Ultra-low power consumption spectrum and vision sensing control system

By introducing environmental perception, task requirement analysis and adaptive control modules in spectral and visual sensing control systems, using machine learning algorithms to identify environmental patterns and optimize sensor parameters, the existing system's flexibility and high energy consumption in complex environments is solved. The problem of this achieves efficient and accurate information acquisition and intelligent control.

CN118939939AInactive Publication Date: 2024-11-12DICUI INTELLIGENT TECH (SHANGHAI) CO LTD
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202410969276.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-19
Publication Date
2024-11-12
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing spectral and visual sensing control systems lack flexibility and adaptability, and cannot adapt to adaptively according to real-time environmental changes, making it difficult to provide accurate and reliable data support in complex environments, and high energy consumption, limiting the system's application scenarios.

Method used

An ultra-low power spectrum plus visual sensing control system was designed, including an environment perception module, a task requirement analysis module and an adaptive control module. The environment parameters are monitored in real time through temperature and humidity, light and air quality sensors to analyze user task requirements. Adjust the sensor working mode, use machine learning algorithms such as ICA and SVM to identify environmental modes, combine AdaGrad and gradient descent optimization algorithm to generate control strategies, and optimize sensor parameters.

Benefits of technology

It realizes efficient and accurate acquisition of spectral and visual information under different environmental conditions, improves the flexibility and robustness of the system, reduces power consumption, extends the service life of the equipment, and improves the level of intelligence and work efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118939939A_ABST
    Figure CN118939939A_ABST
Patent Text Reader

Abstract

The invention discloses an ultra-low power consumption spectrum and vision sensing control system, and relates to the technical field of spectrum and vision data fusion, and the system comprises the following components: an environment sensing module which is responsible for monitoring various parameters of an environment in real time, sampling environment data according to a set sampling frequency, processing the data, and transmitting the processed data to a control module; according to the invention, through integration of all the modules, accurate and efficient acquisition of spectral information and image information of a target object is realized, environment parameters can be monitored in real time, and according to the data provided by the environment sensing module, the self-adaptive control module can be used for realizing real-time monitoring of the target object. The working mode and parameter setting of the sensor are flexibly adjusted through the self-adaptive control module, so that high-quality spectrum and visual information meeting task requirements can be obtained under different environment conditions, the flexibility and robustness of the system are improved through the self-adaptive adjustment capacity, meanwhile, the power consumption of the system is reduced, and the service life of the system is prolonged. And the service life of the equipment is prolonged.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of spectral and visual data fusion, and in particular to an ultra-low power consumption spectral plus visual sensor control system. Background Art

[0002] In the existing technologies, spectral and visual sensing control systems play an important role in multiple fields, such as environmental monitoring, industrial automation, and intelligent security. However, these systems often face some significant shortcomings that limit their performance and efficiency in practical applications.

[0003] First, traditional spectral and visual sensor control systems lack sufficient flexibility and adaptability in environmental perception. They can usually only work according to preset parameters and modes and are unable to make adaptive adjustments based on real-time environmental changes. This makes it difficult for the system to provide accurate and reliable data support when faced with complex and changing environments.

[0004] Secondly, existing systems have limitations in task requirement analysis. They often cannot accurately understand the instructions input by users, especially when the instructions involve complex spectral and visual information requirements. This may cause the system to be unable to perform tasks with the accuracy and speed expected by users, affecting the practicality of the system and user satisfaction. Due to the need for continuous environmental perception, task requirement analysis and data processing operations, a large amount of energy is consumed, which not only increases the operating cost of the system, but also limits the usage scenarios.

