CN-121998956-A - Stevioside product impurity detection method based on image generation

CN121998956ACN 121998956 ACN121998956 ACN 121998956ACN-121998956-A

Abstract

S100, carrying out overall analysis on historically generated images formed in a continuous image generation process according to a time evolution relation, distinguishing and extracting a morphological stable region and a hidden drift region which are formed by accumulation along with time, and constructing a time bearing substrate. According to the invention, the time bearing substrate is constructed, the risk orientation line and the rhythmic updating mechanism are introduced, so that the reference expression is kept in a regulatable evolution state in the time dimension, abnormal long-term solidification is avoided, and meanwhile, the reference image participation weight is periodically received and released by combining the production rhythm, so that high-risk impurities are continuously separated from the reference expression range, and the stability, consistency and long-term reliability of stevioside impurity detection in continuous production are improved.

Inventors

ZHENG YUE
LIU FUFENG
GOU YANLI
LI LI

Assignees

天津科技大学
山东奥晶生物科技有限公司
山东奥创智能科技有限公司
邹城奥晶甜菊糖苷生物研究院

Dates

Publication Date: 20260508
Application Date: 20260129

Claims (10)

1. The stevioside product impurity detection method based on image generation is characterized by comprising the following steps of: s100, integrally analyzing historical generated images formed in a continuous image generation process according to a time evolution relationship, distinguishing and extracting a morphological stable region and a hidden drift region which are formed by accumulation along with time, and constructing a time bearing substrate; S200, based on a time bearing substrate, comparing an image newly entering an image generating process with a morphological stable region, identifying impurity signs which show a deviation trend in the current stage and are absorbed in a history generating process, and extending the impurity signs along a time evolution direction to form a risk pointing line; S300, based on the risk orientation lines, rearranging and controlling the image formation sequence in the subsequent image generation process, so that the image content corresponding to the risk orientation lines is forward displayed in the generation process and forms an opposite relation with the morphological stability region; s400, rhythmically splitting the updated rhythm of the reference image based on the opposite-impact relation, so that the reference expression presents a segmented respiratory change state in the time dimension; s500, based on the sectional respiration change state, periodically compressing and releasing the participation weight of the reference image in the image generation process according to the production rhythm, so that high-risk impurities are continuously separated from the reference expression range in the time advancing process.
2. The method for detecting impurities in an image-based steviol glycoside product according to claim 1, wherein step S100 includes: Uniformly collecting historical generated images formed in the continuous image generation process, and constructing a time sequence relation according to the sequence of image generation; Comparing and analyzing morphological change trends presented in the history generated images based on a time sequence relation, extracting image areas repeatedly appearing in a plurality of continuous time nodes to be used as morphological stable areas, and extracting image areas presenting progressive change characteristics in a time sequence to be used as hidden drift areas; carrying out uniform bearing treatment on the distribution states of the morphological stability region and the hidden drift region in the history generated image based on the time dimension, so that the reference expression states of different time phases form a continuous linked hierarchical structure; on the basis of unified bearing treatment, a time bearing substrate used for representing the evolution state of the reference expression is constructed, and a morphological stable region and a hidden drift region formed in different time stages are brought into the time bearing substrate for continuous recording and updating.
3. The method for detecting impurities in stevioside products based on image generation according to claim 2, wherein in the process of constructing a time-bearing substrate, the original generation state of the historically generated image is maintained, the morphological stability region and the hidden drift region are incorporated into the time-bearing substrate in a continuous mapping manner in a time sequence, and the evolution track of the reference expression is carried in a time dimension, so that the reference expression is converted from a static expression to an integral state with time evolution characteristics.
4. The method for detecting impurities in an image-based steviol glycoside product according to claim 2, wherein step S200 includes: Accessing the images newly entering the image generating process into the terminal time node of the time bearing substrate according to the entering sequence based on the time bearing substrate, so that the images newly entering the image generating process and the history generated images keep a continuous time evolution relation; comparing and analyzing the image newly entering the image generating process with the morphological stable region based on the time bearing substrate, extracting a deviation trend candidate region with a deviation trend with the morphological stable region, and writing in an evolution record of the time bearing substrate; correlating the historical evolution tracks of the deviation trend candidate region and the hidden drift region based on the time bearing substrate, and identifying impurity signs absorbed in the historical generation process; extending along the time evolution direction based on the impurity signs, and forming a risk orientation line used for representing a risk evolution path in the time bearing substrate.
5. The method for detecting impurities in an image-based stevioside product according to claim 4, wherein in the process of forming the risk orientation line, the spatial position of the impurity sign in the time-bearing substrate is continuously mapped with the morphological boundary, so that the risk orientation line maintains a comparison relationship with the morphological stability region in the time evolution direction, and the deviation direction and the deviation range of the impurity sign relative to the morphological stability region are recorded, so as to enhance the indication effect of the risk orientation line on the reference expression evolution state.
6. The method for detecting impurities in an image-based steviol glycoside product according to claim 4, wherein step S300 comprises: associating an image to be entered into a reference expression update sequence in a subsequent image generation process with a risk orientation line based on a time bearing substrate, so that the relationship between the image to be generated and the risk orientation line in the time evolution direction is indexed; Rearranging the image formation sequence in the subsequent image generation process based on the association result, so that the image content associated with the risk orientation line is displayed forward in the generation process; on the basis of image formation sequence rearrangement, the image content corresponding to the risk orientation line presented by forward movement and the morphological stability region are constructed by referencing the reference expression in the same time section, so that a hedging relation is formed; After the opposite-impact relation is formed, continuously acting rearrangement regulation of the image formation sequence on the subsequent image generation process, so that the pointing action of the risk pointing line in the time bearing substrate is kept continuous, and the abnormal content is limited to be fused into the reference expression again.
7. The method for detecting impurities in an image-based stevioside product according to claim 6, wherein in the image formation sequence rearrangement process, advancing of the image content corresponding to the risk orientation line in the time-bearing substrate is continuously performed across a plurality of time sections, and a time juxtaposition relationship with the morphological stability region is maintained in each time section, so that the image content corresponding to the risk orientation line is continuously in a hedging state in the reference expression evolution process.
8. The method for detecting impurities in an image-based steviol glycoside product according to claim 6, wherein step S400 includes: based on the opposite impact relation formed by the risk orientation line and the morphological stability area, carrying out rhythm division on the continuous image generation process in the time dimension, so that the updating process of the reference image is divided into mutually connected time updating sections; Based on the time updating sections, rhythm control is carried out on the updating mode of the reference expression in each time updating section, so that the reference expression absorbs generated images in the sections and keeps a stable state when the sections are switched, and a segmented respiration change state is formed; Embedding the hedging relation between the risk pointing line and the morphological stability area into each time updating section based on the sectional respiration change state, so that the hedging relation acts independently in each time updating section; Based on the sectional respiration change state, the rhythmic updating mode is continuously acted on the subsequent image generation process so as to weaken the long-term dominant effect of a single generation result on the reference expression.
9. The method for detecting impurities in an image-based stevioside product according to claim 8, wherein in a segmented respiratory variation state, the reference expression formed in each time update section is integrally carried as a history expression and participates in the evolution constraint of the subsequent time update section, so that the reference expression maintains continuous association between the stages in the time advancing process, and the generated image in any time update section is limited to form continuous dominance to the reference expression.
10. The method for detecting impurities in an image-based steviol glycoside product according to claim 8, wherein step S500 includes: Dividing the continuous production process into production stages connected with each other according to the production rhythm based on the sectional respiration change state, and enabling each production stage to correspond to a reference expression updating section; compressing participation weights of the reference image in the image generation process based on the production stages in each production stage, so that the guiding effect of the reference image on the new image is reduced in the corresponding stage; Releasing the participation weight of the reference image after the production phase is switched based on the segmented respiration change state, so that the reference image is re-participated in the image generation process in a limited time zone; By alternately implementing compression and release of the participation weight in the continuous production process, the high-risk impurities are continuously separated from the reference expression range in the time advancing process.

