In the field of underwater image processing, the line and rounded objects, like mines and torpedoes, are the most common targets for rec, ognition. Before further analysis, these two image patterns need to be detected...In the field of underwater image processing, the line and rounded objects, like mines and torpedoes, are the most common targets for rec, ognition. Before further analysis, these two image patterns need to be detected and extracted from the underwater images in real-time. Using the subpixel position, direction and curvature information of an edge provided by Zernike Orthogonal Moment （ZOM） edge detection operators, an enhanced Randomized Hough Transform （RHT） to extract straight-lines is developed. This line extraction method consists of two steps ： the rough parameters of a line are obtained robustly at first using RHT with large quantization in the Hough space and then the parameters are refined with line fitting techniques. Therefore both the robustness and high precision can be achieved simultaneously. Particularly, the problem of ellipse extraction is often computationally demanding using traditional Hough Transform, since an ellipse is characterized by five parameters. Based on the generalized K-RASAC algorithm, we develop a new ellipse extraction algorithm using the concept of quadratic curve cluster and random sampling technique. We first develop a new representation of quadratic curves, which facilitates quantization and voting for the parameter A that represents a candidate ellipse among the quadratic curves. Then, after selecting two tangent points and calculating the quadratic parameter equation, we vote for the parameter A to determine an ellipse. Thus the problem of ellipse extraction is reduced into finding the local minimum in the A accumulator array. The methods presented have been applied successfully to the extraction of lines and ellipses from synthetic and real underwater images, serving as a basic computer vision module of the underwater objects recognition system. Compared to the standard RHT line extraction method and K-RANSAC ellipse extraction method, our methods have the attractive advantages of obtaining robustness and high precision simultaneously while preserving the merits of high computation speed and small storage requirement.展开更多
This paper present a new method based on Chaos Genetic Algorithm (CGA) to localize the human iris in a given image. First, the iris image is preprocessed to estimate the range of the iris localization, and then CGA is...This paper present a new method based on Chaos Genetic Algorithm (CGA) to localize the human iris in a given image. First, the iris image is preprocessed to estimate the range of the iris localization, and then CGA is used to extract the boundary of the ~iris . Simulation results show that the proposed algorithms is efficient and robust, and can achieve sub pixel precision. Because Genetic Algorithms (GAs) can search in a large space, the algorithm does not need accurate estimation of iris center for subsequent localization, and hence can lower the requirement for original iris image processing. On this point, the present localization algirithm is superior to Daugman's algorithm.展开更多
Accurate indoor 3D models are essential for building administration and applications in digital city construction and operation.Developing an automatic and accurate method to reconstruct an indoor model with semantics...Accurate indoor 3D models are essential for building administration and applications in digital city construction and operation.Developing an automatic and accurate method to reconstruct an indoor model with semantics is a challenge in complex indoor environments.Our method focuses on the permanent structure based on a weak Manhattan world assumption,and we propose a pipeline to reconstruct indoor models.First,the proposed method extracts boundary primitives from semantic point clouds,such as floors,walls,ceilings,windows,and doors.The primitives of the building boundary,are aligned to generate the boundaries of the indoor scene,which contains the structure of the horizontal plane and height change in the vertical direction.Then,an optimization algorithm is applied to optimize the geometric relationships among all features based on their categories after the classification process.The heights of feature points are captured and optimized according to their neighborhoods.Finally,a 3D wireframe model of the indoor scene is reconstructed based on the 3D feature information.Experiments on three different datasets demonstrate that the proposed method can be used to effectively reconstruct 3D wireframe models of indoor scenes with high accuracy.展开更多
We present a simple yet efficient algorithm for recognizing simple quadric primitives(plane,sphere,cylinder,cone)from triangular meshes.Our approach is an improved version of a previous hierarchical clustering algorit...We present a simple yet efficient algorithm for recognizing simple quadric primitives(plane,sphere,cylinder,cone)from triangular meshes.Our approach is an improved version of a previous hierarchical clustering algorithm,which performs pairwise clustering of triangle patches from bottom to top.The key contributions of our approach include a strategy for priority and fidelity consideration of the detected primitives,and a scheme for boundary smoothness between adjacent clusters.Experimental results demonstrate that the proposed method produces qualitatively and quantitatively better results than representative state-of-the-art methods on a wide range of test data.展开更多