[0005] Therefore, in view of the deficiencies and shortcomings of the prior art, the present invention proposes an ultra-low power consumption spectrum plus visual sensing control system to solve the above problems. Summary of the invention

[0006] The purpose of the present invention is to make up for the shortcomings of the prior art and provide an ultra-low power consumption spectral and visual sensing control system, which can monitor environmental parameters in real time, analyze user task requirements, and adaptively adjust the sensor working mode according to environmental changes to meet the spectral and visual information requirements of different tasks. Through temperature, humidity, light and air quality sensors, it can accurately identify environmental patterns and make control decisions to ensure the efficiency and accuracy of information collection. The system is highly intelligent and adaptable and is suitable for various complex environments and task scenarios.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions: an ultra-low power consumption spectrum plus visual sensing control system, the system comprising the following components:

[0008] Environmental perception module: responsible for real-time monitoring of various environmental parameters, sampling environmental data according to the set sampling frequency, processing the data, and transmitting the processed data to the adaptive control module;

[0009] Task requirement analysis module: receives the task instructions input by the user, and analyzes the type, accuracy and speed requirements of the spectral and visual information required for the task, and makes real-time adjustments to the task requirements based on the information provided by the environmental perception module;

[0010] Adaptive control module: formulates the sensor's working mode and adjustment strategy based on the information provided, generates specific sensor control instructions based on the decision results, and sends the control instructions to the spectral sensor module and the visual sensor module;

[0011] Spectral sensor module: used to capture the spectral information of the target object and adjust the spectral resolution and integration time parameters of the sensor according to the instructions of the adaptive control module to meet the needs of different tasks;

[0012] Visual sensor module: used to capture image information of the target object and adjust the sensor's exposure time, sensitivity, and white balance parameters according to the instructions of the adaptive control module to adapt to the requirements of different environments and tasks.

[0013] Furthermore, the environmental perception module is configured with temperature and humidity sensors, light sensors and air quality sensors, sets the sampling frequency according to actual needs and application scenario requirements, collects environmental data in real time, amplifies and filters the collected original signals to eliminate noise and interference, improves data quality and reliability, cleans the collected data, removes outliers and invalid data, extracts useful features from the original data using independent component analysis (ICA), uses support vector machine (SVM) to analyze and model historical data to identify specific environmental patterns, and transmits the processed data to the adaptive control module.

[0014] Furthermore, the environment perception module uses independent component analysis ICA to extract useful features from the original data and construct an ICA model. The mathematical formula of the ICA model is: X=AS, where: X is an m×n observation data matrix, where m is the number of sensors or features, n is the number of observation points, A is an unknown m×m mixing matrix that describes how the source signal is mixed into the observation signal, S is an m×n source signal matrix, and the source signals are statistically independent. The ICA algorithm attempts to find an unmixing matrix W that maximizes the kurtosis of the output Y=WX. The calculation formula for the kurtosis is: Among them, y i is the i-th component of Y, E[y i ] is y i The output Y of ICA is the feature extracted from the original data.

[0015] Furthermore, the environmental perception module uses a support vector machine (SVM) to analyze and model historical data to identify specific environmental patterns, collect data related to specific environmental patterns from a historical database, extract features related to environmental patterns from raw data, clean and standardize the data, assign a unique label to each environmental pattern, and train an SVM classifier using a labeled data set. The decision function expression of SVM is: f(x) = sign(w·x+b), where: w is a weight vector, x is an input feature vector, b is a bias term, and the sign(·) function is used to determine the sign of the classification. When training the SVM model, a hyperplane is found to maximize the interval between samples of different categories, which is achieved by solving the following inequality: Among them, C is the regularization parameter, which is used to control the degree of penalty for misclassified samples, ξ i is a slack variable that allows samples to violate the interval constraint to a certain extent. i is the class label of the sample, x i It is the corresponding feature vector, collects real-time data to process the data, uses the trained SVM model to classify the real-time data, and identifies the current environmental pattern.