Description

Stevioside product impurity detection method based on image generation Technical Field The invention relates to the technical field of food quality safety detection, in particular to a stevioside product impurity detection method based on image generation. Background The stevioside product impurity detection based on image generation is to construct an analysis basis with visual detection as a core by collecting an appearance image of a product sample in the stevioside production or inspection process, introduce an image generation technology to express and complement sample forms in different states, strengthen and reconstruct the impurity forms, distribution characteristics and appearance conditions which are difficult to stably present at the image level, and then compare and analyze the generated image on the basis, so that foreign matters, abnormal particles or non-target components in the product are identified by a visual detection means. The method does not rely on single static image judgment, but takes visual detection as a main line, and amplifies the difference of the impurity and normal stevioside crystal in morphology, texture and spatial distribution through the generated diversified image expression, so that the impurity characteristics are clearer and distinguishable, and the method is suitable for improving the stability and consistency of the visual detection of the stevioside product impurity in a complex production environment. The prior art has the following defects: Under the prior art condition, the stevioside product impurity detection based on image generation generally adopts a plurality of rounds of image generation results to be overlapped to form a reference image, and the reference image is repeatedly called in the subsequent detection process to serve as a judgment basis. However, when there are small amounts of high risk impurities in the early image generation process that are not accurately distinguished, but are erroneously absorbed into the generation result and solidified as a part of the reference image, the offset expression will be continuously inherited with continued multiplexing of the reference image. Under the condition, the subsequent detection process compares new samples based on the shifted reference expression, so that similar high-risk impurities are repeatedly ignored in a detection scene with continuous time and cannot be recognized and removed in time, and further the impurity risks are continuously accumulated and diffused in a plurality of production batches and detection periods. The above information disclosed in the background section is only for enhancement of understanding of the background of the disclosure and therefore it may include information that does not form the prior art that is already known to a person of ordinary skill in the art. Disclosure of Invention The invention aims to provide an image generation-based stevioside product impurity detection method to solve the problems in the background technology. In order to achieve the above purpose, the invention provides the following technical scheme that the stevioside product impurity detection method based on image generation comprises the following steps: s100, integrally analyzing historical generated images formed in a continuous image generation process according to a time evolution relationship, distinguishing and extracting a morphological stable region and a hidden drift region which are formed by accumulation along with time, and constructing a time bearing substrate for representing a reference expression evolution state; S200, based on a time bearing substrate, comparing an image newly entering an image generating process with a morphological stable region, identifying impurity signs which show a deviation trend in the current stage and are absorbed in a history generating process, and extending the impurity signs along a time evolution direction to form a risk pointing line; S300, based on the risk orientation lines, rearranging and controlling the image formation sequence in the subsequent image generation process, so that image contents corresponding to the risk orientation lines are forwards displayed in the generation process and form a hedging relation with the morphological stability region, and abnormal contents are limited to be fused into the reference expression again; s400, rhythmically splitting the updated rhythm of the reference image based on the opposite-impact relation, so that the reference expression presents a segmented respiratory change state in the time dimension to weaken the long-term dominant effect of a single generation result on the reference expression; S500, based on the sectional respiration change state, periodically compressing and releasing the participation weight of the reference image in the image generation process according to the production rhythm, so that high-risk impurities are continuously separated from the reference expression r