Image Gray Features Research Articles

Learning classification of big medical imaging data based on partial differential equation

Traditional big medical image data classification methods are mostly based on the change of image gray features, extract edge and contour feature information, or perform conversion between medical image coordinate sets. However, the algorithms are complicated, real-time performance is poor, classification speed is slow, and accuracy is low. This paper proposes a classification study of big medical image data based on partial differential equations by combined with deep learning algorithms, and uses partial differential equations in big medical image processing to extract the texture features of medical images. Moreover, according to the texture features of the medical image contrast modulation, this paper filters out the image noise interference. Based on the depth learning algorithm, the image distance stratification, the target object size, the fit and other information, the accurate classification of big medical image data is realized. Experimental results show that the proposed classification method has high efficiency, low error rate, good real-time performance and robustness.

Image segmentation and selective smoothing based on p-harmonic Mumford–Shah functional

In this work, we propose a p-harmonic Mumford–Shah (MS) functional with adaptive variable exponent 1 ≤ p(x) ≤ 2 according to image gray feature, which provides a model for image segmentation and smoothing. The paper analyzes the physical characteristics of the related p-harmonic equation in local coordinates and explains that diffusion behavior of p-harmonic is superior to that of anisotropic diffusion and isotropic diffusion in essence. Thus the proposed model is more suitable for segmentation and smoothing of noisy images with intensity inhomogeneities while simultaneously preserving edges than the piecewise smooth MS (PSMS) model. Then effective numerical scheme is constructed to handle its computation using level set method. The model is finally applied on a wide variety of image segmentation and smoothing. All these results show that the proposed model is effective.

A study of landslide deformation field with digital correlation method

Landslides are one of the main geological disasters that persist in the reservoir after the completion and dynamic water storage of the Three Gorges Dam, and it is a threat to the safety of the residents along the river and shipping. The increase of landslides around the reservoir is affected by the yearly cycle of impoundment and flood water discharged from the dam, and precipitation, etc. Hence, monitoring the reservoir accurately and timely is one of the essential means for preventing disasters like landslide in the Three Gorges. In this work, we apply the photogrammetry method to the field of landslide monitoring on the basis of previous methods. We develop a solving method of landslide displacement field based on the image gray feature matching. First, image capture equipment like camera, vidicon is utilized to photograph the dynamic and static fields or movement process so as to obtain the time series images. Then kinematic variables like position, velocity, acceleration and deformation variables like displacement and strain are calculated by analyzing the images with digital correlation method. The essential idea of the image gray feature matching method is based on the viewpoint of statistics. Images in this method are treated as two-dimensional signals for searching corresponding match between signals by using correlation method. We can match two images around the same site before and after the landslide and obtain the displacement of this point. The Woshaxi landslide of Zigui County, Hubei Province is chosen as the study object, and we obtain the displacement field over time and the displacement vector of the landslide. The overall displacement is up to 50 cm, and the value and direction of regional displacement of the landslide surface are not completely the same. The overall landslide surface move rightward in the horizontal direction, and slid downward to the river in the direction vertical to the flow direction of the river. The vertical displacement is relatively larger in comparison with the horizontal displacement. Therefore, landslide movement is a non-uniform and non-rigid body motion. By integrated use of the speckle method, gray feature search and other related serial methods, we succeed in dealing with landslide problems like small deformation relationship and large displacement movement. As a new observation method, this method also solves problems, such as the registration of collected images and the peering of gray level. We calculate the variation and direction of displacements of all points on the landslide and obtain the distribution map of the landslide displacement. The error in the indoor calibration test was within the acceptable range, which verified the validity of this method. The measuring method of the landslide displacement field proposed in this work possesses advantages of whole field, non-contact and flexible monitoring range like the traditional photogrammetry, as well as other advantages such as large deformation measuring range, good adaptability of environment, etc. Compared with the traditional point measurement method, this method may not only obtain displacement data of more landslide points, but also monitor the global deformation of the landslide and observe and predict the whole movement behavior of the landslide. This method bounds the observation range to every specific landslide, has higher displacement resolution, and is unconstrained by weather condition, attitude angle and flight cycle relative to the photogrammetry based on satellite images. More importantly, the cost of data acquisition is relatively low in this method. We can set up cameras at landslide points and realize unattended data acquisition and analysis, i.e., autonomic monitoring and early warning of landslide displacement. Consequently, this method has a good application prospect.

Tool Wear Detection Based on Visual Saliency Mechanism

Aiming at the characteristics of tool wear, a method is put forward for rapid detection, Firstly, extracted the tool image gray feature and formed gray feature image; Secondly, approximate sub-graph of three layers wavelet decomposition is treated by center-surround difference operations; On this basis, the significant images is formed by the fusion of feature saliency map. Finally, tool wear is detected through the segmentation method of oust variance and eliminated background texture information by region growing method. The experimental results show that the method can be effectively used in the monitoring of tool wear.

Soft Sensor Model for Coal Slurry Ash Content Based on Image Gray Characteristics

Coal slurry ash content is an important variable in the control of a flotation process. It can be concluded that coal slurry image gray feature has a certain correlation with ash content by analyzing coal slurry images at different ash values in this article. Based on image gray features, coal slurry ash content soft sensor models are developed by using BP Neural Network and simple eigenvalue-based Least Square Regression Method (LS), respectively. The simulation results of the two models indicate that the soft sensor model of coal slurry ash content based on the BP Neural Network is more optimal than that based on LS. The model based on BP Neural Network has a high accuracy when coal slurry ash content is higher than 35%. Through this investigation, coal slurry ash can be detected rapidly online.

Local Extreme Extraction of Fabric Gray Feature Wave

Aiming at lack of effective algorithm for image gray feature wave analysis, this paper proposes a new local extreme extraction for image gray feature, which can establish full information of image feature wave for continue analysis. Firstly, the definition of image gray feature wave is given. Secondly, methods of local extreme extraction of feature wave are detail described, and specific calculation formulas are given. Finally, some examples of local extreme extraction of different fabric images are listed. Results show that this method is fast and easy, can exactly judge local extreme of feature wave, and lay a good foundation of data for continue image analysis and recognition.