Embedded FPGA (Field Programmable Gate Array) system based on multi-optical sensing technology and storage medium

Abstract

The invention discloses an embedded FPGA (Field Programmable Gate Array) system based on a multi-optical sensing technology and a storage medium, and relates to the field of FPGAs, and the embedded FPGA system comprises an adaptation module which is used for accessing heterogeneous data output by a target multi-type optical sensor and completing initial packaging and identification of data frames according to a preset format; the preprocessing module is used for extracting characteristic parameters of the access data, removing redundant information through dynamic threshold screening and outputting screened effective data; multi-type optical heterogeneous data can be efficiently accessed, data standard packaging and unique identification are completed through an intelligent adaptive format, feature parameters are accurately extracted, redundant information is dynamically screened, data validity and quality are greatly improved, computing power is dynamically distributed according to data types, scales and computing unit characteristics, and the method is suitable for large-scale data processing. Optimal utilization of computing power resources and improvement of processing efficiency are achieved, storage media and sorting are intelligently selected in combination with data read-write frequency, delay requirements and priorities, and efficient and ordered data storage is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of FPGA technology, specifically to embedded FPGA systems and storage media based on multi-optical sensing technology. Background Technology

[0002] Multi-optical sensing technology, combined with embedded FPGA systems and high-performance storage media, is one of the core technological directions in the current field of intelligent sensing. It can acquire data through multi-spectral, polarization, and other multi-dimensional optical sensing units, and then perform real-time signal processing and feature extraction via FPGA. Currently, this technology is accelerating its application in fields such as industrial inspection and intelligent security. The resolution and response speed of optical sensors continue to improve, further highlighting the parallel computing advantages of FPGAs. However, the adaptability of storage media remains a key support point for the technology's practical application.

[0003] Patent application No. 202110988325.5 discloses a turntable servo control system based on an embedded system and FPGA. This application aims to address the problem that "due to the requirement for rapid servo motion execution in turntables, most turntables employ embedded or real-time operating systems. However, with the increasing demands on turntable functionality and the growing complexity of control algorithms, the computational load of servo operations in the turntable control software is increasing; the improved performance requirements of turntables lead to an increase in the number of parallel tasks that the turntable control software must execute. Simply using an embedded system is insufficient to handle multi-task orchestration. While real-time systems can achieve multi-task parallel processing, most systems use hardware timers to generate a system clock with a period in the millisecond range, making it difficult to achieve a servo operation cycle below the millisecond level. This hinders the improvement of the turntable's dynamic control performance. Furthermore, with increasingly stringent requirements for the localization of weapon systems, imported real-time operating systems cannot meet the requirements for independent controllability."

[0004] In existing technologies, multi-optical sensor data has high bandwidth and heterogeneous characteristics. The processing power of embedded FPGA systems and the read / write speed of storage media often do not match, resulting in data loss or delay.

[0005] To this end, we propose an embedded FPGA system and storage medium based on multi-optical sensing technology. Summary of the Invention

[0006] In view of the above-mentioned shortcomings of the prior art, the present invention provides an embedded FPGA system and storage medium based on multi-optical sensing technology, which can effectively solve the problems of the prior art.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions; This invention discloses an embedded FPGA system based on multi-optical sensing technology, comprising: The system comprises the following modules: an adaptation module for receiving heterogeneous data from various types of optical sensors, and pre-encapsulating and identifying data frames according to a preset format; a preprocessing module for extracting feature parameters from the received data, filtering out redundant information using dynamic thresholds, and outputting the filtered valid data; a scheduling module for dynamically allocating computing resources within the FPGA's internal computing units based on the type and scale of the preprocessed data, and generating an optimal computing power allocation strategy; a storage control module for receiving the optimal computing power allocation strategy generated by the scheduling module, adjusting the read / write timing of the storage medium based on the results, establishing a data priority queue, and applying it to data read / write sorting; a transmission module for dynamically adapting the transmission rate and encoding method based on data characteristics, integrating a signal enhancement unit to transmit the data output from the storage module to the external application terminal in real time; and a collaborative control module for acquiring data processing progress and resource occupancy status signals in the system, generating and issuing synchronization control commands, and regulating the runtime timing and resource allocation of each upper-level module. The upper-level modules of the collaborative control module include an adaptation module, a preprocessing module, a scheduling module, a storage control module, and a transmission module. The adaptation module is interactively connected to the preprocessing module via a local area network. The preprocessing module is interactively connected to the scheduling module via a local area network. The scheduling module is interactively connected to the control module via a local area network. The control module is interactively connected to the transmission module via a local area network. The transmission module is interactively connected to the signal enhancement unit via a local area network. The transmission module is interactively connected to the collaborative control module via a local area network.

[0008] Furthermore, the heterogeneous data accessed by the adapter module consists of optical feature data output by different types of optical sensors; The encapsulation process of the preset format is as follows: Based on the type of heterogeneous data, the field length and data structure are automatically matched. The fields include a data header, a data body and a check bit. The data header integrates the sensor ID, data acquisition timestamp, data type identifier and data integrity check identifier. The identification logic generates a unique identifier code based on the sensor type encoding and data acquisition sequence. The encoding length of the identifier code dynamically adapts to the number of connected sensors, and the mapping relationship between the identifier code and the data frame is updated in real time through a preset dynamic mapping table. The update frequency of the mapping table is synchronized with the data acquisition frequency.

[0009] Furthermore, the feature parameters extracted by the preprocessing module include data amplitude feature parameters, frequency feature parameters, spatiotemporal correlation coefficients between data, and real-time data fluctuation coefficients; The formula for calculating the dynamic threshold is: ; In the formula: For dynamic filtering thresholds; Statistical feature weighting factors; The average amplitude of historical data within a preset time window; This is the variance correction factor; The standard deviation of the amplitude of historical data within the preset time window; Real-time fluctuation influencing factors; It is a correlation correction factor; This is the amplitude fluctuation coefficient between the current data block and the previous data block; This is the average spatiotemporal correlation coefficient between the current data block and historical data blocks of the same type; This is a correction factor for environmental interference. The redundant information is that the feature parameters exceed the threshold. The corresponding numerical range or satisfying Duplicate data blocks exceeding a preset threshold are filtered to select valid data whose feature parameters are within the threshold. Within the corresponding numerical range and Data blocks that do not exceed the preset relevant threshold.

