Research And Application Of Dry Surface Layer Data Collection Method Based On Digital Image Processing

The research of formation model of Dry Surface Layer(DSL) helps to protect water resources in arid regions and the surface of the plant growth. The key to the success of research work for dry sand layer is that how to precisely and comprehensively collect all kinds of data in real time during the process of forming the DSL model.At present, the data collection of DSL model formed is mainly rely on artificial observation records or put into fixed collection devices, although such acquisition method can achieve a certain effect, but simulation has obvious drawbacks: artificial observation data is incomplete, inaccurate and the disadvantage of strong randomicity, and can not reveal the exact model of the formation of dry sand layer; Fixed acquisition device can reduce human labor, but because of it's expensive and complicated to operate, does not apply to such physical simulation experiments.Given the many shortcomings of the traditional methods, in order to provide more reliable, accurate scientific data for the DSL study, the paper presents the computer multimedia technology in which DSL data acquisition, digital image processing technology based on combining pattern recognition and other fields related technologies to collect the data needed to DSL model, such methods have the characteristics of non-contact, rapid, save cost and reduce the workload that can provide reliable data protection for the research. The main work is as follows:(1) For the non-uniform illumination noise of Markov bottle image, the paper proposes a level capable of efficiently removing image noise adaptive filtering method. The algorithm to determine the size relationship between the largest Markov bottle image filtering value, a minimum value filter and median filter value among the filter window to change the size, which can effectively protect the information Markov bottle level scale points, lines, marks and borders of a container.(2) For the binary image problem of the non-uniform illumination Markov bottle, in a variety of binarization algorithm based on comparative analysis, select Bernsen level image processing algorithms, and proposes an improved Bernsen binarization algorithm. Experiments proved that the improved Bernsen algorithm can effectively separate individual tick level image binarization best.(3) For the problem of Markov bottle image identity-cation under the non-uniform illumination, this paper proposed for a detection algorithm marks. It use Loess to smooth the vertical projection area of the curve,calculated for each extreme point, each extreme value point of difference. According to the characteristics of the binary image that the maximum is Markov bottle liquid line location. Experimental verification, under non-uniform illumination, the algorithm of glitter had a good distinguish, can effectively identify Markov bottle level line.(4)For the determination of DSL thickness value, proposed sand boundary recognition based on multiple LBP features improved AdaBoost algorithms. Firstly, the use of multiple LBP features of the sand layer image feature extraction, studied the basis of the traditional classification algorithm, its corresponding improvement, the success of texture features introduced to identify the boundaries of the problem, the final get wet layer thickness values and boundary layer of dry sand. Experimental results show that the improved algorithm in recognition accuracy and performance than existing algorithms some improvement, and the successful realization of automatic observation sand layer thickness.(5) The regression analysis of the DSL thickness and temperature. This paper use GaussNewton algorithm and rational spline interpolation algorithm known thermal infrared images of points temperature and dry layer thickness regression analysis and the establishment of the regression model, to provide more accurate data to support the model for the formation of dry sand layer. Experimental results show the effect of rational interpolation spline fitting algorithm better.(6) Design and implement a data acquisition system model DSL is formed based on image processing,automated hydrogeology in such physical simulation test, continuous and intelligent.