Study On Automatic Counting Of Microorganisms In Sewage

With the acceleration of urbanization and the improvement of people’s lives,urban sewage discharge has increased dramatically.So,it has become a very import issue to effectively improve the efficiency of sewage treatment.Studies have shown that the population structure of microorganisms in activated sludge can well predict the sewage treatment craft and effluent quality Thus,if the structural features and variation tendency of microbial population have been automatically obtained through image processing and analysis technologies,it can provide reliable evidences for scientific and timely adjustment of wastewater treatment parameters.Therefore,this automatic detection technique has higher theoretical research significance and application value.In this paper,we focused on the automatic counting methods of indicative microorganisms,and the main work and results are as follows:Aiming at the Gaussian noise in the microscopic images formed during the imaging process,an adaptive fast non-local mean filtering algorithm is designed.Firstly,we use image blocks instead of individual pixels to calculate weights,so more image edge details can be preserved.Secondly,an optimization model of the filtering parameter h is established,so that the parameters can be adaptively adjusted to improve the limitations of manual adjustment,and enhance the filtering effect at the same time.Finally,the calculation method of the integral graph is introduced to improve the operating efficiency of the algorithm.Experiments show that the algorithm can effectively remove the Gaussian noise in the image and preserve the image edge details better than the traditional filtering algorithms,which provides a good basis for the subsequent segmentation counting research.In view of the uneven grayscale and multi-peak histogram of the microbial images,a microorganism counting method based on multi-threshold optimization is proposed in this paper.Firstly,we design the multi-threshold objective function based on two-dimensional exponential entropy to increase the accuracy of image segmentation.Then,an improved heuristic particle swarm optimization algorithm is introduced to improve the operational efficiency and acquire the optimal segmentation thresholds.Subsequently,the burr spitting and adhesion of targets are removed by morphological processing and pit segmentation.Finally,the breadth-first searchalgorithm is used for targets making and counting.The experimental results show that the method can effectively count the microbial images accurately.Compared with the traditional counting methods based on segmentation,the counting accuracy is increased by about 2.5% on average,and the computational efficiency is faster 1.3s on average.When the contrast between the microbial image target and background are not obvious and the target edges are blurred,it is difficult to extract the accurate target outlines.This paper designs another microbial counting method based on multi-layer self-organizing map neural network.Firstly,a three-layer SOM neural network model is constructed to cluster the gradients of the image and avoid manual selection of gradient thresholds.Secondly,the judgment mechanism in competition layer is designed to fully detect the weak edges of the images.Then,morphological post-processing and pit segmentation methods are used to remove scattered noise and adhesion portions in the resulting image.Finally,by using of the breadth-first search algorithm,we finish the marking and counting works.Experiments show that this method can effectively extract the target creatures in the images with low contrast and blurry edges.