A multi-angle hyperspectral data multi-band cooperative collection method for agricultural products

Abstract

The application discloses a kind of for agricultural product multi-angle hyperspectral data multiband synergic acquisition method, it is related to agricultural product nondestructive testing technical field, including: sample is in position after cross-modal feature perception, obtains quality and outer contour data and solves push scan angle;Based on closed loop gripping mechanism, utilize multi-parameter joint constraint model to solve scanning row number, frame frequency, exposure time and push scan speed synchronously, and parameter static latching is to hardware bottom register;Under the control of logic interlock, automatically circulates and completes multi-side joint push scan acquisition;Finally, space self-registration and space-time label binding are carried out by asynchronous data calculation node, output geometric alignment hyperspectral data cube, the present application effectively solves the space-time dislocation and geometric distortion problem when multiband camera shares movement mechanism, significantly improves the collection efficiency and geometric fidelity of high-throughput, multi-modal agricultural product hyperspectral data, provides high-dimensional digital model construction foundation for agricultural product quality detection.

Description

Technical Field

[0001] This invention relates to the field of non-destructive testing technology for agricultural products, specifically a method for multi-angle hyperspectral data acquisition of agricultural products using multiple bands. Background Technology

[0002] Hyperspectral imaging technology integrates image and spectral information, enabling the simultaneous acquisition of the external morphological features of an object and the corresponding spectral responses of its internal components. It holds broad application prospects in fields such as agricultural product quality analysis, growth monitoring, and variety identification. For agricultural products, by collecting hyperspectral data from their entire surface, the distribution of their physicochemical properties can be comprehensively inverted, providing a data foundation for subsequent quality model construction.

[0003] Hyperspectral cameras can be classified into visible and near-infrared (400~1000nm) and short-wave infrared (900~1700nm) types according to their wavelength range. Different wavelength cameras have different response characteristics to the internal components of agricultural products. By using multiple hyperspectral cameras in combination, the spectral fingerprint information of agricultural products can be obtained more comprehensively, providing a rich data source for multimodal data analysis.

[0004] Furthermore, in agricultural product modeling or testing, agricultural products or fruits vary in shape across different areas, and defects (such as bruises and insect holes) or internal components (such as sugar content and moisture) are often spatially unevenly distributed. Relying solely on hyperspectral information from a single angle is insufficient to cover the overall information of the agricultural product. Therefore, utilizing hyperspectral cameras with multiple spectral bands to acquire hyperspectral images from multiple angles (such as acquiring images from multiple angles around the agricultural product) can construct richer and more complete spectral fingerprints and spatial information, providing comprehensive hyperspectral data for applications such as crop growth and fruit quality prediction.

[0005] However, the following technical bottlenecks exist in practical applications:

[0006] 1. Low efficiency of multi-angle imaging: Traditional hyperspectral imaging often uses the "push-broom" method, in which the camera and the sliding rail move to scan the space line by line. This method can often only acquire data from a single angle of the target. Multi-angle acquisition requires flipping the target, which is difficult for machine operation. It usually relies on manual placement of samples, manual triggering of acquisition, and manual flipping of samples to change the shooting angle. This process is extremely time-consuming and labor-intensive, which seriously restricts the efficiency of acquiring multi-angle spectral data and makes it difficult to build large-scale datasets.

[0007] 2. Difficulties in Multi-Camera Collaboration: If two cameras with different wavelengths (such as visible light near-infrared and short-wave infrared) are needed to photograph the same agricultural product from multiple angles, multiple independent imaging systems are usually required for separate control. This not only increases the hardware cost of the equipment but also easily leads to asynchronous data from multiple perspectives and wavelengths in the temporal dimension, significantly increasing the difficulty of image registration and multimodal fusion in the later stages. In addition, when multiple cameras are directly connected in parallel and share the same rotating platform, the optimal imaging parameters of each camera differ significantly. Exposure time, frame rate, and turntable angular velocity are mutually restrictive and difficult to match, making it impossible to achieve synchronous acquisition under differentiated parameter conditions. This results in temporal misalignment and spatial mismatch of images, further increasing the complexity of subsequent image processing and data fusion.

