The System Of Intelligent Water Level Detection Based On SoC Technology

With the development of image processing technology, the application of machinevision is more and more widely, the research development direction of the water levelmonitoring technology has turned to intelligent visual inspection which based on imageprocessing technology. The smart water level visual inspection system based on imageprocessing technology has the advantages of non-contact, little affected by theenvironment, which is the focus of the water level detection technology in recent years.Because of the existing water level detection method that based on image processingtechnology is vulnerable to external factors, which is not yet mature, and its precision ofmeasurement needs to be improved. To the problem that existing in the water leveldetection method based on image processing technology, this paper puts forward a kind ofalgorithm which takes the advantage of the phenomenon of mirror reflection on watersurface, and achieves the water level based on neural network, in addition, use thisalgorithm to design a set of water level detection system based on SoC, which provides anew idea for smart water level visual inspection technology.In this paper, we have accomplished the basic design of smart water level detectionsystem based on SoC, specific work is divided into two parts. The first part is to putforward the algorithm of the water level visual inspection which based on the principle ofmirror reflection and verify the algorithm is feasible. At first, collecting images fromdifferent water level as the training sample to establish neural network model of thecorresponding of the mapping which between image pixel coordinates and the actual waterlevel; Then, using auxiliary illuminant to produce the phenomenon of mirror reflection onthe shore, collecting water level image and extract the corresponding image coordinates;At last, according to the established neural network model, putting the image pixelcoordinates as the test sample to calculate the water level the actual height. the new waterlevel detection algorithm has completed on the MATLAB software, compared with otherwater visual detection algorithm, this algorithm need not to set up scale, get rid of the edgedetection and the water line extraction process, reducing the interference factors such aslight or light and shadow. In the precision of the range of400mm, the average error of theneural network model is0.02mm, the average error of calculation results is1.59mm. Thesecond part of the work is to realize the smart water level visual inspection system basedon SoC technology. First of all, dividing the new water level detection algorithm intospecific models by function, and then, using the VHDL language to design each model onthe Quartus II platform. CMOS camera and OV7620image converter have been adoptedas the water level image acquisition device in this System, the SoC development boardwith the main control chip EP2C35F484C8have been selected as the processor. The simulation results show that the new water level detection method is feasible. This designhas a certain practicability and reference value, which could provide the new designscheme and design thinking for the realization of the smart water level visual inspectiontechnology.