CN-122023286-A - Three-dimensional microscopic image processing method in stevioside crystallization process

CN122023286ACN 122023286 ACN122023286 ACN 122023286ACN-122023286-A

Abstract

The invention relates to the technical field of image processing and discloses a three-dimensional microscopic image processing method in a stevioside crystallization process, which solves systematic defects of crystal identity breakage, grain count statistical distortion and insufficient traceability precision in the prior art by event-driven dynamic identification, substance pedigree graph construction and embedded traceability, realizes triple spanning from static tracing to dynamic pedigree, batch tracing to event positioning and open loop observation to closed loop correction, provides an interpretable, verifiable and optimizable intelligent analysis means for a continuous crystallization process, ensures that crystal identification accuracy and quality traceability efficiency are substantially improved, and provides an event-level accurate regulation basis for process optimization.

Inventors

ZHENG YUE
LIU FUFENG
LI LI
GOU YANLI

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: 20260512
Application Date: 20260112

Claims (10)

1. A three-dimensional microscopic image processing method in the stevioside crystallization process is characterized by comprising the following specific steps: S1, synchronously acquiring and marking multiple physical fields, namely arranging a three-dimensional microscopic imaging device, a concentration detection probe and a temperature sensor in a crystallization tank, synchronously acquiring crystal morphology, mother liquor concentration and temperature data, predicting a process parameter change trend based on a time sequence, and marking an image frame as a steady state or transient mode; S2, dynamic coding and event capturing, namely, extracting the form and component characteristics of crystals to construct dynamic identifications, freezing the original crystal identifications and generating event marks when crushing, agglomerating or curing events are detected, simulating characteristic evolution paths through a prediction model in a steady state mode, and judging the crystals deviating from a predicted value as abnormal events; S3, establishing a father-son node relation chain for the crystals with events, expanding the crystal identification into pedigree nodes containing characteristic information, father node pointers, son node lists, event types and process parameters, storing by adopting a graph database, updating in real time, and checking the embedded grain number and a mass conservation operator; S4, embedding and tracing physical information, namely inputting process parameter changes into a prediction model to update growth dynamics parameters on line, reversely tracing back along a father node chain from a target node when a quality defect is detected, calculating causal contribution by combining process parameters and event types of ancestor nodes, and positioning a key event; s5, self-adaptive optical calibration and correction, namely monitoring photoelectric signal fluctuation of an imaging device in real time, automatically triggering dark field correction and noise model recalibration when detecting light source power drift, and synchronously correcting a photo-thermal compensation item in a growth dynamics model; And S6, generating a quality traceability report, namely extracting a defect event causal chain from the pedigree graph after each batch is finished, and automatically generating the quality traceability report containing the problem crystal pedigree, the key technological parameter nodes and the physical mechanism.
2. The method for three-dimensional microscopic image processing in stevioside crystallization process according to claim 1, wherein in the step S2, the specific steps of dynamic coding and event capturing are as follows: S2.1, extracting crystal morphology features and component features, and recording an original crystal mark and generating an event mark when a crystal breaking, agglomeration or curing event occurs; S2.2, simulating a crystal characteristic evolution path through a prediction model in a steady state mode, and judging a crystal deviating from a predicted value as an abnormal event.
3. The method for three-dimensional microscopic image processing in stevioside crystallization process according to claim 2, wherein the step S2.1 comprises the following specific steps: S2.11, extracting crystal outline morphological characteristics from a three-dimensional microscopic image, extracting crystal component characteristics from concentration detection data, and carrying out weighted fusion on the two types of characteristics to form a crystal dynamic identifier with time sequence self-adaptive evolution capability, wherein a weight coefficient is dynamically adjusted along with the change trend of working condition parameters; S2.12, detecting crystal breaking, agglomerating or curing events based on mutation characteristics of the dynamic identification, and immediately freezing the original crystal identification and generating an event mark containing event type, time stamp and material conservation check code when the event is detected.
4. The method for three-dimensional microscopic image processing in stevioside crystallization process according to claim 3, wherein the step S2.2 comprises the following specific steps: s2.21, under a steady state mode, constructing a time sequence prediction model based on historical data of crystal dynamic identification, and simulating a normal evolution path of crystal characteristics; S2.22, comparing the actually observed crystal characteristic value with the model predicted value, and judging that the crystal characteristic value is abnormal when the deviation exceeds a preset threshold value.
5. The method for three-dimensional microscopic image processing in stevioside crystallization process according to claim 4, wherein in the step S3, the specific steps of pedigree diagram construction and conservation check are as follows: S3.1, establishing a pedigree node and father-son relationship chain, establishing a father node and son node association relationship chain for crystals with broken, agglomerated or cured events, and expanding crystal identification into a pedigree node structure containing characteristic information, father node pointers, son node lists, event types and process parameters; s3.2, storing and conservation checking a graph database, wherein the graph database stores a pedigree node structure and updates a father-son relationship chain in real time, a grain number conservation and mass conservation checking mechanism is embedded in the storage process, and a data quality alarm is sent when a conservation residual error exceeds a set range.
6. The method for three-dimensional microscopic image processing in stevioside crystallization process according to claim 5, wherein the step S3.1 comprises the following specific steps: s3.11, receiving event marks and working condition parameter change trends generated in the previous steps, dynamically creating pedigree nodes according to event types and working condition drift amplitude when crystal crushing, agglomeration or curing events are detected, and building association relation chains of father nodes and child nodes; S3.12, expanding the crystal identification into a pedigree node structure containing characteristic information, father node pointers, child node lists, event types and process parameters, synchronously embedding grain number conservation and mass conservation checking mechanisms in the structure construction process, and triggering node structure correction when conservation residual errors exceed a set range.
7. The method for three-dimensional microscopic image processing in stevioside crystallization process according to claim 6, wherein the step S3.2 comprises the following specific steps: S3.21, adopting an event-driven self-adaptive graph database storage architecture to perform incremental storage and dynamic topology reconstruction on the pedigree node structure, optimizing the index structure of the father-son relationship chain in real time and establishing an event cause and effect index; S3.22, constructing a multi-physical field coupling bidirectional conservation check mechanism, synchronously embedding grain number conservation and mass conservation check operators in the storage process, triggering closed loop feedback correction through residual analysis, and starting a node structure self-repairing program and sending out a data quality alarm when the conservation residual exceeds a set range.
8. The method for three-dimensional microscopic image processing in stevioside crystallization process according to claim 7, wherein in the step S4, the specific steps of physical information embedding and tracing are as follows: s4.1, inputting a working condition parameter change trend into a prediction model, and updating crystal growth kinetic parameters on line to enable the model to be adaptively matched with working condition drift; S4.2, reversely tracing back from the pedigree target node along the father node chain when the quality defect is detected, calculating causal contribution by combining working condition parameters of all nodes and event types, and positioning key event nodes causing the defect.
9. The method for three-dimensional microscopic image processing in stevioside crystallization according to claim 8, wherein in the step S5, the adaptive optical calibration and correction comprises the following specific steps: s5.1, monitoring photoelectric signal fluctuation of an imaging device, judging the power drift amplitude of a light source, and generating a correction instruction when the drift amplitude exceeds a set range; S5.2, carrying out dark field correction and noise model recalibration according to the correction instruction, and synchronously correcting a photo-thermal compensation term in the growth dynamics model.
10. The method for three-dimensional microscopic image processing in stevioside crystallization process according to claim 9, wherein in the step S6, the specific steps of generating the quality traceability report are as follows: S6.1, identifying defect event nodes related to quality problems from the pedigree diagram, and extracting a complete causal chain along a parent-child relation chain; s6.2, converting causal chain data into a quality traceability report containing a problem crystal pedigree, key process parameter nodes and a physical mechanism.