[0016] Furthermore, the task requirement parsing module receives task instructions input by the user through the user interface, uses natural language processing technology to understand and analyze the user input to ensure the integrity and validity of the instructions, and parses the required spectral and visual information types and accuracy requirements, as well as the task's requirements for information acquisition and processing speed according to the task instructions, obtains the current environment status information from the environmental perception module, uses a decision tree to make predictions and decisions based on the environmental information, and adjusts the information type, accuracy and speed requirements in the task requirements in real time according to the adjustment strategy and transmits them to the adaptive control module.

[0017] Furthermore, the task requirement parsing module uses a decision tree to make predictions and decisions based on environmental information, obtains the state information of the current environment from the environmental perception module as the input feature of the decision tree, and selects the optimal partitioning feature based on the information gain index. The calculation formula of the information gain is: Where: Entropy(D) is the entropy of the data set D, V is the number of values ​​of feature A, D vIt is a sample subset whose feature A takes the value of v. Until the number of node samples is too small, the decision tree is trained using historical data. During the training process, the decision tree will learn to make decisions based on the input environmental status information. When new environmental status information is input, the decision tree will traverse the tree structure from top to bottom, select the corresponding branch according to the value of the input feature, and output the action or decision that should be taken based on the current environmental information. Since the environment changes over time, the decision tree is adjusted according to the new environmental information through retraining, and the output of the decision tree is transmitted to the adaptive control module.

[0018] Furthermore, the adaptive control module receives various parameter information of the current environment from the environmental perception module, receives task requirements from the task requirement analysis module, integrates the environmental information and task requirements to form a comprehensive input data set, predefines a series of control strategies based on different environments and task requirements, uses the AdaGrad algorithm to generate new control strategies in real time, matches the predefined control strategy library according to the integrated input data set, and finds the control strategy that best meets the current environment and task requirements, selects the working mode of the spectral sensor and the visual sensor according to the control strategy, and uses the gradient descent optimization algorithm to find the best adjustment parameter combination, converts the best adjustment parameter group into specific sensor control instructions, and sends them to the spectral sensor module and the visual sensor module.

[0019] Furthermore, the adaptive control module uses the gradient descent optimization algorithm to find the best parameter combination, sets the initial value θ0 of the control strategy parameter, and defines the loss function L(θ,X t ) measures the model's performance under given parameters θ and current environment information X t Under the performance, the environment perception module obtains various parameter information of the current environment, which is represented as a vector X t , receives the specific information required for the task from the task requirement parsing module, which is used to define the loss function, using the current parameters θ and the environment information X t , calculate the loss function L(θ,X t ), calculate the gradient of the loss function with respect to the parameter θ Using gradient descent, the gradient descent update formula is: Where: η is the learning rate, which determines the step size of parameter update. Repeat the above steps until the preset maximum number of iterations is reached. After multiple iterative optimizations, the algorithm will converge to a set of parameter values ​​that are considered to be the best parameter combination.

[0020] Furthermore, the spectral sensor module sets the initial integration time and spectral resolution parameters according to preset values, and receives the adjustment requirements of the spectral resolution and integration time transmitted from the adaptive control module. The integration time adjustment is based on the spectral characteristics of the target object, the brightness of the light source, and the required signal-to-noise ratio. The mathematical expression of the integration time adjustment is: Wherein: SNR represents signal-to-noise ratio, SNR target represents target signal-to-noise ratio, SNR current is current signal-to-noise ratio, t current is current integration time, and adjustment of spectral resolution is based on analysis requirements and requirements for detail of spectral information. The light source is triggered according to the instruction to ensure that the target object is properly illuminated, and the spectral information of the target object is captured by the detector. The collected spectral data is corrected, and the corrected spectral data is analyzed to extract the spectral characteristics of the target object. The spectral characteristics obtained by analysis are output to other modules, and feedback from other modules is received to understand the effect and performance of the current working state. According to the feedback results, the working mode and parameters of the spectral sensor module are fine-tuned to optimize performance or meet new task requirements.

