Design And Implement Of Turbidity Detection System Based On Image Recognition

With the improving of living standards, people’s environmental awareness isalso growing. Water, as the vital source for life, naturally attracts more attention.Activities, as protecting the mother river and governing the water source, have beencontinuously carried out. In all the activites, water quality detection as a necessaryprecautionary measure to forbid the re-contamination of water, is quite important.Detection for turbidity (one of the main indicators in water quality detection) isindispensible too. The current means of detection are complicated and expensive. Inthis paper, a water turbidity monitoring system based on digital video is proposed.At first, the basic principles of the current turbidity measurement are analyzed.And the overall program flow of the water turbidity detection system based onimage recognition technology is designed associated with the basic principles.Secondly, the water turbidity image information extraction algorithm centering onthe reference light source is concretely designed. This system offers two ways toextract the image information based on the spatial domain (the way using thegradient of the image) and the frequency domain (the way using the wavelettransform), both of which were proved effectively to extract the turbidity. Thirdly,the neural network for measuring the turbidity with a back propagation structure isdesigned. The internal structure is determined by test. The neural network only getsa error rate of0.5%and the maximum error rate is lower than5%. Finally, thesystem is realized and tested in practical applications. Both of the average errorrates of the two ways are greater than0.5%and the maximum error rates reach3%.Considering the error rate of the reference turbid meter, the accuracy of the systemis about8%, which is less than10%and could fulfill the turbidity monitoring needs.