[0008] 3. Limited Data Dimensions: Existing acquisition devices typically only collect spectral images, neglecting key physical indicators such as weight. While incorporating weight information can enrich the data dimensions in multimodal data analysis, the lack of automated methods for correlating weight acquisition with spectral data makes it difficult to collaboratively acquire information from multiple sources, thus failing to fully leverage the advantages of multimodal learning.

[0009] Therefore, there is an urgent need for an automated acquisition device that can achieve multi-camera collaboration, automatic sample rotation, and automatic weight collection, enabling multi-angle and multi-band hyperspectral data acquisition. This device would efficiently acquire multi-angle and multi-band spectral data and physical indicators of agricultural products, providing a high-quality data foundation for subsequent quality testing research based on multimodal data analysis. Summary of the Invention

[0010] The purpose of this invention is to provide a method for multi-band collaborative acquisition of multi-angle hyperspectral data of agricultural products. By using multi-dimensional sensing, heterogeneous computing power separation, and a static latching collaborative mechanism for core parameters, it breaks through the parameter coupling bottleneck when different band hyperspectral cameras share a turntable, and realizes high-throughput, distortion-free, fully automatic multimodal data acquisition and shallow 3D quality joint modeling, thereby solving the problems mentioned in the background art.

[0011] To achieve the above objectives, the present invention provides the following technical solution: a method for multi-band collaborative acquisition of multi-angle hyperspectral data of agricultural products. This method relies on an automated device comprising at least two hyperspectral imaging units sharing the same pushbroom drive unit, a pose adjustment unit for carrying the sample to be measured, a quality sensing unit, a morphology sensing unit, and a collaborative control system. Specifically, it includes the following steps:

[0012] S1 Sample Placement and Cross-Modal Feature Sensing: After the sample to be tested is placed in the pose adjustment unit, the collaborative control system receives the initial mass data of the sample from the mass sensing unit and simultaneously triggers the morphology sensing unit to scan the physical outer contour of the sample; the collaborative control system calculates the push-broom relative motion angle or displacement required to cover the sample based on the received contour data. Complete data initialization;

[0013] S2 Multi-parameter Joint Cooperative Configuration Based on Closed-Loop Handshake: The cooperative control system will calculate the... By importing the internally preset multi-parameter joint constraint model and combining the different band responses and field-of-view attributes of each hyperspectral imaging unit, the number of scan lines that satisfy both spatial distortion-free and temporal strong synchronization is jointly calculated. Frame rate Exposure time and the uniform relative motion velocity of the push-broom drive unit Subsequently, the collaborative control system issues the above parameter instructions and writes them into the underlying registers of each corresponding hardware unit for static latching, forming a strong mapping relationship between software and hardware parameters.

[0014] S3 Automatic Cyclic Synchronous Acquisition Based on Logic Interlocking: In the static latching state of parameters, the cooperative control system receives the "pose lock ready" signal sent by the pose adjustment unit, using this as a trigger condition to synchronously send acquisition enable signals to the pushbroom drive unit and each hyperspectral imaging unit, driving them to... Complete the joint push-broom of the current side; after the current side acquisition is completed, the collaborative control system blocks the acquisition enable signal and instead sends a drive command to the pose adjustment unit to change the preset angle until the next "pose lock ready" signal is received. This step is repeated until all side acquisitions are completed.

[0015] S4 Asynchronous Alignment and Multidimensional Spatiotemporal Association: During the acquisition cycle, the collaborative control system binds the hyperspectral image data stream acquired from each side with the corresponding angular spatial coordinates and the initial quality data acquired in step S1, according to the sample identity identifier, and pushes it to the asynchronous data computing node. The node then performs spatial self-registration of the multi-band hyperspectral image in a background operation mode.

[0016] Preferably, in step S2, the multi-parameter joint constraint model receives data from the morphology sensing unit to construct interactive constraints that minimize spatial distortion; to ensure that the image obtained by push-broom is consistent with the actual object in spatial geometric proportions, the collaborative control system adjusts the field of view of each hyperspectral imaging unit i according to the field of view angle of each hyperspectral imaging unit i. and horizontal pixel count Force a limit on the number of rows scanned. The formula must be satisfied:

[0017] .