[0021] Furthermore, the visual sensor module sets the initial sensitivity and exposure time parameters according to the preset values, and receives the exposure time and sensitivity setting values ​​from the adaptive control module. The sensitivity is based on the ambient light intensity and the image quality requirements. The mathematical calculation formula of the sensitivity is: Among them: ISO stands for sensitivity, ISO min and ISO max They represent the minimum and maximum values ​​of the sensitivity, L represents the ambient light intensity, and L min and L max They represent the minimum and maximum values ​​of the ambient light intensity respectively. The exposure time is based on the ambient light intensity and the preset image brightness target. The camera captures the image of the target object and transmits it to the image processor, which pre-processes the captured image, uses the image processing algorithm to extract the key features in the image, uses the deep learning model to perform target detection and recognition, and sends the processed image data to other modules. Based on the feedback or new instructions received, it further adjusts its working mode and parameters to optimize performance.

[0022] Compared with the existing technology, this ultra-low power spectrum plus visual sensing control system has the following beneficial effects:

[0023] 1. The present invention realizes accurate and efficient acquisition of spectral information and image information of target objects by integrating various modules, can monitor environmental parameters in real time, and flexibly adjust the working mode and parameter settings of the sensor through the adaptive control module according to the data provided by the environmental perception module, thereby ensuring that high-quality spectral and visual information that meets the task requirements can be obtained under different environmental conditions. This adaptive adjustment capability not only improves the flexibility and robustness of the system, but also reduces the system power consumption and extends the service life of the equipment.

[0024] 2. The present invention uses a machine learning algorithm support vector machine (SVM) to model and classify environmental data, which can accurately identify specific environmental patterns and select corresponding control strategies based on the identified environmental patterns. The environmental pattern recognition technology based on machine learning enables the system to automatically adjust the working states of spectral sensors and visual sensors according to the needs under different environmental conditions, thereby realizing intelligent environmental perception and adaptive control. It not only improves the intelligence level of the system, but also better meets the needs of actual application scenarios, thereby improving work efficiency and reliability.

[0025] Other advantages, objectives and features of the present invention will be set forth in part in the following description and, in part, will be apparent to those skilled in the art based on an examination of the following or may be taught from the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0027] Figure 1 The flowchart of an ultra-low power spectrum plus visual sensing control system. DETAILED DESCRIPTION

[0028] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] Embodiment 1:

[0030] This embodiment provides an ultra-low power spectrum plus visual sensing control system, which monitors environmental parameters in real time through the environmental perception module and uses machine learning algorithms to analyze historical data to identify specific environmental patterns. The task requirement parsing module is responsible for receiving user instructions, ensuring the effectiveness of the instructions through natural language processing technology, and adjusting task requirements in real time according to environmental information. The adaptive control module formulates the working mode and adjustment strategy of the sensor based on this information to achieve precise control of the spectral sensor and the visual sensor.

[0031] Environmental perception module: Perform preliminary processing on the collected data, remove outliers and invalid data, ensure the accuracy of subsequent analysis, and build an ICA model. The mathematical formula of the ICA model is: X = AS. The algorithm is used to find an unmixing matrix W so that the kurtosis of the output Y = WX is maximized. The calculation formula of the kurtosis is: The output Y of ICA is the useful features extracted from the original data. These features can more accurately reflect the essential changes in the environment. The features extracted by ICA are standardized to eliminate the dimensional differences between different features, which is convenient for the subsequent training of the SVM model. Data related to specific environmental patterns are collected from the historical database, and a unique label is assigned to each environmental pattern. The SVM classifier is trained using the labeled data set. The decision function expression of SVM is: f(x) = sign(w·x+b). When training the SVM model, a hyperplane is found to maximize the interval between samples of different categories. This is achieved by solving the following optimization problem: Collect and process real-time data, classify the real-time data using the trained SVM model, identify the current environmental pattern, and transmit the processed data to the adaptive control module.

