The System Of Microorganism Detection Based On Image Processing

The aquatic microorganism real-time detection based on image processing is the key step of water pollution control and water quality environmental monitoring;also it is a research focus in the image processing application fields nowadays.Because of the selecting difficult of classification feature brought by aquatic microorganism deformation,now there is still a lack of research and developed product which can carry out real-time and on-line detection.It is highly necessary to build a real-time detection system aiming at the feature of aquatic microorganism so as to make a preliminary probing algorithm research on it.This project builds a system first,adding stepwise microorganism image processing algorithm then making comprehensive test and detection test.The hardware includes image acquisition and display device,carrying stage and the observation dish;Based on Visual C++ the software adopts the way UML overall design and HIPO modular development.The recognition algorithm mainly takes the deformable Red Worm and Root-knot Nematode as research object,extracting the feature of area and perimeter for recognition algorithm research.The segmentation uses OTSU double threshold to make binary image out of the gray image,after morphology de-noising,zigzag scan grades,target modification,the extracted features are put into Bayes recognition.By factorial experiment and variance analysis,the detection test chooses recognition rate as the test indexes,and the training sample capacity,exposure time,target deformation as the factor A,B,C respectively.This microorganism detection system can execute static image/dynamic video detection, continuous acquisition,can make real-time output,save images;also it can carry out the on-line/off-line image preprocessing,pattern recognition,statistical classification,and further more,generate graph-text report.By doing the factorial experiment and variance analysis of the three factors in twenty-one groups for three times,comes out the significance statistics F of each factor.The detection test shows:The training pattern capacity and exposure time have significant effect on recognition rate,while target deformation has little or no effect,that is, this system has preliminary anti-deformation capability against the microorganism object researched in this project.Under the best operation condition,the recognition rate of random test for mixed pattern meets the system designing demand.