[0018] Preferably, the multi-parameter joint constraint model in step S2 constructs time synchronization interaction constraints for multiple imaging units; the collaborative control system is controlled via a hardware bus to force each hyperspectral imaging unit to acquire data. The total time required for each frame must be equal and consistent with the execution time T of the controlled motion of the push-broom drive unit, satisfying the formula: And the unified relative motion speed command value The controlled system is set to .

[0019] Preferably, the cooperative control system achieves the minimum total time for jointly solving the system. At that time, the following logical judgment and selection mechanism shall be established:

[0020] Under the basic hardware constraints of hyperspectral imaging unit i Under this premise, the maximum value of the constraint boundary of all hyperspectral imaging units is extracted as the minimum instruction cycle of the control bus:

[0021] ;

[0022] In the formula, The optimal signal-to-noise ratio exposure time is dynamically locked by the collaborative control system based on the current illumination feedback. This represents the maximum frame rate limit provided by the hardware feedback of the hyperspectral imaging unit. The system's preset instruction execution redundancy is a small positive number.

[0023] Preferably, the optimal signal-to-noise ratio exposure time The dynamic locking mechanism is as follows: In step S1, the collaborative control system calls the morphology sensing unit to perform local micro-motion pre-scan. Based on the feedback of the acquired image histogram, the collaborative control system automatically masks and removes interference areas. Based on the peak value of the extracted high reflectivity uniform area histogram, the dynamic optimization algorithm is used to adjust and finally determine the locking mechanism. This value is then input into the multi-parameter joint constraint model.

[0024] Preferably, the static latching state of the parameters in step S3 embodies a disturbance rejection feedback mechanism; within the multi-faceted cyclic acquisition cycle of the same sample, the cooperative control system actively blocks dynamic speed adjustment interruption requests for the pushbroom drive unit, forcing the motor driver to act according to the latched parameters. Maintaining a constant current output cuts off the transmission path of image micro-geometric distortion caused by nonlinear fluctuations in mechanical acceleration and deceleration.

[0025] Preferably, the collaborative control system is configured as a dual-node heterogeneous interactive architecture, including a real-time motion control node and an asynchronous data computing node; steps S2 and S3 are performed by the real-time motion control node directly scheduling the hardware unit through an industrial real-time bus; when an interrupt signal generated by each side in step S3 is triggered, the real-time motion control node maps the raw data with timestamps across nodes to the asynchronous data computing node through a direct memory access mechanism to execute step S4.

[0026] Preferably, the spatial self-registration performed by the asynchronous data computing node is controlled by the spatial reference command issued by the real-time motion control node; the asynchronous data computing node automatically extracts the unit with the smallest field of view value among all hyperspectral imaging units as the spatial anchoring reference, and performs hardware-level clipping coordinate mapping and resampling calculation on other data sources accordingly, thereby outputting a full-band geometrically aligned hyperspectral data cube.

[0027] Preferably, the logical interlock between the pose adjustment unit and the push-broom drive unit in step S3 is specifically manifested as follows: after the cooperative control system issues the pose change command, it continuously polls the encoder feedback value of the pose adjustment unit. Only when the encoder difference is detected to be zero and the dwell time exceeds the system tuning threshold will the enable flag bit that triggers the movement of the push-broom drive unit be activated to prevent field jitter.

[0028] Preferably, after step S4, the collaborative control system retrieves the associated packaged multidimensional spatiotemporal data to perform a fusion analysis step: using multi-band, multi-angle spectral data as the first input vector to invert the spatial distribution of the physicochemical characteristics of the sample, using the quality data fed back by the quality sensing unit as the second input vector to jointly reconstruct the overall density attribute of the sample, and outputting a multidimensional heterogeneous fusion digital model for quality evaluation through cross-validation.

[0029] In summary, the beneficial effects of this invention are:

[0030] This invention establishes for the first time a multi-parameter joint constraint model based on minimizing spatial distortion and interlocking time synchronization. This model transforms isolated parameters such as exposure time, frame rate, and push-broom speed into a strongly correlated cooperative system, thus completely solving the problem of temporal misalignment and spatial mismatch when multiple band cameras share the same motion mechanism from the underlying logic.