[0010] Furthermore, the optimal computing power allocation strategy is generated in the scheduling module as follows: Calculate the quantized value of the computing power requirement per unit time for the k-th type of data. ; based on Calculate the computing power allocation ratio of the i-th processing unit. ; In the formula: The processing priority weight for the k-th type of data; Let k be the transmission volume per unit time for the k-th type of data; This represents the computational complexity coefficient corresponding to the k-th data type. The allowable processing delay for the k-th type of data; For data format adaptation coefficients; The adaptation coefficient for the processing of the k-th type of data by the i-th processing unit; This represents the percentage of idle computing power in the i-th computing unit in real time. is the energy efficiency coefficient of the i-th processing unit; This represents the total number of computing units involved in scheduling within the FPGA. This is a resource redundancy adjustment factor. This represents the system's current total remaining computing power. The scheduling module according to The computing resources of each computing unit are allocated to the corresponding data processing tasks, and the system is recalculated in real time when new data types are added, the status of computing units changes, or the computing power demand fluctuates beyond the preset range. and To dynamically update the computing power allocation strategy.

[0011] Furthermore, the storage medium in the storage control module includes static random access memory, dynamic random access memory, and flash memory; During the operation of the storage control module, the storage medium is dynamically selected based on the data read / write frequency and allowed access latency: data with high read / write frequency and low latency requirements are preferentially stored in static random access memory, data with medium read / write frequency and latency requirements are stored in dynamic random access memory, and data with low read / write frequency and large capacity requirements are stored in flash memory. The data priority queue is sorted based on comprehensive priority. ,in Weighting for data real-time requirements, For subsequent data processing, urgency weights, The weights are determined by the proportion of data size, and , , The sum is 1. Scoring based on data real-time requirements, Assess the urgency of subsequent data processing. Score the percentage of data size; The storage control module adapts the optimal computing power allocation strategy by adjusting the read / write clock frequency and data bus width of the storage medium. When multiple data initiate read / write requests simultaneously, the read / write operations are executed in descending order of comprehensive priority P. The P value of each data is updated in real time at preset time windows to dynamically adjust the read / write order.

[0012] Furthermore, the aforementioned , , The calculation logic is as follows: ; In the formula: The effective duration of data storage; The duration the data has been stored; This is the quantized value of the sensor real-time performance level corresponding to the kth type of data; This represents the maximum quantization value for the real-time performance level of the sensors supported by the system. , Weighting factors include effective duration of operation as a percentage of weight and sensor real-time performance level as a percentage of weight. This is the offset of the real-time scoring threshold; This is the scoring adjustment factor; ; In the formula: This is the deadline for subsequent data processing tasks; The current system time; For the time the data was generated; , , Weights are assigned based on the proportion of processing time remaining, the completion rate of dependent tasks, and the inherent urgency level. The number of tasks that have been completed in the prerequisite tasks required for subsequent data processing; This represents the total number of prerequisite tasks required for subsequent data processing. The inherent urgency coefficient for the k-th type of data; This is a calibration factor for urgency. ; In the formula: This represents the storage capacity of the current data block. Weights for data storage density; This represents the total number of data blocks to be processed in the current system. Let m be the storage capacity of the m-th data block to be processed; This is a storage capacity adjustment item; The amplitude is affected by storage pressure; This refers to the storage pressure response coefficient. This represents the current actual occupancy rate of the storage medium. The preset safe occupancy threshold for the storage medium; The , , After calculation using the above formula and normalization calibration, , , The values ​​are all in the range of 0 to 1.

[0013] Furthermore, the dynamic adaptation of the transmission rate of the transmission module conforms to: Under the premise of meeting the data transmission delay requirements, the transmission rate is dynamically adjusted based on the current amount of data to be transmitted, the available bandwidth of the channel, and the channel bit error rate to maximize the data transmission efficiency, and the adjustment range of the transmission rate does not exceed the preset range of the maximum supported transmission rate and the minimum stable transmission rate of the channel. The encoding methods include low-density parity-check codes, convolutional codes, and adaptive variable-length codes. The encoding method is dynamically selected based on the structured attributes of the data and a preset bit error rate threshold. For structured data, low-density parity-check codes should be used preferentially. For unstructured data, adaptive variable-length coding should be used preferentially; When the data is semi-structured, or the channel bit error rate is within a preset intermediate range, and data transmission needs to balance accuracy and efficiency, convolutional codes are preferred. Furthermore, the key parameters of the convolutional codes are dynamically optimized based on real-time channel conditions and data characteristics. These dynamically optimized key parameters include: Convolutional code bit rate ,in This is the fault tolerance adjustment coefficient. For bitrate calibration factor, >0, used to balance encoding efficiency and decoding complexity; constraint length ,in Based on the length of the basic constraint, For the constraint length increment coefficient, This indicates the floor function. Score the importance of the data.

[0014] Furthermore, the resource occupancy status signals monitored by the collaborative control module include the computing power utilization rate of the computing unit, the storage space occupancy rate of the storage medium, the bandwidth occupancy rate of the transmission link, and the processing latency of each module. The generation of the synchronization control command is based on the processing progress difference and resource utilization deviation between modules: ; In the formula: To address the schedule discrepancy; This represents the processing progress of the i-th module; The processing progress of the j-th module associated with the i-th module; This is due to deviation in resource utilization. This refers to the actual utilization rate of resources. The preset security threshold for resources; when Exceeding the preset schedule deviation range or When the deviation exceeds the preset occupancy rate range, the collaborative control module generates synchronization control commands by adjusting the operating clock frequency, data input rate, or computing power allocation ratio of the corresponding module.

[0015] On the other hand, a storage medium has a built-in processor and a computer program stored on it. When the computer program is executed by the processor, it implements the running program of an embedded FPGA system based on multi-optical sensing technology.