[0032] In the task requirement analysis module, the user inputs the required spectral and visual information type, accuracy requirements, and information acquisition and processing speed requirements of the task through the user interface. The natural language processing unit understands and analyzes the task instructions input by the user to ensure the integrity and effectiveness of the instructions. Through the environmental perception interface, the current environmental status information is obtained from the environmental perception module. The decision tree prediction unit uses the decision tree model trained with historical data to make predictions and decisions based on the current environmental status information and selects the optimal partitioning features based on the information gain index. The information gain calculation formula is: Each non-leaf node is divided until the number of node samples is too small. When new environmental status information is input, the decision tree will traverse the tree structure from top to bottom and select the corresponding branch according to the value of the input feature. The leaf node finally reached represents the decision result. According to the decision tree prediction and decision results, the information type, accuracy and speed requirements in the task requirements are adjusted in real time, and the adjusted task requirements are transmitted to the adaptive control module through the adaptive control interface.

[0033] The environment perception module continuously collects various parameter information of the current environment. The task requirement analysis module analyzes the type, accuracy and speed requirements of the spectral and visual information required for the task, and sends this information to the adaptive control module. The adaptive control module integrates the environmental parameter information sent by the environment perception module and the task requirements sent by the task requirement analysis module to form a comprehensive input data set. If there is no control strategy that fully meets the current environment and task requirements in the predefined control strategy library, the adaptive control module will use the AdaGrad algorithm to generate a new control strategy in real time. According to the selected or generated control strategy, the adaptive control module selects the working mode of the spectral sensor and the visual sensor, and uses the gradient descent optimization algorithm to find the best parameter combination to adjust the various parameters of the sensor. By setting the initial value θ0 of the control strategy parameter, the loss function L(θ,X t ) measures the model’s performance under given parameters θ and current environment information X t Under the performance, the environment perception module obtains various parameter information of the current environment, which is represented as a vector X t , receives the specific information required for the task from the task requirement parsing module, which is used to define the loss function L(θ,X t ), using the current parameters θ and environmental information X t , calculate the loss function L(θ,X t ), calculate the gradient of the loss function with respect to the parameter θ Using gradient descent, the gradient descent update formula is: The above steps are repeated until the preset maximum number of iterations is reached. After multiple iterations of optimization, the algorithm will converge to a set of parameter values ​​that are considered to be the optimal parameter combination. The adaptive control module converts the selected working mode and adjusted parameters into specific sensor control instructions and sends them to the spectral sensor and visual sensor.

[0034] The spectral sensor module sets the initial integration time and spectral resolution parameters according to the preset values, and receives the adjustment requirements of the spectral resolution and integration time transmitted from the adaptive control module. The integration time adjustment is based on the spectral characteristics of the target object, the brightness of the light source, and the required signal-to-noise ratio. The mathematical expression of the integration time adjustment is: The adjustment of spectral resolution is based on the analysis requirements and the requirements for the detail of spectral information. The adjustment calculation formula of spectral resolution is: The light source triggering module triggers the light source according to the instruction to ensure that the target object is properly illuminated. The spectral sensor module captures the spectral information of the target object through the detector and converts it into an electrical signal. The data correction module corrects the collected spectral data to eliminate system errors and background interference. The data analysis module analyzes the corrected spectral data to extract the spectral characteristics of the target object. The feedback receiving module receives feedback from the adaptive control module or other modules, and fine-tunes the working mode and parameters of the spectral sensor module according to the feedback results.

[0035] The visual sensor module sets the initial sensitivity and exposure time parameters according to the preset values, and receives the exposure time and sensitivity setting values ​​from the adaptive control module. The sensitivity is based on the ambient light intensity and image quality requirements. The mathematical calculation formula of the sensitivity is: The exposure time is based on the ambient light intensity and the preset image brightness target. The exposure time is calculated as: Under the adjusted parameters, the camera of the visual sensor module captures the image of the target object and converts the captured light signal into an electrical signal. The image processor receives the electrical signal, performs preprocessing operations on the image, and uses the image processing algorithm to extract key features from the preprocessed image. Based on the extracted features, a classifier or deep learning model is used to perform target detection and recognition. The processed image data is sent to other modules through the communication interface. Based on the feedback or new instructions received, the visual sensor module can further adjust its operating mode and parameters to optimize performance.

