CN-121980361-A - Method for rapidly detecting seed vitality in tree seed breeding process

CN121980361ACN 121980361 ACN121980361 ACN 121980361ACN-121980361-A

Abstract

The invention relates to the technical field of tree seed selection, in particular to a rapid detection method for seed vigor in a tree seed selection process. The technical problems that in the prior art, the signal change caused by physical characteristic fluctuation is difficult to effectively distinguish from the substantial spectrum difference generated by biochemical deterioration in the seed, and the operation efficiency and the detection accuracy are seriously affected are solved. The method comprises the steps of determining a reference parameter used for correcting diffuse reflection spectrums of seeds based on spectrum data of a plurality of seeds collected in an initial stage of material changing, determining a first difference value based on an internal substance characteristic spectrum, a preset reference spectrum characteristic and a preset graph theory matching algorithm, performing compensation calculation on the first difference value based on a thickness characterization parameter, the reference parameter and a preset nonlinear attenuation model to determine a second difference value, and performing sorting operation on target seeds based on the second difference value and a preset sorting threshold. The method is used for the tree seed selection scene.

Inventors

HE BING
LI TIANJUN
FENG TONG
WANG CHAO
SUN LIQUN
CHENG YUFEI
XU QIUHUI
ZHANG WENGUANG
XIE DONGMING

Assignees

重庆西科巴岳农庄现代农业有限公司

Dates

Publication Date: 20260505
Application Date: 20260407

Claims (10)

1. The rapid detection method for the seed vitality in the tree seed breeding process is characterized by comprising the following steps: determining reference parameters for correcting diffuse reflection spectra of seeds based on spectrum data of a plurality of seeds collected at the initial stage of material changing, wherein the reference parameters comprise reference vectors for representing batch common scattering interference and thickness reference parameters for representing batch average seed coat thickness; correcting a real-time spectrum of a target seed based on the reference vector and the thickness reference parameter, and determining an internal substance characteristic spectrum of the target seed and a thickness characterization parameter, wherein the thickness characterization parameter is used for characterizing the seed coat thickness of the target seed; determining a first difference value based on the internal substance characteristic spectrum, a preset reference spectrum characteristic and a preset graph theory matching algorithm, wherein the first difference value is used for representing the deterioration degree of the internal substance of the target seed; Performing compensation calculation on the first difference value based on the thickness characterization parameter, the thickness reference parameter and a preset nonlinear attenuation model to determine a second difference value; and performing sorting operation on the target seeds based on the second difference value and a preset sorting threshold value.
2. The method for rapidly detecting seed vigor in a tree seed selection process according to claim 1, wherein the determining a reference parameter for correcting a diffuse reflectance spectrum of a seed based on spectrum data of a plurality of seeds collected in an initial stage of refueling comprises: Constructing an initial spectrum set based on spectrum data of seeds of a preset number in an initial refueling stage; Carrying out standardized correction on the initial spectrum set, and determining a corrected spectrum set after eliminating surface multiplicative errors; Performing preset matrix decomposition processing on the correction spectrum set, and determining a feature vector corresponding to the maximum feature value as the reference vector; The thickness reference parameter is determined based on a set of projection values of each spectrum of the set of corrected spectra onto the reference vector.
3. The method for rapidly detecting seed vigor in a tree seed selection process according to claim 2, wherein the determining the thickness reference parameter based on the projection value set of each spectrum in the corrected spectrum set on the reference vector comprises: sorting the projection values in the set of projection values based on a numerical value size; determining projection values of a preset proportion of the front end and the rear end of the ordered sequence as edge values; and determining an arithmetic average value of the residual projection values after the edge values are removed as the thickness reference parameter.
4. The method for rapid detection of seed vigor in a tree seed selection process according to claim 1, wherein the correcting the real-time spectrum of the target seed based on the reference vector and the thickness reference parameter, determining the internal substance characteristic spectrum of the target seed and the thickness characterization parameter, comprises: Collecting an original diffuse reflection spectrum of the target seed; Carrying out standardized correction on the original diffuse reflection spectrum, and determining a corrected spectrum after eliminating surface multiplicative errors; Determining a projection scalar of the correction spectrum on the reference vector as the thickness characterization parameter; the baseline component determined by the thickness characterization parameter and the reference vector is subtracted from the corrected spectrum to determine the internal material characteristic spectrum.
5. The method for rapid detection of seed vigor in a tree seed selection process according to claim 1, wherein the determining the first difference value based on the internal substance characteristic spectrum, a preset reference spectrum characteristic, and a preset graph-theory matching algorithm comprises: Extracting local maximum point sets in a preset characteristic wave band from the internal substance characteristic spectrum and the preset reference spectrum characteristic respectively; constructing a first local maximum point set of the target seed and a second local maximum point set of the preset reference spectrum characteristic into two node sets of a bipartite graph; when the node numbers of the two node sets are inconsistent, supplementing virtual nodes in the node set with less node numbers, and giving preset punishment weight to the connection line related to the virtual nodes; The weights of all possible connecting lines between the two node sets are determined, wherein the weights of the possible connecting lines are determined based on the difference of the two nodes in the wavelength dimension and the absorbance dimension; and determining a node pairing scheme for minimizing the weight sum of all the matched connecting lines, and determining the minimum weight sum as the first difference value.
6. The method for rapid detection of seed vigor in a tree seed selection process according to claim 5, wherein the extracting the local maximum point set in the preset characteristic band from the internal substance characteristic spectrum and the preset reference spectrum characteristic respectively includes: smoothing and filtering the internal substance characteristic spectrum and the data of the preset reference spectrum characteristic in the preset characteristic wave band to obtain smoothed spectrum data; Determining the positions of a plurality of local maxima based on the smoothed spectrum data; Determining saliency of the plurality of local maxima relative to adjacent local minima; determining a local maximum value with the saliency larger than a preset threshold value from the local maxima as an effective local maximum value; and constructing the local maximum point set based on the wavelength position and the absorbance value of the effective local maximum.
7. The method for rapid detection of seed vigor in a tree seed selection process of claim 5, further comprising: judging whether the first local maximum point set is an empty set or not; and if the first local maximum point set is an empty set, skipping the step of determining a second difference value by performing compensation calculation based on the first difference value, and determining a preset penalty constant as the second difference value, wherein the preset penalty constant is larger than the preset sorting threshold.
8. The method for rapidly detecting seed vigor in a tree seed selection process according to claim 1, wherein the determining the second difference value based on the thickness characterization parameter and the thickness reference parameter and a preset nonlinear attenuation model by performing compensation calculation on the first difference value comprises: determining an excessive thickness based on a comparison result between the thickness characterization parameter and the reference parameter; Determining an attenuation coefficient through a preset attenuation function based on the thickness excess, wherein the attenuation function is configured to take a value of one when the thickness excess indicates that the thickness characterization parameter does not exceed the thickness reference parameter, take a value of no less than a preset confidence base line when the thickness excess indicates that the thickness characterization parameter exceeds the thickness reference parameter, and take a value of no less than a preset confidence base line; and carrying out attenuation processing on the first difference value based on the attenuation coefficient, and determining the second difference value.
9. The method for rapid detection of seed vigor in a tree seed selection process according to claim 8, wherein the performing a sorting operation on the target seed based on the second difference value and a preset sorting threshold value comprises: Comparing the second difference value with the preset sorting threshold value; and if the second difference value is larger than the preset sorting threshold value, generating a seed rejecting control signal.
10. The method for rapid detection of seed vigor in a tree seed selection process of claim 9, further comprising: Determining the physical distance between the spectrum acquisition position and the rejection execution position of the target seed and the running speed of a conveyor belt; Determining an action delay time based on the physical distance and the running speed; And executing a rejecting operation on the target seed based on the action delay time.