[0016] Compared with the known prior art, the technical solution provided by this invention has the following beneficial effects: This invention provides an embedded FPGA system and storage medium based on multi-optical sensing technology. During execution, the system can efficiently access various types of heterogeneous optical data, complete data standardization and unique identification through intelligent format adaptation, accurately extract feature parameters and dynamically filter redundant information, significantly improving data validity and quality. At the same time, it dynamically allocates computing power according to data type, scale and computing unit characteristics to achieve optimal utilization of computing resources and improve processing efficiency. Furthermore, it intelligently selects and sorts storage media based on data read / write frequency, latency requirements and priority to ensure efficient and orderly data storage. Furthermore, the transmission rate and encoding method are dynamically adapted based on data characteristics and channel status to enhance signal transmission stability and accuracy. By monitoring the processing progress and resource occupancy status in real time, the operation sequence of each link is precisely controlled to achieve full-process collaborative linkage, effectively reduce resource consumption, improve system adaptability and operational stability, and fully meet the real-time, accuracy and reliability requirements of optical data processing in different scenarios. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0018] Figure 1 This is a schematic diagram of an embedded FPGA system based on multi-optical sensing technology. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0020] The present invention will be further described below with reference to embodiments.

[0021] Example: The embedded FPGA system based on multi-optical sensing technology in this embodiment, such as Figure 1 As shown, it includes: The adapter module is used to receive heterogeneous data output from various types of optical sensors of the target and to complete the initial encapsulation and identification of the data frames according to the preset format. In this embodiment, the adaptation module can access different types of optical sensing data according to different application scenarios. In the basic configuration, the system supports accessing light intensity sensing data, spectral sensing data, phase sensing data, and polarization state sensing data; however, the module can select to access two or more combinations of the above data types according to specific application requirements. For example, in laser welding quality monitoring applications, the system accesses light intensity sensing data corresponding to laser reflected light, spectral sensing data corresponding to plasma radiation, and light intensity sensing data of a specific band corresponding to molten pool thermal radiation. The pre-defined format encapsulation process is as follows: Based on the type of heterogeneous data, the field length and data structure are automatically matched. The fields include a data header, a data body and a check bit. The data header integrates the sensor ID, data acquisition timestamp, data type identifier and data integrity check identifier. The identification logic generates a unique identifier based on the sensor type encoding and the data acquisition order. The encoding length of the identifier is dynamically adapted to the number of connected sensors, and the mapping relationship between the identifier and the data frame is updated in real time through a preset dynamic mapping table. The update frequency of the mapping table is synchronized with the data acquisition frequency. The preprocessing module is used to extract feature parameters from the access data, filter out redundant information through dynamic thresholds, and output the filtered valid data. The preprocessing module extracts feature parameters including data amplitude feature parameters, frequency feature parameters, spatiotemporal correlation coefficients between data, and real-time data fluctuation coefficients. The frequency feature parameters are spectral features extracted after frequency domain transformation of continuously sampled time-series optical data, including but not limited to signal dominant frequency, spectral energy distribution, and the proportion of high-frequency / low-frequency components. Through frequency domain analysis methods such as Fourier transform and wavelet transform, the time-domain optical signal is converted to the frequency domain, extracting feature parameters reflecting the signal's periodic changes, fluctuation patterns, and frequency components. For example, in laser welding monitoring, the spectral characteristics of the plasma radiation signal can reflect the stability of the welding process; high-frequency component anomalies are usually related to transient defects such as spatter and porosity, while low-frequency component changes reflect the overall dynamic evolution of the molten pool. The formula for calculating the dynamic threshold is: ; In the formula: For dynamic filtering thresholds; Statistical feature weighting factors; The average amplitude of historical data within a preset time window; This is the variance correction factor; The standard deviation of the amplitude of historical data within the preset time window; Real-time fluctuation influencing factors; It is a correlation correction factor; This is the amplitude fluctuation coefficient between the current data block and the previous data block; This is the average spatiotemporal correlation coefficient between the current data block and historical data blocks of the same type; This is a correction factor for environmental interference. The above formula integrates the mean and standard deviation of historical data amplitudes, amplitude fluctuations between the current data block and the previous data block, average spatiotemporal correlation of similar historical data blocks, and noise levels in the optical sensing environment. By dynamically adjusting the screening criteria with weights and correction coefficients related to data reliability, sensor confidence level, fluctuation degree, correlation fit, and noise level, it achieves accurate removal of redundant information and efficient retention of effective data, thereby adapting to screening needs under different data quality and environmental conditions. Redundant information is when feature parameters exceed the threshold. The corresponding numerical range or satisfying Duplicate data blocks exceeding a preset threshold are filtered to select valid data whose feature parameters are within the threshold. Within the corresponding numerical range and Data blocks that do not exceed the preset relevant thresholds; in, The preset value range is 0.8-1.5. The higher the data reliability level, the larger the value, and the lower the data reliability level, the smaller the value. The preset value range is 0.5-2.0. The higher the confidence level corresponding to the sensor's measurement accuracy level, the larger the value; the lower the confidence level, the smaller the value. The preset value range is 0.3-1.8, representing the amplitude fluctuation coefficient between the current data block and the previous data block. The larger the value, the larger the value. The smaller the value, the smaller the value. The preset value range is 0.4-1.6, representing the average spatiotemporal correlation coefficient between the current data block and historical data blocks of the same type. The smaller the value, the larger the range. The larger the value, the smaller the value. ≥0, , These represent the average magnitude of the current data block and the average magnitude of the previous data block, respectively. The value range is preset to [0,1]. Its value increases with the improvement of the spatial distribution characteristics of the current data block and the historical data block of the same type, the consistency of the time change trend, and the stability of the data acquisition environment. It decreases with the increase of the spatial distribution difference between the two, the deviation of the time change pattern, and the fluctuation of the data acquisition environment. The preset value range is 0.7-1.3. The higher the noise level of the optical sensing environment, the larger the value; the lower the noise level, the smaller the value. The scheduling module is used to dynamically allocate computing resources of the FPGA internal computing units according to the type and scale of the preprocessed data, and generate the optimal computing power allocation strategy. The FPGA's internal computing units within the scheduling module include an arithmetic logic unit, a digital signal processing unit, and a dedicated optical data processing accelerator. The arithmetic logic unit performs basic arithmetic and logical operations during data processing. It receives data processing instructions based on the system's instruction set and completes addition, subtraction, multiplication, division, and logical AND, OR, and NOT operations according to a preset timing sequence. The digital signal processing unit handles signal processing tasks such as signal filtering and Fourier transform in optical data. It receives the data to be processed through a built-in dedicated multiplier-accumulator array and performs operations in parallel according to a preset signal processing flow. The dedicated optical data processing accelerator performs specialized processing on optical-specific data such as spectra, phases, and polarization states. Based on hardware circuits embedded with optical data processing algorithms, it directly receives the adapted optical data and executes preset processing steps. The optimal computing power allocation strategy is generated as follows: Calculate the quantized value of the computing power requirement per unit time for the k-th type of data. ; The above formula integrates the processing priority weight of the k-th type of data, the transmission scale per unit time, the computational complexity coefficient, the allowable processing delay, and the data format adaptation coefficient to quantify the data computing power requirements