[0031] To address the potential equipment vibration caused by dynamic parameter adjustment, this invention innovatively introduces a morphology-aware pre-detection mechanism and a static parameter latching mechanism. Within a single acquisition cycle, the motor speed parameters are forcibly latched, thus physically cutting off the transmission path of microscopic wave-like deformation in the image caused by frequent acceleration and deceleration, ensuring extremely high image geometric fidelity.

[0032] Abandoning the traditional centralized control architecture, it adopts a dual-node heterogeneous architecture of "real-time motion control node" and "asynchronous data computing node". The front-end node is only responsible for high-speed triggering and latching control at the microsecond level, while the back-end node can easily handle the registration and fusion of massive data cubes in the background, truly realizing a non-downtime high-throughput operation process of "place and collect, continue collecting when changing samples".

[0033] Multidimensional heterogeneous data resonance to build high-value digital twins: Breaking through the detection limitations of single spectrum or single weight, by establishing a low-level spatiotemporal correlation between multi-angle and multi-band spectral data and absolute physical quality data, it can accurately invert the three-dimensional gradient of the physicochemical components of the shallow surface of agricultural products, and jointly deduce the internal overall density defects, providing a highly dimensional multimodal digital model foundation for modern agricultural sorting and quality assessment. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of the invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is a schematic diagram of the overall process framework of a multi-band collaborative acquisition method for multi-angle hyperspectral data of agricultural products according to the present invention;

[0036] Figure 2 This is a schematic diagram of a data acquisition device for a multi-band collaborative acquisition method of multi-angle hyperspectral data of agricultural products according to the present invention;

[0037] Figure 3 This is a schematic diagram of the collaborative control connection for a multi-band collaborative acquisition method for multi-angle hyperspectral data of agricultural products according to the present invention;

[0038] Figure 4 This is a schematic diagram of the camera settings in the shooting interface of a multi-angle hyperspectral data multi-band collaborative acquisition method for agricultural products according to the present invention;

[0039] Figure 5 This is a log illustration of the shooting interface in a multi-band collaborative acquisition method for multi-angle hyperspectral data of agricultural products according to the present invention.

[0040] Figure 6 This is a hyperspectral image of the first hyperspectral camera used in the present invention for a multi-angle hyperspectral data multi-band collaborative acquisition method for agricultural products;

[0041] Figure 7This is a hyperspectral image showing the hyperspectral data from the second hyperspectral camera used in the multi-angle hyperspectral data multi-band collaborative acquisition method for agricultural products according to the present invention.

[0042] Figure 8 This is a schematic diagram comparing the combined parameter tuning and untuned parameter tuning of a multi-band collaborative acquisition method for multi-angle hyperspectral data of agricultural products according to the present invention. Detailed Implementation

[0043] The present invention will now be described in further detail with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. 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. These drawings are simplified schematic diagrams, which are only used to illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0044] To facilitate understanding of the present invention, a more complete description of the invention will be given below with reference to the accompanying drawings, which illustrate several embodiments of the invention. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that the disclosure of the invention will be more thorough and complete.

[0045] All features disclosed in this specification, or all steps in all disclosed methods or processes, may be combined in any way, except for mutually exclusive features and / or steps.

[0046] Any feature disclosed in this specification (including any appended claims, abstract, and drawings) may be replaced by other equivalent or similar features for a similar purpose, unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is merely one example of a series of equivalent or similar features.

[0047] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection, a direct connection, or an indirect connection through an intermediate medium; they can refer to the internal communication of at least two elements or the interaction relationship of at least two elements, unless otherwise explicitly limited. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0048] Please see Figure 1 - Figure 8The present invention provides an embodiment of a method for multi-band collaborative acquisition of multi-angle hyperspectral data of agricultural products, wherein the acquisition device is configured inside a dark box to isolate external ambient light interference and is equipped with a light source such as a halogen tungsten lamp to provide full-band illumination, specifically including:

[0049] Push-broom drive unit: Specifically implemented as a precision electronically controlled rotary table or a high-precision linear guide, preferably a camera turntable. This unit is controlled by the underlying industrial bus and is responsible for driving the camera assembly to perform smooth and uniform scanning relative motion in the horizontal direction.

[0050] Hyperspectral imaging unit: Specifically implemented as at least two hyperspectral cameras with responses to different wavelengths. For example, the first hyperspectral camera is a visible-near-infrared camera with a response range of 400-1000 nm, and the second hyperspectral camera is a short-wave infrared camera with a response range of 900-1700 nm. The two cameras are vertically stacked with their optical axes parallel, and fixed on the camera turntable using a dedicated structural component.