[0036] Embodiment 2:

[0037] An ultra-low power spectrum plus visual sensing control system in this embodiment particularly emphasizes the adaptive control module;

[0038] Receive various parameter information of the current environment, such as temperature, humidity, light intensity, and air quality, from the environmental perception module; receive the spectral and visual information type, accuracy, and speed requirements required for the task from the task requirement analysis module; integrate these environmental information and task requirements to form a comprehensive input data set.

[0039] Based on different environments and task requirements, a series of control strategies are predefined, which specify how spectral sensors and visual sensors should work in different environments and tasks.

[0040] The AdaGrad algorithm is used to analyze and learn historical data and generate new control strategies in real time. The AdaGrad algorithm automatically adjusts the learning rate and uses different learning rates to update different parameter components to speed up convergence and avoid falling into local optimal solutions.

[0041] According to the integrated input data set, matching is performed in the predefined control strategy library to find the control strategy that best meets the current environment and task requirements. According to the matched control strategy, the working mode of the spectral sensor and the visual sensor is selected, and the gradient descent optimization algorithm is used to find the best parameter combination.

[0042] The selected working mode and the adjusted parameters are converted into specific sensor control instructions, and the control instructions are sent to the spectral sensor module and the visual sensor module to adjust their working states.

[0043] Embodiment three:

[0044] Based on the first and second embodiments, the specific process steps of the system are as follows:

[0045] (1) Monitor various environmental parameters in real time according to the set sampling frequency, process the collected environmental data, extract key feature information, and transmit the processed data to the adaptive control module;

[0046] (2) The task requirement analysis module receives the task instructions input by the user, analyzes the type, accuracy and speed requirements of the spectral and visual information required for the task, and adjusts the task requirements in real time based on the information provided by the environmental perception module to ensure that the task can be successfully executed in the current environment;

[0047] (3) The adaptive control module formulates the sensor working mode and parameter adjustment strategy according to the environmental perception data and task requirements, uses the decision tree to make predictions and decisions based on the environmental information, determines the optimal spectral and visual sensor settings in the current environment, and sends them to the spectral sensor module and the visual sensor module;

[0048] (4) The spectrum sensor module adjusts the parameters of spectral resolution and integration time according to the control instructions to capture the spectral information of the target object. The visual sensor module adjusts the parameters of exposure time, sensitivity, and white balance according to the control instructions to capture the image information of the target object.

[0049] (5) Process and analyze the collected spectral and visual data, extract useful information, and output or store the processed data in an appropriate format according to task requirements.

[0050] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. An ultra-low power spectrum plus visual sensing control system, It is characterized in that The system consists of the following components: Environmental perception module: responsible for real-time monitoring of various environmental parameters, sampling environmental data according to the set sampling frequency, processing the data, and transmitting the processed data to the adaptive control module; Task requirement analysis module: receives the task instructions input by the user, and analyzes the type, accuracy and speed requirements of the spectral and visual information required for the task, and makes real-time adjustments to the task requirements based on the information provided by the environmental perception module; Adaptive control module: formulates the sensor's working mode and adjustment strategy based on the information provided, generates specific sensor control instructions based on the decision results, and sends the control instructions to the spectral sensor module and the visual sensor module; Spectral sensor module: used to capture the spectral information of the target object and adjust the spectral resolution and integration time parameters of the sensor according to the instructions of the adaptive control module to meet the needs of different tasks; Visual sensor module: used to capture the image information of the target object and adjust the sensor's exposure time, sensitivity, and white balance parameters according to the instructions of the adaptive control module to adapt to the requirements of different environments and tasks.

