Hyperspectral quality detection method for fruits based on photon transport simulation of double-layer plate model for fruits

Abstract

The invention discloses a hyperspectral quality detection method for fruits based on photon transport simulation of a double-layer plate model for the fruits. The method comprises the following steps of establishing a double-layer plate model for fruits; carrying out photon transport simulation to obtain a luminance distribution diagram on the basis of a monte carlo method; inputting the luminance distribution diagram into a preset convolutional neural network for carrying out convolution and deconvolution; analyzing convolution and deconvolution results to obtain a source exploration distance; and carrying out quality detection on the fruits by using the source exploration distance. The double-layer plate model can be used for replacing a curved surface model, so that the complexity of the curved surface model is reduced, and meanwhile, the accuracy is ensured. Compared with the curved surface model for an apple, the simplified double-layer plate model has the advantages that the simulation time can be shortened, and meanwhile, the method has better universality (applicable to stone fruits, such as pears, with thick flesh).

Description

technical field [0001] The invention relates to the detection field, in particular to a fruit hyperspectral quality detection method based on photon transmission simulation based on a fruit double-layer plate model. Background technique [0002] Portable spectral detectors are widely used because of their small size, portability, and simple operation. At present, there have been many researches on the portable fruit detector, but there are few studies on the distance between the light source and the probe of the portable fruit detector. The research group of Tsinghua University Ding Haishu used the Monte Carlo method to align light in a multi-layer biological tissue model The average penetration depth of the photons, the average flight distance of photons and the spatial sensitivity distribution were simulated and analyzed. The results showed that the distance between the light source and the photodetector is an important factor affecting the detection sensitivity. For speci...

AI Technical Summary

Problems solved by technology

There are two disadvantages in using measured data for research. One is that the research in this area cannot obtain a large amount of sufficient data, and is affected by the variety batches, so the results obtained are very limited.

Second, in the case of less data, there are still inevitable errors in the measured data, which further leads to inaccurate results.

[0006] This scheme uses the ellipsoidal surface equation to construct the apple model. The apple surface model has a higher degree of reduction in the simulation of the actual apple tissue, and the simulated data is more accurate. However, this model is too complicated in operation. The degree method is not very practical in some biological research

In addition, only for one fruit structure model, the universality is very low

Images