Description

Three-dimensional microscopic image processing method in stevioside crystallization process Technical Field The invention relates to the technical field of image processing, in particular to a three-dimensional microscopic image processing method in a stevioside crystallization process. Background Industrial process analysis technology is used as a core support in the pharmaceutical and fine chemical fields, and is deepened into on-line tracking of crystal morphology in the operation of a crystallization unit. The industrial crystallization process control captures the morphology of crystals in real time through a three-dimensional microscopic imaging device, distributes unique identification for each crystal, establishes a space-time continuous growth file from nucleation to final products, realizes on-line monitoring of crystal size distribution and grain number balance, and forms a specific technical realization path in stevioside purification scenes. The prior art builds on the rigid object assumption, forcing the maintenance of crystal identity. When industrial high shear initiates crushing, agglomeration or curing, identity chain breakage causes systematic distortion of a particle number balance model, the particle number is increased virtually after crushing, the particle number is attenuated abnormally by agglomeration, and an impurity wrapping source cannot be traced to a specific process fluctuation event. In a continuous crystallizer, the stirring paddle shearing enables the crystal breaking proportion to reach a higher level, and the existing system misjudges fragments as new crystals, so that the crystal size distribution data lose substance traceability significance. Therefore, we propose a three-dimensional microscopic image processing method in the stevioside crystallization process, so as to solve the above-mentioned problems. Disclosure of Invention The invention aims to provide a three-dimensional microscopic image processing method in a stevioside crystallization process, which aims to solve the problems that when industrial high shear is used for inducing crushing, agglomeration or curing, an identity chain is broken to cause systematic distortion of a grain number balance model, the grain number is increased virtually after crushing, the agglomeration is used for causing abnormal attenuation of the grain number, and an impurity wrapping source cannot be traced back to a specific process fluctuation event. In order to achieve the purpose, the invention provides the following technical scheme that the three-dimensional microscopic image processing method in the stevioside crystallization process comprises the following specific steps: S1, synchronously acquiring and marking multiple physical fields, namely arranging a three-dimensional microscopic imaging device, a concentration detection probe and a temperature sensor in a crystallization tank, synchronously acquiring crystal morphology, mother liquor concentration and temperature data, predicting a process parameter change trend based on a time sequence, and marking an image frame as a steady state or transient mode; S2, dynamic coding and event capturing, namely, extracting the form and component characteristics of crystals to construct dynamic identifications, freezing the original crystal identifications and generating event marks when crushing, agglomerating or curing events are detected, simulating characteristic evolution paths through a prediction model in a steady state mode, and judging the crystals deviating from a predicted value as abnormal events; S3, establishing a father-son node relation chain for the crystals with events, expanding the crystal identification into pedigree nodes containing characteristic information, father node pointers, son node lists, event types and process parameters, storing by adopting a graph database, updating in real time, and checking the embedded grain number and a mass conservation operator; S4, embedding and tracing physical information, namely inputting process parameter changes into a prediction model to update growth dynamics parameters on line, reversely tracing back along a father node chain from a target node when a quality defect is detected, calculating causal contribution by combining process parameters and event types of ancestor nodes, and positioning a key event; s5, self-adaptive optical calibration and correction, namely monitoring photoelectric signal fluctuation of an imaging device in real time, automatically triggering dark field correction and noise model recalibration when detecting light source power drift, and synchronously correcting a photo-thermal compensation item in a growth dynamics model; And S6, generating a quality traceability report, namely extracting a defect event causal chain from the pedigree graph after each batch is finished, and automatically generating the quality traceability report containing the problem crystal pedigree, the key technological par