[0051] The pose adjustment unit is specifically implemented as a sample turntable equipped with a high-precision servo motor. This turntable is positioned within the camera's imaging field of view and is used to carry and drive the agricultural products to perform intermittent fixed-angle rotations around the vertical axis, for example, rotating 90° each time.

[0052] Quality sensing unit: Specifically implemented as a high-precision electronic weighing module or force sensor integrated at the bottom of the sample turntable, which can acquire high-precision physical quality data imperceptibly the moment the sample is placed.

[0053] Morphology sensing unit: Specifically implemented as a low-cost industrial RGB camera or TOF (time-of-flight) depth sensor, set on the top or side of the dark box, used to instantly acquire three-dimensional point cloud or two-dimensional global outer contour image of agricultural products before the push-broom.

[0054] Collaborative control system: Specifically implemented as an industrial control computer system employing a dual-node heterogeneous processing architecture. It includes "real-time motion control nodes" responsible for underlying hardware I / O scheduling, such as PLCs or motion control cards equipped with RTOS, and "asynchronous data computing nodes" dedicated to massive data processing, such as workstations equipped with independent GPUs.

[0055] The specific data collection method includes the following steps:

[0056] S1 Sample Placement and Cross-Modal Feature Perception

[0057] The sample to be tested is placed on the pose adjustment unit, and the quality sensing unit automatically reads the initial quality data and sends it to the collaborative control system.

[0058] The collaborative control system synchronously triggers the shape perception unit (RGB camera / TOF sensor) to scan the outer contour of the sample. An edge extraction algorithm calculates the actual width and height of the sample, and then determines the required camera turntable push-broom rotation angle to cover the sample. For example, for a fruit of a specific size, calculate... .

[0059] The collaborative control system invokes the morphology sensing unit to perform local micro-motion pre-scanning. Based on the acquired image histogram, it automatically masks and removes interfering areas such as fruit stems, highly reflective spots, and dark lesions, locking in a region of interest (ROI) with uniform reflectivity. For this region, microsecond-level exposure trials are performed, and histogram peak values ​​are extracted to dynamically determine the optimal signal-to-noise ratio exposure time for each camera. .

[0060] S2 Multi-parameter Joint Coordination Configuration Based on Closed-Loop Handshake

[0061] To ensure that the images do not suffer from spatial geometric distortion and that absolute time synchronization is maintained when all cameras share the same turntable, the cooperative control system imports the above parameters into a multi-parameter joint constraint model for solution:

[0062] Spatial distortion minimization constraint: To ensure that the aspect ratio of the image obtained by push-broom is consistent with that of the actual object in space (without stretching or compression), the number of scan rows required for each hyperspectral imaging unit i is determined. The following formula must be satisfied:

[0063] ;

[0064] in, This represents the horizontal pixel count of the camera. This represents the camera's field of view.

[0065] Multi-imaging unit time synchronization interlock: each camera acquires data The total time required for each frame must be equal, and equal to the rotation time T of the push-broom drive unit. The formula is as follows:

[0066] ;

[0067] And uniform push-broom angular velocity satisfy .

[0068] Solving for optimal timing parameters: while satisfying the basic hardware constraints of each camera. Under the premise that the system extracts the extreme value as the bus period, the minimum total time is calculated. :

[0069] ;

[0070] In the formula Set a small redundancy positive number for the system, such as 0.01s.

[0071] Static latching mechanism: Solve for the optimal T, and select the actual... Frame rate of each camera and angular velocity Subsequently, the collaborative control system sends instructions to the underlying registers of each hardware component for static latching. During the multi-faceted acquisition cycle of the same sample, dynamic speed control interrupt requests for the motor are actively masked, forcing the motor driver to act according to the latched parameters. Maintain constant torque / current output to completely cut off the transmission path of microscopic wavy distortion in the image caused by mechanical vibration.

[0072] S3 Automatic Cyclic Synchronous Acquisition Based on Logic Interlocks

[0073] Pose lock verification: The real-time motion control node of the cooperative control system polls the encoder feedback value of the pose adjustment unit. Only when the encoder difference is zero and the dwell time exceeds the set threshold is "pose lock ready".