2. The ultra-low power spectrum plus visual sensing control system according to claim 1, It is characterized in that The environmental perception module is equipped with temperature and humidity sensors, light sensors and air quality sensors. The sampling frequency is set according to actual needs and application scenario requirements, and environmental data is collected in real time. The collected original signals are amplified and filtered to eliminate noise and interference, improve data quality and reliability, clean the collected data, remove outliers and invalid data, extract useful features from the original data using independent component analysis (ICA), analyze and model historical data using support vector machine (SVM) to identify specific environmental patterns, and transmit the processed data to the adaptive control module.

3. The ultra-low power spectrum plus visual sensing control system according to claim 2, It is characterized in that The environment perception module uses independent component analysis ICA to extract useful features from the original data and construct an ICA model. The mathematical formula of the ICA model is: X=AS, where: X is an m×n observation data matrix, where m is the number of sensors or features, n is the number of observation points, A is an unknown m×m mixing matrix that describes how source signals are mixed into observation signals, S is an m×n source signal matrix, and the source signals are statistically independent. The ICA algorithm attempts to find an unmixing matrix W that maximizes the kurtosis of the output Y=WX. The calculation formula for the kurtosis is: Among them, y i is the i-th component of Y, E[y i ] is y i The output Y of ICA is the feature extracted from the original data.

4. The ultra-low power spectrum plus visual sensing control system according to claim 2, It is characterized in that The environmental perception module uses a support vector machine (SVM) to analyze and model historical data to identify specific environmental patterns, collect data related to specific environmental patterns from a historical database, extract features related to environmental patterns from raw data, clean and standardize the data, assign a unique label to each environmental pattern, and use a labeled data set to train an SVM classifier. The decision function expression of SVM is: f(x) = sign(w·x+b), where: w is a weight vector, x is an input feature vector, b is a bias term, and the sign(·) function is used to determine the sign of the classification. When training the SVM model, a hyperplane is found to maximize the interval between samples of different categories, which is achieved by solving the following inequality: ξ i ≥0,i=1,2,…,N, where C is the regularization parameter used to control the degree of penalty for misclassified samples, ξ i is a slack variable that allows samples to violate the interval constraint to a certain extent. i is the class label of the sample, x i It is the corresponding feature vector, collects real-time data to process the data, uses the trained SVM model to classify the real-time data, and identifies the current environmental pattern.

5. The ultra-low power spectrum plus visual sensing control system according to claim 1, It is characterized in that The task requirement parsing module receives task instructions input by the user through the user interface, uses natural language processing technology to understand and analyze the user input to ensure the integrity and effectiveness of the instructions, and parses the required spectral and visual information types and accuracy requirements, as well as the task's requirements for information acquisition and processing speed according to the task instructions. It obtains the status information of the current environment from the environmental perception module, uses a decision tree to make predictions and decisions based on the environmental information, and adjusts the information type, accuracy and speed requirements in the task requirements in real time according to the adjustment strategy and transmits them to the adaptive control module.

6. The ultra-low power spectrum plus visual sensing control system according to claim 5, It is characterized in that The task requirement parsing module uses a decision tree to make predictions and decisions based on environmental information, obtains the current environmental status information from the environmental perception module as the input feature of the decision tree, and selects the optimal partitioning feature based on the information gain index. The calculation formula of information gain is: Where: Entropy(D) is the entropy of the data set D, V is the number of values ​​of feature A, D v It is a sample subset whose feature A takes the value of v. Until the number of node samples is too small, the decision tree is trained using historical data. During the training process, the decision tree will learn to make decisions based on the input environmental status information. When new environmental status information is input, the decision tree will traverse the tree structure from top to bottom, select the corresponding branch according to the value of the input feature, and output the action or decision that should be taken based on the current environmental information. Since the environment changes over time, the decision tree is adjusted according to the new environmental information through retraining, and the output of the decision tree is transmitted to the adaptive control module.

