Research On Image Segmentation Method Of Industrial Smoke Based On Improved CV Model

In recent years,with the development of the economy,the problem of environmental pollution has become increasingly prominent.With the development of industrial cities,the emission of industrial smoke has been increasing,leading to a series of environmental pollution problems such as ecological imbalance and environmental degradation.According to the survey,industrial pollution sources are still the main source of air pollution,so the detection of industrial smoke emissions has a very important role in protecting the environment.The blackness value of flue gas can reflect the degree of industrial smoke and dust pollution,usually using the Ringlingman blackness method for detection,but its human influence is large and low efficiency.In recent years,with the rapid development of image recognition technology,the use of image technology to detect the blackness value of smoke has become our new research direction.In the process of detection,how to separate the smoke area becomes the key factor.In this paper,we will study the method of industrial smoke segmentation for this problem and show it visually in a systematic form.The main research undertaken in this paper is as follows.(1)A new algorithm for industrial smoke image segmentation based on improved CV model is investigated.The new algorithm is designed for industrial smoke images with a small difference between the background and the target gray level,with interferents,and a large variety of smoke images.The active contour model based on the improved CV model is obtained by combining the CV model and the cross-entropy model.The model consists of a regularization term and a fidelity term.The regularization term is obtained by an iterative convolution-thresholding algorithm by transforming the fidelity term into the product of the feature function and the cross-entropy function,and then using the general function of the feature function to obtain the regularization term by a thermonuclear convolutional approximation.The iterative convolution-thresholding algorithm is used to minimize the energy generalization and speed up the convergence.The experimental results with seven other models show that the new algorithm with improved CV model yields better results and high segmentation efficiency for industrial smoke segmentation.(2)Based on the improved level set energy regular term model,the industrial smoke image is segmented.To address the problem that the level set can generate local spikes and deep valleys,or overly flat regions during its evolution,which leads to errors in the numerical computation and thus affects the stability of the iterative process and the accuracy of the segmentation.By adding a distance regularity term to the energy function,the level set function is always kept close to the symbolic distance function.By combining the logarithmic function and polynomial,the deviation of the evolutionary process is reduced to ensure the stability and effectiveness of the industrial smoke image segmentation.By comparing with other five distance regular term models,the segmentation robustness of industrial smoke image segmentation model based on improved level set energy regular term is proved.(3)Design and implementation of a visual inspection system for industrial fume based on an improved CV model.In order to show the two segmentation methods mentioned above more visually in front of the staff,a visual inspection system is designed and implemented for industrial fume and dust images.The system requirements mainly include business requirements,user requirements,functional requirements and operation environment requirements,then the overall architecture of the visual inspection system is designed,and an industrial smoke inspection system with human-computer interface is developed through matlab GUI interface.After optimization and improvement,the system can be applied in practical applications to help the government and related staff detect industrial smoke and dust emissions,and provide reference for winning the battle of blue sky.