per unit time in multiple dimensions. Among them, the priority weight matches the importance of the application scenario, the transmission scale and computational complexity reflect the resource consumption of data processing, the allowable processing delay constrains the time requirements, and the data format adaptation coefficient adapts to different encapsulation structures, thereby achieving an accurate characterization of the computing power requirements of various types of data and providing an effective basis for subsequent computing power allocation. based on Calculate the computing power allocation ratio of the i-th processing unit. ; The above formula is based on the quantified value of the computing power requirement of the k-th type of data, combined with the processing adaptation coefficient of the i-th computing unit for this type of data, the proportion of real-time idle computing power and the energy efficiency coefficient, and introduces the total remaining computing power of the system and the resource redundancy adjustment factor. The computing power allocation ratio of each computing unit is determined through normalization calculation, which not only ensures that the computing power allocation is accurately matched with the data demand, but also takes into account the adaptability, idle state and energy efficiency of the computing units, and can also cope with changes in the remaining computing power of the system, so as to achieve the optimal allocation of computing power resources. In the formula: The processing priority weight for the k-th type of data; Let k be the transmission volume per unit time for the k-th type of data; This represents the computational complexity coefficient corresponding to the k-th data type. The allowable processing delay for the k-th type of data; For data format adaptation coefficients; The adaptation coefficient for the processing of the k-th type of data by the i-th processing unit; This represents the percentage of idle computing power in the i-th computing unit in real time. is the energy efficiency coefficient of the i-th processing unit; This represents the total number of computing units involved in scheduling within the FPGA. This is a resource redundancy adjustment factor. This represents the system's current total remaining computing power. The scheduling module is based on The computing resources of each computing unit are allocated to the corresponding data processing tasks, and the system is recalculated in real time when new data types are added, the status of computing units changes, or the computing power demand fluctuates beyond the preset range. and To dynamically update the computing power allocation strategy; in, ∈ (0,1), the higher the importance level of the application scenario corresponding to the data, the larger the value; the lower the importance level of the application scenario, the smaller the value. The unit is bits per second or bytes per second; ∈[0.1,5], the more instruction cycles required for data processing, the larger the value; the fewer instruction cycles required for data processing, the smaller the value. ∈[0.8,1.5], the more fields a data encapsulation format has and the more complex its structure, the larger the value; the fewer fields a data encapsulation format has and the simpler its structure, the smaller the value. ∈ (0,1), the higher the matching degree between the hardware architecture of the operation unit and the processing type of the k-th data, the larger the value; the lower the matching degree, the smaller the value. ∈[0,1,0,9], the lower the unit computing power energy consumption of the computing unit, the larger the value; the higher the unit computing power energy consumption of the computing unit, the smaller the value. ∈ (0.05, 0.3), the higher the total remaining computing power of the system, the larger the value; the lower the total remaining computing power of the system, the smaller the value. The storage control module is used to receive the optimal computing power allocation strategy generated by the scheduling module, adjust the read and write timing of the storage medium based on the results, establish a data priority queue and apply it to the data read and write sorting. The storage media in the storage control module include static random access memory, dynamic random access memory, and flash memory; During the operation of the storage control module, the storage medium is dynamically selected based on the data read / write frequency and allowed access latency: data with high read / write frequency and low latency requirements are preferentially stored in static random access memory, data with medium read / write frequency and latency requirements are stored in dynamic random access memory, and data with low read / write frequency and large capacity requirements are stored in flash memory. The data priority queue is sorted based on overall priority. ,in Weighting for data real-time requirements, For subsequent data processing, urgency weights, The weights are determined by the proportion of data size, and , , The sum is 1. Scoring based on data real-time requirements, Assess the urgency of subsequent data processing. Score the percentage of data size; The above formula assigns corresponding weights to the real-time requirements of data, the urgency of subsequent processing, and the proportion of data size, and the sum of the weights of the three is 1. The weighted sum of the three types of scores yields the comprehensive priority of the data. The weight allocation can adapt to the emphasis on real-time, urgency, and storage scale in different scenarios, realize the scientific quantification of data read and write sorting, and provide a basis for efficient read and write scheduling of storage media. The storage control module adapts to the optimal computing power allocation strategy by adjusting the read and write clock frequency and data bus width of the storage medium. When multiple data initiate read and write requests at the same time, the read and write operations are executed in order of comprehensive priority P from high to low. The P value of each data is updated in real time at each preset time window to dynamically adjust the read and write order. , , The calculation logic is as follows: ; In the formula: The effective duration of data storage; The duration the data has been stored; This is the quantized value of the sensor real-time performance level corresponding to the kth type of data; This represents the maximum quantization value for the real-time performance level of the sensors supported by the system. , Weighting factors include effective duration of operation as a percentage of weight and sensor real-time performance level as a percentage of weight. This is the offset of the real-time scoring threshold; This is the scoring adjustment factor; The above formula combines the ratio of the effective storage time of the data to the stored time, the ratio of the real-time level of the corresponding sensor to the maximum real-time level of the system, and sets the weight of the effective storage time ratio and the weight of the sensor real-time level. Combined with the real-time scoring threshold offset and the scoring adjustment coefficient, the scoring can sensitively reflect the changes in the real-time performance of the data, while adapting to the differences in the sensitivity of different data to real-time performance, and ensuring that the scoring fluctuates smoothly within a reasonable range. ; In the formula: This is the deadline for subsequent data processing tasks; The current system time; For the time the data was generated; , , Weights are assigned based on the proportion of processing time remaining, the completion rate of dependent tasks, and the inherent urgency level. The number of tasks that have been completed in the prerequisite tasks required for subsequent data processing; This represents the total number of prerequisite tasks required for subsequent data processing. The inherent urgency coefficient for the k-th type of data; This is a calibration factor for urgency. The above formula calculates the proportion of processing time remaining and the completion rate of the preceding dependent tasks. Combined with the inherent urgency coefficient of the data, the three are assigned corresponding weights and summed, and then multiplied by the urgency calibration coefficient. The proportion of time remaining reflects the urgency of the remaining time of the task, the completion rate of the dependent tasks reflects the maturity of the processing start conditions, the inherent urgency reflects the priority of the data itself, and the calibration coefficient adapts to the computational complexity of different processing tasks, so as to achieve accurate quantification of the urgency of subsequent processing. ; In the formula: This represents the storage capacity of the current data block. Weights for data storage density; This represents the total number of data blocks to be processed in the current system. Let m be the storage capacity of the m-th data block to be processed; This is a storage capacity adjustment item; The amplitude is affected by storage pressure; This refers to the storage pressure response coefficient. This represents the current actual occupancy rate of the storage medium. The preset safe occupancy threshold for the storage medium; The above formula combines the storage capacity of the current data block and the total scale of the data to be processed in the system, introduces the data storage density weight, fine-tunes the score through the storage capacity correction item, and associates the actual occupancy rate of the storage medium with the preset safe occupancy rate threshold. By using the amplitude of the storage pressure influence and the storage pressure response coefficient, it comprehensively quantifies the impact of the data size ratio on storage scheduling, so that the score can not only reflect the storage scale ratio of a single data block, but also dynamically adapt to the system storage resource status, providing support for data read and write sorting. , , After calculation using the above formula and normalization calibration, , , The values ​​are all within the range of 0 to 1. The calibration process involves