7. The ultra-low power spectrum plus visual sensing control system according to claim 1, It is characterized in that The adaptive control module receives various parameter information of the current environment from the environmental perception module, receives task requirements from the task requirement analysis module, integrates the environmental information and task requirements to form a comprehensive input data set, predefines a series of control strategies based on different environments and task requirements, uses the AdaGrad algorithm to generate new control strategies in real time, matches the predefined control strategy library based on the integrated input data set, and finds the control strategy that best meets the current environment and task requirements. According to the control strategy, the working mode of the spectral sensor and the visual sensor is selected, and the gradient descent optimization algorithm is used to find the best adjustment parameter combination, and the best adjustment parameter group is converted into specific sensor control instructions, which are sent to the spectral sensor module and the visual sensor module.

8. The ultra-low power spectrum plus visual sensing control system according to claim 7, It is characterized in that The adaptive control module uses gradient descent to find the best parameter combination using an optimization algorithm and sets the initial value θ of the control strategy parameter 0 , define the loss function L(θ,X t ) measures the model's performance under given parameters θ and current environment information X t Under the performance, the environment perception module obtains various parameter information of the current environment, which is represented as a vector X t , receives the specific information required for the task from the task requirement parsing module, which is used to define the loss function, using the current parameters θ and the environment information X t , calculate the loss function L(θ,X t ), calculate the gradient of the loss function with respect to the parameter θ Using gradient descent, the gradient descent update formula is: Where: η is the learning rate, which determines the step size of parameter update. Repeat the above steps until the preset maximum number of iterations is reached. After multiple iterative optimizations, the algorithm will converge to a set of parameter values ​​that are considered to be the best parameter combination.

9. The ultra-low power spectrum plus visual sensing control system according to claim 1, It is characterized in that The spectral sensor module sets the initial integration time and spectral resolution parameters according to the preset values, and receives the adjustment requirements of the spectral resolution and integration time transmitted from the adaptive control module. The integration time adjustment is based on the spectral characteristics of the target object, the brightness of the light source, and the required signal-to-noise ratio. The mathematical expression of the integration time adjustment is: Wherein: SNR represents signal-to-noise ratio, SNR target represents target signal-to-noise ratio, SNR current is current signal-to-noise ratio, t current is current integration time, and adjustment of spectral resolution is based on analysis requirements and requirements for detail of spectral information. The light source is triggered according to the instruction to ensure that the target object is properly illuminated, and the spectral information of the target object is captured by the detector. The collected spectral data is corrected, and the corrected spectral data is analyzed to extract the spectral characteristics of the target object. The spectral characteristics obtained by analysis are output to other modules, and feedback from other modules is received to understand the effect and performance of the current working state. According to the feedback results, the working mode and parameters of the spectral sensor module are fine-tuned to optimize performance or meet new task requirements.

10. The ultra-low power spectrum plus visual sensing control system according to claim 1, It is characterized in that The visual sensor module sets the initial sensitivity and exposure time parameters according to the preset values, and receives the exposure time and sensitivity setting values ​​from the adaptive control module. The sensitivity is based on the ambient light intensity and image quality requirements. The mathematical calculation formula of the sensitivity is: ISO = ISO min + Among them: ISO stands for sensitivity, ISO min and ISO max They represent the minimum and maximum values ​​of the sensitivity, L represents the ambient light intensity, and L min and L max They represent the minimum and maximum values ​​of the ambient light intensity respectively. The exposure time is based on the ambient light intensity and the preset image brightness target. The camera captures the image of the target object and transmits it to the image processor, which pre-processes the captured image, uses the image processing algorithm to extract the key features in the image, uses the deep learning model to perform target detection and recognition, and sends the processed image data to other modules. Based on the feedback or new instructions received, it further adjusts its working mode and parameters to optimize performance.

Citation Information

Cited By

  • ROS2-based high-performance brushless motor control system and method, and medium

    CN119276160A

  • Traditional Chinese medicine extraction control system and method based on AI recognition

    CN121253457A