文摘In the field of underwater image processing, the line and rounded objects, like mines and torpedoes, are the most common targets for rec, ognition. Before further analysis, these two image patterns need to be detected and extracted from the underwater images in real-time. Using the subpixel position, direction and curvature information of an edge provided by Zernike Orthogonal Moment （ZOM） edge detection operators, an enhanced Randomized Hough Transform （RHT） to extract straight-lines is developed. This line extraction method consists of two steps ： the rough parameters of a line are obtained robustly at first using RHT with large quantization in the Hough space and then the parameters are refined with line fitting techniques. Therefore both the robustness and high precision can be achieved simultaneously. Particularly, the problem of ellipse extraction is often computationally demanding using traditional Hough Transform, since an ellipse is characterized by five parameters. Based on the generalized K-RASAC algorithm, we develop a new ellipse extraction algorithm using the concept of quadratic curve cluster and random sampling technique. We first develop a new representation of quadratic curves, which facilitates quantization and voting for the parameter A that represents a candidate ellipse among the quadratic curves. Then, after selecting two tangent points and calculating the quadratic parameter equation, we vote for the parameter A to determine an ellipse. Thus the problem of ellipse extraction is reduced into finding the local minimum in the A accumulator array. The methods presented have been applied successfully to the extraction of lines and ellipses from synthetic and real underwater images, serving as a basic computer vision module of the underwater objects recognition system. Compared to the standard RHT line extraction method and K-RANSAC ellipse extraction method, our methods have the attractive advantages of obtaining robustness and high precision simultaneously while preserving the merits of high computation speed and small storage requirement.
文摘This paper present a new method based on Chaos Genetic Algorithm (CGA) to localize the human iris in a given image. First, the iris image is preprocessed to estimate the range of the iris localization, and then CGA is used to extract the boundary of the ~iris . Simulation results show that the proposed algorithms is efficient and robust, and can achieve sub pixel precision. Because Genetic Algorithms (GAs) can search in a large space, the algorithm does not need accurate estimation of iris center for subsequent localization, and hence can lower the requirement for original iris image processing. On this point, the present localization algirithm is superior to Daugman's algorithm.
基金supported by the National Key Research and Development Program of China(Grant No.2021YFB2501103)the National Science Foundation of China(Grant No.42271429 and 42130106)the Key Research and Development Projects of Shanghai Science and Technology Commission(Grant No.21DZ1204103).
文摘Accurate indoor 3D models are essential for building administration and applications in digital city construction and operation.Developing an automatic and accurate method to reconstruct an indoor model with semantics is a challenge in complex indoor environments.Our method focuses on the permanent structure based on a weak Manhattan world assumption,and we propose a pipeline to reconstruct indoor models.First,the proposed method extracts boundary primitives from semantic point clouds,such as floors,walls,ceilings,windows,and doors.The primitives of the building boundary,are aligned to generate the boundaries of the indoor scene,which contains the structure of the horizontal plane and height change in the vertical direction.Then,an optimization algorithm is applied to optimize the geometric relationships among all features based on their categories after the classification process.The heights of feature points are captured and optimized according to their neighborhoods.Finally,a 3D wireframe model of the indoor scene is reconstructed based on the 3D feature information.Experiments on three different datasets demonstrate that the proposed method can be used to effectively reconstruct 3D wireframe models of indoor scenes with high accuracy.
基金the National Natural Science of Foundation for Outstanding Young Scholars(12022117)the National Natural Science Foundation of China(61872354)+1 种基金the Beijing Natural Science Foundation(Z190004)the Intelligent Science and Technology Advanced subject project of University of Chinese Academy of Sciences(115200S001)。
文摘We present a simple yet efficient algorithm for recognizing simple quadric primitives(plane,sphere,cylinder,cone)from triangular meshes.Our approach is an improved version of a previous hierarchical clustering algorithm,which performs pairwise clustering of triangle patches from bottom to top.The key contributions of our approach include a strategy for priority and fidelity consideration of the detected primitives,and a scheme for boundary smoothness between adjacent clusters.Experimental results demonstrate that the proposed method produces qualitatively and quantitatively better results than representative state-of-the-art methods on a wide range of test data.