mapping each score value to a preset standard score interval. The mapping relationship is dynamically optimized based on the score distribution characteristics of the system's historical processed data. in, ∈ (0,1), the higher the sensitivity of the data to real-time, the larger the value; the lower the sensitivity to real-time, the smaller the value. ∈ (0,1), the greater the influence of the sensor's real-time performance level on the overall real-time performance of the data, the larger the value; the smaller the influence, the smaller the value. ∈[0.5,5], the higher the system's sensitivity to real-time requirements, the more demanding it is. The more sensitive the change, the larger the value; the lower the sensitivity, the more... The more gradual the change, the smaller the value. ∈ (0,1), the greater the impact of the processing time margin on the timely completion of subsequent tasks, the larger the value; the smaller the impact, the smaller the value. ∈ (0,1), the greater the constraint of the completion status of the preceding dependent tasks on the initiation of subsequent processing, the larger the value; the smaller the constraint, the smaller the value. ∈ (0,1), the higher the weight of the inherent urgency of the data on the priority of subsequent processing, the larger the value; the lower the weight of the influence, the smaller the value. As preset; ∈[0.8, 1.2], the higher the computational complexity of the task and the greater the score deviation that needs to be corrected, the further the value deviates from 1.0; the lower the computational complexity and the smaller the score deviation that needs to be corrected, the closer the value is to 1.0. ∈[0.5, 2], the lower the data compression ratio, the larger the value; the higher the compression ratio, the smaller the value. The default value range is 0-10. -6 The smaller the total capacity of the data to be processed in the current system, the larger the value; the larger the total capacity, the smaller the value, even approaching 0. ∈[0,0.5], the more strained the system's storage resources are and the greater the impact of storage pressure on the score needs to be, the larger the value will be; the more abundant the storage resources are, the smaller the value will be. ∈[1,10], the more sensitive the change in storage occupancy rate is to the rating, the larger the value should be; the lower the sensitivity requirement for this effect, the smaller the value should be. The transmission module is used to dynamically adapt the transmission rate and encoding method based on data characteristics, and integrates a signal enhancement unit to transmit the data output by the storage module to the application terminal connected to the system in real time. The dynamic adaptation of the transmission rate of the transmission module conforms to: Under the premise of meeting the data transmission delay requirements, the transmission rate is dynamically adjusted based on the current amount of data to be transmitted, the available bandwidth of the channel, and the channel bit error rate to maximize the data transmission efficiency, and the adjustment range of the transmission rate does not exceed the preset range of the maximum supported transmission rate and the minimum stable transmission rate of the channel. Encoding methods include low-density parity-check codes, convolutional codes, and adaptive variable-length codes. The encoding method is dynamically selected based on the structured attributes of the data and a preset bit error rate threshold. For structured data, low-density parity-check codes should be used preferentially. For unstructured data, adaptive variable-length coding should be used preferentially; When the data is semi-structured, or the channel bit error rate is within a preset intermediate range, and data transmission needs to balance accuracy and efficiency, convolutional codes are preferred. Furthermore, the key parameters of the convolutional codes are dynamically optimized based on real-time channel conditions and data characteristics. These dynamically optimized key parameters include: Convolutional code bit rate ,in This is the fault tolerance adjustment coefficient, which is positively correlated with the data's fault tolerance level. The fault tolerance level is set based on the error tolerance of the data application scenario. For bitrate calibration factor, >0, used to balance encoding efficiency and decoding complexity; The above formula takes the fault tolerance level of the data as the core, adjusts the denominator of the base code rate through the fault tolerance adjustment coefficient which is positively correlated with the fault tolerance level, and then uses the code rate calibration factor to balance the encoding efficiency and decoding complexity, so that the code rate of the convolutional code can dynamically adapt to the error tolerance of different data application scenarios, and optimize the resource consumption of encoding and decoding while ensuring the accuracy of data transmission. constraint length ,in Based on the length of the basic constraint, For the constraint length increment coefficient, ∈[1,5], the value increases when the data importance score and channel bit error rate are higher, and decreases when the data importance score and channel bit error rate are lower. This indicates the floor function. Data importance scoring, data importance scoring The impact weight of the data on the subsequent application terminal processing results is dynamically calculated and normalized to a preset range before application. The above formula takes the basic constraint length as the benchmark and introduces an incremental coefficient of constraint length related to data importance score and channel bit error rate. The product of data importance score and incremental coefficient is rounded down and then superimposed on the basic constraint length, so that the constraint length can be dynamically adjusted with data importance and channel bit error rate. The higher the data importance and the higher the channel bit error rate, the larger the constraint length, which improves the reliability of data transmission. Conversely, a moderate constraint length is maintained to balance transmission efficiency. The signal enhancement unit adopts a combination of adaptive noise suppression and dynamic gain adjustment. It identifies the noise frequency band by extracting the spectral characteristics of the transmitted signal, performs targeted noise suppression processing by adaptive notch filtering, and adjusts the gain coefficient in real time based on the signal strength so that the signal-to-noise ratio of the transmitted signal always meets the reception requirements of the application terminal. The adjustment range of the gain coefficient is inversely proportional to the signal strength and does not exceed the preset gain safety range. The collaborative control module is used to acquire data processing progress and resource occupancy status signals in the system, generate and issue synchronization control commands, and regulate the runtime sequence and resource allocation of each superior module. The resource occupancy status signals monitored by the collaborative control module include the computing power utilization rate of the computing unit, the storage space occupancy rate of the storage medium, the bandwidth occupancy rate of the transmission link, and the processing latency of each module. The generation of synchronization control instructions is based on the difference in processing progress and resource utilization between modules: ; In the formula: To address the schedule discrepancy; This represents the processing progress of the i-th module; The processing progress of the j-th module associated with the i-th module; This is due to deviation in resource utilization. This refers to the actual utilization rate of resources. The preset security threshold for resources; The calculation formula directly reflects the processing timing synchronization status between modules by calculating the processing progress difference between the i-th module and the associated j-th module. This provides a core basis for determining whether it is necessary to adjust the module's running clock frequency, data input rate, or computing power allocation ratio, ensuring that the processing progress between modules matches and avoiding timing errors. The calculation formula quantifies the degree to which resource occupancy deviates from a reasonable range by comparing the actual resource occupancy rate with a preset safety threshold. This provides a key reference for the collaborative control module to generate synchronization control commands, enabling the module to respond promptly to abnormal resource occupancy and ensuring the stable utilization of system resources. when Exceeding the preset schedule deviation range or When the deviation exceeds the preset occupancy rate range, the collaborative control module generates a synchronization control command by adjusting the running clock frequency, data input rate, or computing power allocation ratio of the corresponding module. The command is issued in a time-division multiplexing manner to ensure the real-time and accuracy of the command received by each module. The execution effect of the command is verified in real time by the status signal fed back by the module. The command parameters are dynamically optimized based on the verification results to achieve precise synchronization of the running sequence between modules. The upper-level modules of the collaborative control module include an adaptation module, a preprocessing module, a scheduling module, a storage control module, and a transmission module. The adaptation module is interconnected with the preprocessing module via a local area network. The preprocessing module is interconnected with the scheduling module via a local area network. The scheduling module is interconnected with the control module via a local area network. The control module is interconnected with the transmission module via a local area network. The transmission module is interconnected with the signal enhancement unit via a local area network. The transmission module is interconnected with the collaborative control module via a local area network.

