Multidimensional Information Fusion Representation Technology Based On Phase And Hyperspectral Data

With the development of the economy and the improvement of living standards,people have higher and higher requirements for the quality of fruits and vegetables,and also put forward higher requirements for their quality inspection.Taking tomato fruit as an example,the appearance characteristics and internal physiological information are two important aspects of its quality inspection.At present,the quality inspection of tomatoes mainly relies on manual work,which has problems such as low efficiency,high cost,and high misjudgment rate,which restricts the development of tomato production.With the development of technologies such as machine vision and neural networks,automated tomato defect detection methods have been widely studied and applied,providing new ideas and methods for tomato quality inspection.Aiming at the problem of incomplete characterization data under a single detection method,this study focuses on phase imaging technology,combined with hyperspectral imaging technology,and takes tomato as the research object to carry out research on the fusion detection method of agricultural product fruit quality under the two technologies.The main research contents are as follows:Based on the principle of phase imaging,using structured light imaging technology and neural network technology,the rapid reconstruction of the threedimensional shape of tomato fruit can be realized in a non-destructive and contactfree manner.First of all,this paper attempts to explore and improve the filtering link of the fringe light intensity distribution information in the traditional physical reconstruction algorithm,discusses the selection of the size of the filtering window,and proposes a self-defining algorithm for the adaptive width of the filtering window.The selection of filtering area in the traditional algorithm reduces the difficulty and subjectivity of manual operation.Then,based on the semantic segmentation network model,the deformed stripe pattern is predicted,and the color characteristics of the semantic information are used to map the height information,so as to realize the single-step reconstruction process from a single frame stripe image to a 3D height.At the same time,considering that the implementation of this method has high requirements for the data set,this research also discusses the optimization of the data set.It is expected to provide more ideas and solutions for solving the difficult problems in traditional physical algorithms and the appearance shape detection of agricultural products.Based on hyperspectral imaging technology,the spectral information of tomato fruit is collected and the relationship between its biochemical defects and spectral data is analyzed.On the basis of obtaining 3D morphological data and spectral data,an image based on edge point features and line features is proposed.registration method.Use edge extraction technology and least squares fitting algorithm to restore the spatial position and pixel area of the reference object,and then use this as a registration element to restore the spatial position of the object to be measured,and realize rapid registration of heterogeneous images through simple operations,so as to realize the real-time fusion and characterization of the three-dimensional shape information and spectral information of the sample.The deep learning 3D reconstruction algorithm proposed in this paper is based on a single structured light fringe image to directly extract the shape distribution,and to perform heterogeneous image registration based on point-line features.Threedimensional shape reconstruction and multi-dimensional information fusion research provide certain reference,and provide technical support for automatic real-time detection of agricultural product quality.