[0074] Triggered push-broom: The real-time motion control node synchronously sends an enable signal, and the push-broom drive unit drives the push-broom at a constant angular velocity. It drives multiple cameras to perform joint push-broom imaging of the current side.

[0075] Cyclic Side Swapping: After a single side is scanned, the acquisition enable is blocked, and the pose adjustment unit is driven to rotate by a preset angle to the next station. This interlocking action is executed cyclically until the automated acquisition of all sides is completed.

[0076] S4 Asynchronous Alignment and Multidimensional Spatiotemporal Correlation

[0077] Raw data cross-node mapping: The real-time motion control node uses the direct memory access (DMA) mechanism to map the timestamped spectral raw data stream, angular coordinates, and initial quality data obtained in step S1 generated after each side is completed to the asynchronous data computing node at high speed across nodes.

[0078] Spatial self-registration: The asynchronous data computing node executes the algorithm in the background, using the unit with the smallest field of view value among all hyperspectral imaging units as the spatial anchoring reference, and performs hardware-level cropping coordinate mapping and resampling on other data sources, outputting a multi-dimensional hyperspectral data cube with absolute geometric alignment across the entire band.

[0079] Digital twin model fusion and inversion: Using aligned multi-band and multi-angle spectral data, the three-dimensional spatial distribution of physicochemical characteristics (such as sugar content and moisture content) within a few millimeters of the surface and subcutaneous layer of agricultural products is inverted; at the same time, the overall physical weight data measured by the quality sensing unit is combined to deduce the overall physical density properties of the sample at this volume, thereby identifying large-area hollow or pithy defects; finally, through cross-validation, a multi-dimensional heterogeneous fusion digital twin model for comprehensive quality evaluation is constructed and output.

[0080] Furthermore, to further clarify the collaborative solution process and working principle of the multi-parameter joint constraint model in this invention, a detailed operational example of this device is provided below, using a specific multi-band hyperspectral acquisition task for agricultural products as an example:

[0081] First hyperspectral imaging unit (visible and near-infrared camera): Horizontal pixel count Field of view Maximum hardware frame rate .

[0082] Second hyperspectral imaging unit (short-wave infrared camera): number of horizontal pixels Field of view Maximum hardware frame rate .

[0083] The preset number of sampling surfaces is set to 4 sides, and the sample turntable rotates 90° each time.

[0084] 2. Example Execution Steps S1: Sample Placement and Cross-Modal Feature Perception

[0085] The agricultural product to be tested was placed on the sample turntable. The quality sensing unit instantly read and recorded the weight as 2.5 kg.

[0086] The shape sensing unit acquires the outer contour of the agricultural product, and the collaborative control system calculates the push-broom angle of the camera turntable required to cover the agricultural product. .

[0087] Through localized micro-motion pre-scanning and exposure optimization, the optimal exposure time for the first camera was determined. (i.e., 0.008s), the optimal exposure time for the second camera (i.e., 0.012s).

[0088] 3. Example Execution Step S2: Closed-Loop Parameter Solving and Static Latching

[0089] The collaborative control system substitutes the above parameters into the joint constraint model for rapid solution:

[0090] Calculate the required number of scan rows (minimize distortion):

[0091] Number of rows required for the first camera: OK.

[0092] Number of rows required for the second camera: OK.

[0093] Find the minimum total time (Time synchronization interlock):

[0094] According to the formula For the sake of simplicity, let's assume... Approaching 0:

[0095] Lower limit of first camera exposure time: Frame rate lower limit: 1152 / 100 = 11.52s.

[0096] Lower limit of exposure time for the second camera: Frame rate lower limit: 960 / 60 = 16.0s.

[0097] The system extracts the global maximum value and determines the total push-broom time to be T = 16.0 s.

[0098] Issue and latch the final coordination parameters:

[0099] Push-broom angular velocity .

[0100] First camera operating frame rate: It satisfies the basic exposure constraint of 72 < 1 / 0.008.

[0101] Second camera operating frame rate: It satisfies the basic exposure constraint of 60 < 1 / 0.012.

[0102] The collaborative control system will Write to the underlying hardware registers and force latch.