[0022] A storage medium having a built-in processor and storing a computer program thereon, wherein when the computer program is executed by the processor, it implements the running program of an embedded FPGA system based on multi-optical sensing technology.

[0023] In this embodiment, the adaptation module receives heterogeneous data from multiple types of optical sensors, performs preliminary encapsulation and identification of data frames according to a preset format, and the preprocessing module extracts feature parameters from the received data, removes redundant information through dynamic threshold filtering, and outputs the filtered valid data. The scheduling module then dynamically allocates computing resources of the FPGA's internal computing units based on the type and scale of the preprocessed data, generating an optimal computing power allocation strategy. The storage control module further receives the optimal computing power allocation strategy generation result from the scheduling module, adjusts the read / write timing of the storage medium based on the result, establishes a data priority queue and applies it to data read / write sorting, and dynamically adapts the transmission rate and encoding method based on data characteristics through the transmission module. The integrated signal enhancement unit transmits the data output by the storage module to the external application terminal in real time. Finally, the collaborative control module obtains the data processing progress and resource occupancy status signals in the system, generates and issues synchronization control commands to regulate the running timing and resource allocation of each upper-level module.

[0024] In the above embodiments, the system can efficiently integrate multiple types of optical sensing data, accurately filter effective information, eliminate redundancy, rationally allocate computing resources to improve processing efficiency and reduce energy consumption, intelligently adapt to storage and transmission needs, ensure the real-time performance, accuracy and stability of data processing, and enable application terminals to quickly obtain reliable data.