Description

Method for rapidly detecting seed vitality in tree seed breeding process Technical Field The invention relates to the field of tree seed selection, in particular to a rapid detection method for seed vigor in a tree seed selection process. Background In automated nursery operations, tree seeds are typically screened for high throughput using a conveyor belt system equipped with near infrared spectral probes and pneumatic sorting devices, and their internal quality and vigor is assessed by analyzing the diffuse reflectance spectral characteristics of the seeds for near infrared light. Because the tree seeds have obvious geographical seed source heterogeneity, the seeds of different harvesting batches have population differences in physical characteristics such as shell thickness, basic water content and the like, so that the near infrared spectrum data generate complex signal interference in the acquisition process. In the prior art, during sorting operation of a processing assembly line, collected full-band spectrum data is generally directly input into a preset comparison model for matching analysis, and a sorting executing mechanism is controlled to reject inferior seeds according to analysis results. However, when the seed lot is replaced in the actual production process, the spectral baseline drift caused by the seed coat scattering effect, the characteristic absorption peak wavelength micrometric shift caused by the internal water loss, and the high-frequency dark current noise amplification caused by the effective photon attenuation caused by the thick-shell seeds all cause interference in accurately reading the spectral data. The traditional method is difficult to effectively distinguish the signal change caused by physical characteristic fluctuation and the substantial spectrum difference generated by biochemical deterioration in the seeds, pipeline operation is usually required to be interrupted, the model threshold value is recalibrated by relying on a manual germination test, and the continuous operation efficiency and the detection accuracy of automatic sorting equipment are seriously affected. Disclosure of Invention In order to solve the technical problems that in the prior art, signal changes caused by physical characteristic fluctuation and substantial spectrum differences generated by biochemical deterioration in seeds are difficult to distinguish effectively, pipeline operation is usually required to be interrupted, a model threshold is recalibrated by a manual germination test, and continuous operation efficiency and detection accuracy of automatic sorting equipment are seriously influenced, the invention aims to provide a rapid seed vitality detection method for a tree seed breeding process, and the adopted technical scheme is as follows: According to the method, a reference parameter used for correcting diffuse reflection spectra of seeds is determined based on spectrum data of a plurality of seeds collected in an initial stage of material changing, the reference parameter comprises a reference vector used for representing batch common scattering interference and a thickness reference parameter used for representing batch average seed coat thickness, real-time spectra of target seeds are corrected based on the reference vector and the thickness reference parameter, internal substance characteristic spectra and thickness characteristic parameters of the target seeds are determined, the thickness characteristic parameters are used for representing seed coat thicknesses of the target seeds, a first difference value is determined based on the internal substance characteristic spectra, preset reference spectrum characteristics and a preset graph theory matching algorithm, the first difference value is used for representing the deterioration degree of internal substances of the target seeds, a second difference value is determined based on the thickness characteristic parameters and the thickness reference parameter and a preset nonlinear attenuation model in a compensation calculation mode, and sorting operation is performed on the target seeds based on the second difference value and a preset sorting threshold. With reference to the first aspect, in one possible implementation manner, the method specifically includes constructing an initial spectrum set based on spectrum data of a preset number of seeds in an initial period of material changing, performing standardized correction on the initial spectrum set, determining a corrected spectrum set after surface multiplicative errors are eliminated, performing preset matrix decomposition processing on the corrected spectrum set, determining a feature vector corresponding to a maximum feature value as a reference vector, and determining a thickness reference parameter based on a projection value set of each spectrum in the corrected spectrum set on the reference vector. With reference to the first aspect, in one possible implementation manner, th