Segmenting the touching objects in an image has been remaining as a hot subject due to the problematic complexities, and a vast number of algorithms designed to tackle this issue have come into being since a decade ago. In this paper, a new granule segmentation algorithm is developed using saddle point as the cutting point. The image is binarized and then sequentially eroded to form a gray-scale topographic counterpart, followed by using Hessian matrix computation to search for the saddle point. The segmentation is performed by cutting through the saddle point and along the maximal gradient path on the topographic surface. The results of the algorithm test on the given real images indicate certain superiorities in both the segmenting robustness and execution time to the referenced methods.
Acquiring the size gradation for particle aggregates is a common practice in the granule related industry,and mechanical sieving or screening has been the normal method. Among many drawbacks of this conventional means,the major ones are time-consuming,labor-intensive,and being unable to provide real-time feedback for process control. In this letter,an optical sieving approach is introduced. The two-dimensional images are used to develop methods for inferring particle volume and sieving behavior for gradation purposes. And a combination of deterministic and probabilistic methods is described to predict the sieving behaviors of the particles and to construct the gradation curves for the aggregate sample. Comparison of the optical sieving with standard mechanical sieving shows good correlation.
An alternative method is proposed in this letter for describing the arbitrary shape and size for granules in 2D image.After image binarization, the edge points on contour are detected, by which the centroid of the shape in question is sought using the moment calculation.Using Principal Component Analysis(PCA), the major and minor diameters are computed.Based on the signature curve-fitting, the first-order derivative is taken so as to seek all the characteristic vertices.By connecting the vertices found, the simplified polygon is formed and utilized for shape and size descriptive purposes.The developed algorithm is run on two given real particle images, and the execution results indicate that the computed parameters can technically well describe the shape and size for the original particles, being able to provide a ready-to-use database for machine vision system to perform related data processing tasks.
Classifying the texture of granules in 2D images has aroused manifold research atten-tion for its technical challenges in image processing areas.This letter presents an aggregate texture identification approach by jointly using Gray Level Co-occurrence Probability(GLCP) and BP neural network techniques.First, up to 8 GLCP-associated texture feature parameters are defined and computed, and these consequent parameters next serve as the inputs feeding to the BP neural network to calculate the similarity to any of given aggregate texture type.A finite number of aggregate images of 3 kinds, with each containing specific type of mineral particles, are put to the identification test, experimentally proving the feasibility and robustness of the proposed method.
Chen KenWang YicongZhao PanLarry E. BantaZhao Xuemei
In this paper the design and implementation of Multi-Dimensional (MD) filter, particularly 3-Dimensional (3D) filter, are presented. Digital (discrete domain) filters applied to image and video signal processing using the novel 3D multirate algorithms for efficient implementation of moving object extraction are engineered with an example. The multirate (decimation and/or interpolation) signal processing algorithms can achieve significant savings in computation and memory usage. The proposed algorithm uses the mapping relations of z-transfer functions between non-multirate and multirate mathematical expressions in terms of time-varying coefficient instead of traditional polyphase de- composition counterparts. The mapping properties can be readily used to efficiently analyze and synthesize MD multirate filters.
This paper proposes a Back Propagation (BP) neural network with momentum enhancement aiming to achieving the smooth convergence for aggregate volumetric estimation purpose. Network inputs are first selected by optically measuring the eight geometry-related parameters from the given particle image. To simplify the network structure, principal component analysis technique is applied to reduce the input dimension. The specific network structure is finalized based on both empirical expertise and analysis on selecting the appropriate number of neurons in hidden layer. The network is trained using the finite number of randomly-picked particles. The training and test results suggest that, compared to the generic BP network, the training duration of the proposed neural network is greatly attenuated, the complexity of the network structure is largely reduced, and the estimation precision is within 2%, being sufficiently up to technical satisfaction.
In granule processing industries, acquisition of particle size and shape parameters is a common procedure, and volumetric measurement is of great importance in dealing with particle sizing and gradation. To eradicate the major drawbacks with manual gauge, this paper proposes an optical approach using Back Propagation （BP） neural network to estimate the particle volume based on the two-Dimensional （2D） image information. To achieve the better network efficiency and structure simplicity, Principal Component Analysis （PCA） is adopted to reduce the dimensions of network inputs To overcome the shortcomings of generic BP network for being slow to converge and vulnerable to being trapped in local minimum, Levenberg-Marquardt （LM） algorithm is applied to achieve a higher speed and a lower error rate. The real particle data is utilized in training and testing the presented network. The experimental result suggests that the proposed neural network is capable of estimating aggregate volume with satisfactory precision and superior to the generic BP network in terms of perforxnance capacity.
Metric measurement of digitized shapes is commonly applied in optical measuring systems.In this letter,three shape-related factors defined by the authors are used in the construction of amultiple linear regression model which is utilized to compute the circumference of the convex shapes inmillimeter unit.The model is first built upon the relationship hypothesis and then its adequacy ismathematically validated.The results of applying the developed model to the given number of convexshapes in a finite circumferential length range suggest that,in terms of percent error,the model pre-cision is to satisfaction by being within±4%.The test also shows the model’s robustness against theshape’s orientation anisotropy.