[0025] Referring to the system in the above embodiments, the following is an application example of this system in a real-time monitoring scenario of laser welding quality: This system can serve as the data processing and control core of a real-time laser welding quality monitoring system. Based on the technical requirements of laser welding monitoring, the system connects to heterogeneous optical data output from various types of photoelectric sensors via an adapter module. In this application example, the data types connected to the system and their corresponding relationships are as follows: Light intensity sensing data: Laser reflected light characteristic data: The laser intensity signal reflected from the welding area is collected by a photodetector, reflecting the laser power fluctuation and the reflection characteristics of the weldment surface; this data is a single wavelength or narrowband light intensity numerical sequence. Spectral sensing data: Metal vapor / plasma radiation characteristic data: The radiation spectrum of plasma generated by the ionization of metal vapor during welding is collected by a spectrometer in a wide range of wavelengths, reflecting the plasma temperature, density and metal vapor ionization state; this data is the spectral intensity distribution of multiple wavelength channels; Molten pool thermal radiation spectral data: The thermal radiation spectral distribution of the molten pool in a specific wavelength range is collected by a spectrometer, reflecting the temperature field distribution and morphological evolution characteristics of the molten pool. This application example primarily utilizes a combination of the aforementioned light intensity sensing data and spectral sensing data. It should be noted that in higher-precision welding quality monitoring scenarios, the system can also be extended to access phase sensing data (obtaining phase information of the weld surface morphology through interferometry) or polarization state sensing data (detecting stress distribution and material anisotropy in the welding area through polarization analysis), which will not be described in detail in this embodiment. During system operation, the adaptation module encapsulates data frames according to the preset format described in the above embodiments, assigns a unique identifier to each sensor, and adds a timestamp to ensure the time synchronization of multi-sensor data. Specifically, light intensity sensing data and spectral sensing data differ in their data frame structures: light intensity data is a single-channel scalar sequence, with each frame containing a single intensity value and a timestamp; spectral data is a multi-channel vector sequence, with each frame containing a matrix of intensity values ​​for multiple wavelength channels. The adaptation module automatically matches the field length based on the data type, completing the standardized encapsulation of heterogeneous data. The preprocessing module uses a dynamic threshold filtering algorithm to extract welding defect feature parameters: extracting features such as abnormal fluctuations in laser power and sudden changes in reflectivity from light intensity sensing data; extracting features such as abnormal plasma radiation intensity, missing feature spectral lines, and abnormal peak values ​​of molten pool temperature from spectral sensing data; and, based on the dynamic threshold calculation formula in the above embodiment, combined with historical statistical features, real-time fluctuation coefficients, and environmental interference factors (such as welding spatter and dust interference), eliminating redundant information and accurately locating defect features. The scheduling module dynamically allocates computing resources of the FPGA internal computing units according to the type and scale of the defect feature data: the processing of light intensity sensing data is preferentially allocated to the ALU array for fast numerical calculation; the processing of spectral sensing data is preferentially allocated to the DSP core for multi-channel parallel spectrum analysis; according to the computing power allocation strategy generation formula in the above embodiment, the optimal resource scheduling is achieved by comprehensively considering data processing priority, transmission scale, computational complexity and latency requirements. The storage control module, based on the comprehensive priority scoring mechanism in the above embodiments, stores high-priority defect feature data (such as the identification results of key defects such as cold solder joints, gaps, and holes) in static random access memory for fast access; stores large-capacity raw spectral data in flash memory for offline in-depth analysis and quality traceability; and, due to its small data volume, light intensity sensing data can be flexibly allocated to dynamic random access memory or flash memory according to the access frequency.

[0026] The transmission module transmits the processed welding quality analysis results (including defect type, location coordinates, severity level, time series, etc.) to the quality monitoring terminal through dynamically adapted transmission rates and encoding methods. The signal enhancement unit ensures the transmission stability of data in the electromagnetic interference environment of the welding site through adaptive noise suppression and dynamic gain adjustment. The collaborative control module monitors the processing progress and resource usage of each module in real time: to address the processing time differences between light intensity data and spectral data (spectral data takes longer to process due to the large number of channels), it precisely regulates the running sequence of each stage through synchronous control commands; ensuring the time consistency of multi-sensor data and avoiding defect location errors caused by timestamp misalignment between light intensity data and spectral data. In summary, the system described in the above embodiments can serve as a low-level data processing and computing power scheduling platform in laser welding quality monitoring applications. Through FPGA hardware architecture, it achieves efficient fusion of heterogeneous data from multiple optical sensors, real-time defect feature extraction, intelligent storage management, and stable data transmission, providing a complete embedded solution for real-time monitoring of laser welding quality.

[0027] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An embedded FPGA system based on multi-optical sensing technology, characterized in that, include: The adapter module is used to receive heterogeneous data output from various types of optical sensors of the target and to complete the initial encapsulation and identification of the data frames according to the preset format. The preprocessing module is used to extract feature parameters from the access data, filter out redundant information through dynamic thresholds, and output the filtered valid data. The scheduling module is used to dynamically allocate computing resources of the FPGA internal computing units according to the type and scale of the preprocessed data, and generate the optimal computing power allocation strategy. The storage control module is used to receive the optimal computing power allocation strategy generated by the scheduling module, adjust the read and write timing of the storage medium based on the results, establish a data priority queue and apply it to the data read and write sorting. The transmission module is used to dynamically adapt the transmission rate and encoding method based on data characteristics, and integrates a signal enhancement unit to transmit the data output by the storage module to the application terminal connected to the system in real time. The collaborative control module is used to acquire data processing progress and resource occupancy status signals in the system, generate and issue synchronization control commands, and regulate the runtime sequence and resource allocation of each superior module. The upper-level modules in the collaborative control module include an adaptation module, a preprocessing module, a scheduling module, a storage control module, and a transmission module.

2. The embedded FPGA system based on multi-optical sensing technology according to claim 1, characterized in that, The heterogeneous data accessed by the adapter module consists of optical feature data output by different types of optical sensors; The encapsulation process of the preset format is as follows: Based on the type of heterogeneous data, the field length and data structure are automatically matched. The fields include a data header, a data body and a check bit. The data header integrates the sensor ID, data acquisition timestamp, data type identifier and data integrity check identifier. The identification logic generates a unique identifier code based on the sensor type encoding and data acquisition sequence. The encoding length of the identifier code dynamically adapts to the number of connected sensors, and the mapping relationship between the identifier code and the data frame is updated in real time through a preset dynamic mapping table. The update frequency of the mapping table is synchronized with the data acquisition frequency.

3. The embedded FPGA system based on multi-optical sensing technology according to claim 1, characterized in that, The feature parameters extracted by the preprocessing module include data amplitude feature parameters, frequency feature parameters, spatiotemporal correlation coefficients between data, and real-time data fluctuation coefficients. The formula for calculating the dynamic threshold is: ； In the formula: For dynamic filtering thresholds; Statistical feature weighting factors; The average amplitude of historical data within a preset time window; This is the variance correction factor; The standard deviation of the amplitude of historical data within the preset time window; Real-time fluctuation influencing factors; It is a correlation correction factor; This is the amplitude fluctuation coefficient between the current data block and the previous data block; This is the average spatiotemporal correlation coefficient between the current data block and historical data blocks of the same type; This is a correction factor for environmental interference. The redundant information is that the feature parameters exceed the threshold. The corresponding numerical range or satisfying Duplicate data blocks exceeding a preset threshold are filtered to select valid data whose feature parameters are within the threshold. Within the corresponding numerical range and Data blocks that do not exceed the preset relevant threshold.