[0103] 4. Example Execution Step S3: Automatic Loop Synchronous Data Acquisition

[0104] The push-broom drive unit drives the two cameras to rotate synchronously at an absolutely constant speed of 2.8125° / s, completing a 45° push-broom of the first side in 16.0 seconds. Due to the constant speed and strict alignment with the frame rate, the acquired 1152-line and 960-line images are completely consistent with the actual agricultural products in terms of aspect ratio.

[0105] After the push-broom operation is complete, the camera turntable resets, and the sample turntable rotates precisely 90° and locks its position.

[0106] Repeat the above sweeping process until all four sides are sampled, without any mechanical vibration caused by parameter changes.

[0107] 5. Example Execution Step S4: Background Alignment and Twin Modeling

[0108] While the physical page conversion is being performed on the front end, the asynchronous data computing node has already received the data from the first page in the back end.

[0109] The system recognizes that the field of view of the second camera is 30°, which is less than that of the first camera (40°). Therefore, based on the spatial coordinates of the second camera, it performs precise edge cropping and sub-pixel level resampling on the 1024×1152 pixel data acquired by the first camera, so that it achieves hardware-level spatial overlap with the 640×960 pixel data of the second camera.

[0110] Finally, the system binds the aligned four-plane dual-band hyperspectral data with the absolute physical weight of 2.5 kg using spatiotemporal tags, imports it into the preset quality detection model, and outputs a three-dimensional distribution map of the sugar content of the agricultural product and a density assessment report on whether there are cavities inside.

[0111] The above description is merely a specific embodiment of the invention, but the scope of protection of the invention is not limited thereto. Any variations or substitutions conceived without inventive effort should be included within the scope of protection of the invention. Therefore, the scope of protection of the invention should be determined by the scope defined in the claims.

Claims

1. A method for multi-band collaborative acquisition of multi-angle hyperspectral data of agricultural products, characterized in that: Includes the following steps: S1: After the sample to be tested is placed in the pose adjustment unit, the collaborative control system receives the initial mass data of the sample from the mass sensing unit and simultaneously triggers the shape sensing unit to scan the physical outer contour of the sample; the collaborative control system calculates the push-broom relative motion angle or displacement required to cover the sample based on the received contour data. Complete data initialization; S2: The cooperative control system will obtain the solution... By importing the internally preset multi-parameter joint constraint model and combining the different band responses and field-of-view attributes of each hyperspectral imaging unit, the number of scan lines that satisfy both spatial distortion-free and temporal strong synchronization is jointly calculated. Frame rate Exposure time and the uniform relative motion velocity of the push-broom drive unit Subsequently, the collaborative control system issues the above parameter instructions and writes them into the underlying registers of the corresponding hardware units for static latching, forming a strong mapping relationship between software and hardware parameters. S3: In the static latching state, the cooperative control system receives the pose lock ready signal sent by the pose adjustment unit, and synchronously sends acquisition enable signals to the pushbroom drive unit and each hyperspectral imaging unit, driving them to... Complete the joint push-broom of the current side; after the current side is acquired, the pose adjustment unit changes the preset angle and repeats this step until all sides are acquired. S4: The collaborative control system binds the hyperspectral image data stream acquired for each side to the corresponding angular spatial coordinates and the initial quality data acquired in step S1, according to the sample identity identifier, and pushes it to the asynchronous data computing node. The node then performs spatial self-registration of the multi-band hyperspectral image in the background operation mode.

2. The method for multi-band collaborative acquisition of multi-angle hyperspectral data of agricultural products according to claim 1, characterized in that: In step S2, the multi-parameter joint constraint model receives data from the morphology sensing unit to construct interactive constraints that minimize spatial distortion; to ensure that the image obtained by push-broom is consistent with the actual object in spatial geometry, the collaborative control system determines the spatial distortion based on the field of view of each hyperspectral imaging unit i. and horizontal pixel count Force a limit on the number of rows scanned. The formula must be satisfied: 。 3. The method for multi-band collaborative acquisition of multi-angle hyperspectral data of agricultural products according to claim 2, characterized in that: The multi-parameter joint constraint model in step S2 constructs time synchronization interaction constraints for multiple imaging units; the collaborative control system is controlled by a hardware bus, forcing each hyperspectral imaging unit to acquire data. The total time required for each frame must be equal and consistent with the execution time T of the controlled motion of the push-broom drive unit, satisfying the formula: And the unified relative motion speed command value The controlled system is set to .

