively, improve real-time performance, which is a try beneficial to apply new methods for off-line signature identification, having the certain promotion application value.