4. The embedded FPGA system based on multi-optical sensing technology according to claim 1, characterized in that, The optimal computing power allocation strategy in the scheduling module is generated as follows: Calculate the quantized value of the computing power requirement per unit time for the k-th type of data. ; based on Calculate the computing power allocation ratio of the i-th processing unit. ; In the formula: The processing priority weight for the k-th type of data; Let k be the transmission volume per unit time for the k-th type of data; This represents the computational complexity coefficient corresponding to the k-th data type. The allowable processing delay for the k-th type of data; For data format adaptation coefficients; The adaptation coefficient for the processing of the k-th type of data by the i-th processing unit; This represents the percentage of idle computing power in the i-th computing unit in real time. is the energy efficiency coefficient of the i-th processing unit; This represents the total number of computing units involved in scheduling within the FPGA. This is a resource redundancy adjustment factor. This represents the system's current total remaining computing power. The scheduling module according to The computing resources of each computing unit are allocated to the corresponding data processing tasks, and the system is recalculated in real time when new data types are added, the status of computing units changes, or the computing power demand fluctuates beyond the preset range. and To dynamically update the computing power allocation strategy.

5. The embedded FPGA system based on multi-optical sensing technology according to claim 1, characterized in that, The storage medium in the storage control module includes static random access memory, dynamic random access memory, and flash memory. During the operation of the storage control module, the storage medium is dynamically selected based on the data read / write frequency and allowed access latency: data with high read / write frequency and low latency requirements are preferentially stored in static random access memory, data with medium read / write frequency and latency requirements are stored in dynamic random access memory, and data with low read / write frequency and large capacity requirements are stored in flash memory. The data priority queue is sorted based on comprehensive priority. ,in Weighting for data real-time requirements, For subsequent data processing, urgency weights, The weights are determined by the proportion of data size, and , , The sum is 1. Scoring based on data real-time requirements, Assess the urgency of subsequent data processing. Score the percentage of data size; The storage control module adapts the optimal computing power allocation strategy by adjusting the read / write clock frequency and data bus width of the storage medium. When multiple data initiate read / write requests simultaneously, the read / write operations are executed in descending order of comprehensive priority P. The P value of each data is updated in real time at preset time windows to dynamically adjust the read / write order.

6. The embedded FPGA system based on multi-optical sensing technology according to claim 5, characterized in that, The , , The calculation logic is as follows: ； In the formula: The effective duration of data storage; The duration the data has been stored; This is the quantized value of the sensor real-time performance level corresponding to the kth type of data; This represents the maximum quantization value for the real-time performance level of the sensors supported by the system. , Weighting factors include effective duration of operation as a percentage of weight and sensor real-time performance level as a percentage of weight. This is the offset of the real-time scoring threshold; This is the scoring adjustment factor; ； In the formula: This is the deadline for subsequent data processing tasks; The current system time; For the time the data was generated; , , Weights are assigned based on the proportion of processing time remaining, the completion rate of dependent tasks, and the inherent urgency level. The number of tasks that have been completed in the prerequisite tasks required for subsequent data processing; This represents the total number of prerequisite tasks required for subsequent data processing. The inherent urgency coefficient for the k-th type of data; This is a calibration factor for urgency. ； In the formula: This represents the storage capacity of the current data block. Weights for data storage density; This represents the total number of data blocks to be processed in the current system. Let m be the storage capacity of the m-th data block to be processed; This is a storage capacity adjustment item; The amplitude is affected by storage pressure; This refers to the storage pressure response coefficient. This represents the current actual occupancy rate of the storage medium. The preset safe occupancy threshold for the storage medium; The , , After calculation using the above formula and normalization calibration, , , The values ​​are all in the range of 0 to 1.

7. The embedded FPGA system based on multi-optical sensing technology according to claim 1, characterized in that, The dynamic adaptation of the transmission rate of the transmission module conforms to: Under the premise of meeting the data transmission delay requirements, the transmission rate is dynamically adjusted based on the current amount of data to be transmitted, the available bandwidth of the channel, and the channel bit error rate to maximize the data transmission efficiency, and the adjustment range of the transmission rate does not exceed the preset range of the maximum supported transmission rate and the minimum stable transmission rate of the channel. The encoding methods include low-density parity-check codes, convolutional codes, and adaptive variable-length codes. The encoding method is dynamically selected based on the structured attributes of the data and a preset bit error rate threshold. For structured data, low-density parity-check codes should be used preferentially. For unstructured data, adaptive variable-length coding should be used preferentially; When the data is semi-structured, or the channel bit error rate is within a preset intermediate range, and data transmission needs to balance accuracy and efficiency, convolutional codes are preferred. Furthermore, the key parameters of the convolutional codes are dynamically optimized based on real-time channel conditions and data characteristics. These dynamically optimized key parameters include: Convolutional code bit rate ,in This is the fault tolerance adjustment coefficient. For bitrate calibration factor, >0, used to balance encoding efficiency and decoding complexity; constraint length ,in Based on the length of the basic constraint, For the constraint length increment coefficient, This indicates the floor function. Score the importance of the data.

8. The embedded FPGA system based on multi-optical sensing technology according to claim 1, characterized in that, The resource occupancy status signals monitored by the collaborative control module include the computing power utilization rate of the computing unit, the storage space occupancy rate of the storage medium, the bandwidth occupancy rate of the transmission link, and the processing latency of each module. The generation of the synchronization control command is based on the processing progress difference and resource utilization deviation between modules: ； In the formula: To address the schedule discrepancy; This represents the processing progress of the i-th module; The processing progress of the j-th module associated with the i-th module; This is due to deviation in resource utilization. This refers to the actual utilization rate of resources. The preset security threshold for resources; when Exceeding the preset schedule deviation range or When the deviation exceeds the preset occupancy rate range, the collaborative control module generates synchronization control commands by adjusting the operating clock frequency, data input rate, or computing power allocation ratio of the corresponding module.

9. The embedded FPGA system based on multi-optical sensing technology according to claim 1, characterized in that, The adaptation module is interactively connected to the preprocessing module via a local area network. The preprocessing module is interactively connected to the scheduling module via a local area network. The scheduling module is interactively connected to the control module via a local area network. The control module is interactively connected to the transmission module via a local area network. The transmission module is interactively connected to the signal enhancement unit via a local area network. The transmission module is interactively connected to the collaborative control module via a local area network.

10. A storage medium, characterized in that, The storage medium has a built-in processor and stores a computer program on it. When the computer program is executed by the processor, it implements the running program of the embedded FPGA system based on multi-optical sensing technology as described in any one of claims 1 to 9.

Images