4. The method for multi-band collaborative acquisition of multi-angle hyperspectral data of agricultural products according to claim 3, characterized in that: The collaborative control system achieves the minimum total time in the joint solution system. At that time, the following logical judgment and selection mechanism shall be established: Under the basic hardware constraints of hyperspectral imaging unit i Under this premise, the maximum value of the constraint boundary of all hyperspectral imaging units is extracted as the minimum instruction cycle of the control bus: ; In the formula, The optimal signal-to-noise ratio exposure time is dynamically locked by the collaborative control system based on the current illumination feedback. This represents the maximum frame rate limit provided by the hardware feedback of the hyperspectral imaging unit. The system's preset instruction execution redundancy is a small positive number.

5. A method for multi-band collaborative acquisition of multi-angle hyperspectral data of agricultural products according to claim 4, characterized in that: The optimal signal-to-noise ratio exposure time The dynamic locking mechanism is as follows: In step S1, the collaborative control system calls the morphology sensing unit to perform local micro-motion pre-scan. Based on the feedback of the acquired image histogram, the collaborative control system automatically masks and removes interference areas. Based on the peak value of the extracted high reflectivity uniform area histogram, the dynamic optimization algorithm is used to adjust and finally determine the locking mechanism. This value is then input into the multi-parameter joint constraint model.

6. The method for multi-band collaborative acquisition of multi-angle hyperspectral data of agricultural products according to claim 1, characterized in that: The static latching state of the parameters in step S3 serves as an anti-disturbance feedback mechanism; within the multi-faceted cyclic acquisition cycle of the same sample, the collaborative control system actively blocks dynamic speed adjustment interruption requests for the pushbroom drive unit, forcing the motor driver to operate according to the latched parameters. Maintaining a constant current output cuts off the transmission path of image micro-geometric distortion caused by nonlinear fluctuations in mechanical acceleration and deceleration.

7. A method for multi-band collaborative acquisition of multi-angle hyperspectral data of agricultural products according to claim 1, characterized in that: The collaborative control system is configured as a dual-node heterogeneous interactive architecture, including a real-time motion control node and an asynchronous data computing node; steps S2 and S3 are directly scheduled by the real-time motion control node through an industrial real-time bus to control the hardware unit; when an interrupt signal generated by each side in step S3 is triggered, the real-time motion control node maps the raw data with timestamps across nodes to the asynchronous data computing node through a direct memory access mechanism to execute step S4.

8. A method for multi-band collaborative acquisition of multi-angle hyperspectral data of agricultural products according to claim 7, characterized in that: The spatial self-registration performed by the asynchronous data computing node is controlled by the spatial reference command issued by the real-time motion control node. The asynchronous data computing node automatically extracts the unit with the smallest field of view value among all hyperspectral imaging units as the spatial anchoring reference, and performs hardware-level clipping coordinate mapping and resampling calculation on other data sources accordingly, thereby outputting a full-band geometrically aligned hyperspectral data cube.

9. A method for multi-band collaborative acquisition of multi-angle hyperspectral data of agricultural products according to claim 1, characterized in that: The logical interlock between the pose adjustment unit and the push-broom drive unit in step S3 is specifically manifested as follows: after the cooperative control system issues the pose change command, it continuously polls the encoder feedback value of the pose adjustment unit. Only when the encoder difference is detected to be zero and the dwell time exceeds the system tuning threshold will the enable flag bit that triggers the movement of the push-broom drive unit be activated.

10. A method for multi-band collaborative acquisition of multi-angle hyperspectral data of agricultural products according to claim 1, characterized in that: After step S4, the collaborative control system retrieves the associated packaged multidimensional spatiotemporal data and performs a fusion analysis step: using multi-band and multi-angle spectral data as the first input vector to invert the spatial distribution of the physicochemical characteristics of the sample, using the quality data fed back by the quality sensing unit as the second input vector to jointly reconstruct the overall density attribute of the sample, and outputting a multidimensional heterogeneous fusion digital model for quality evaluation through cross-validation.

Images