A system and method for detecting plant leaf wetness characteristics and determining weeds in a field

Abstract

The application discloses a kind of field plant leaf wetting characteristic detection and weed determination system and method, belong to plant leaf optical detection technical field, method includes: obtaining the physical property information of detection droplet, control detection droplet application device to the plant leaf to be detected with detection droplet;With the time of application as time reference, obtain the polarization image sequence and leaf surface spatial posture information in the process of leaf surface wetting response before detection droplet application;According to the polarization image before application, determine the polarization response baseline before application, determine wetting response according to the time sequence change of polarization image after application relative to the polarization response baseline before application;According to leaf surface spatial posture information, wetting response is corrected to leaf surface spatial posture, and wetting polarization response parameter is determined;Obtain crop wetting characteristic reference information, compare wetting polarization response parameter with crop wetting characteristic reference information, obtain weed determination result.The application is favorable to reduce the influence of leaf surface spatial posture difference on parameter comparison result.

Description

Technical Field

[0001] This invention relates to the field of optical detection technology for plant leaves, and in particular to a system and method for detecting the wetting characteristics of plant leaves and identifying weeds in the field. Background Technology

[0002] Field weed identification typically distinguishes target crops from other plants based on the color, shape, texture, or spectral reflectance characteristics of plant leaves. However, when target crops and weeds have similar leaf appearances in the early stages of growth, or when there are changes in light intensity, leaf shading, or background interference in the field, it is difficult to reliably characterize the physical differences in plant leaf surfaces based solely on static appearance characteristics.

[0003] The waxy layer, microstructure, and surface roughness of plant leaves affect the processes of droplet expansion, pinning, retention, rolling, shrinkage, and polarization response recovery on the leaf surface. Therefore, detecting leaf wetting characteristics can help distinguish different plants. However, in a field environment, different leaves typically exhibit differences in leaf normal, leaf tilt angle, incident angle, and observation angle. Furthermore, the area of ​​influence of the detected droplet, the morphology of the contact boundary, and the polarization response of the same leaf surface can vary with the leaf's spatial orientation. Consequently, the directly obtained wetting response includes both differences in the leaf's inherent wetting characteristics and differences in its spatial orientation, making it difficult to determine whether parameter deviations are caused by differences in plant species or by changes in leaf tilt and imaging angle.

[0004] In addition, existing leaf wetting detection methods mostly focus on a single static or geometric parameter in the contact angle, droplet area, or rolling state, which makes it difficult to simultaneously characterize the dynamic relationship between the contact boundary movement and the change in leaf polarization state. They also lack a complete judgment process that compares the detection results of the plant leaf to be tested with the reference leaf of the target crop under corresponding detection conditions.

[0005] Therefore, the technical problem that the existing technology needs to solve is: how to distinguish between changes in the leaf's own wetting characteristics and changes in the geometric and polarization responses caused by the leaf's spatial orientation when there are differences in the spatial orientation of leaves in the field, so that the wetting responses of the plant leaves to be tested and the reference leaves of the target crop can be compared under a unified orientation benchmark, and the weed determination results for the target crop planting area can be obtained accordingly. Summary of the Invention

[0006] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a method for detecting the leaf wettability characteristics of plants in the field and identifying weeds, used to identify the leaves of plants to be tested in a field where target crops are planted, comprising the following steps:

[0008] The target crop type is determined and the physical properties of the detection droplets are obtained. The detection droplet application device is controlled to apply the detection droplets to the predetermined detection area of ​​the leaf of the plant to be tested at a determined application time.

[0009] Using the application time as a time reference, the polarization image sequence and corresponding leaf spatial attitude information of the plant leaf before the application of the detection droplet and from the application time are obtained according to the acquisition time sequence.

[0010] The pre-application polarization response baseline is determined based on the pre-application polarization image in the polarization image sequence, and the wetting response is determined based on the temporal change of the post-application polarization image relative to the pre-application polarization response baseline.

[0011] The wetting response is corrected based on the leaf spatial attitude information, and converted to a predetermined leaf attitude reference. The wetting polarization response parameters are then determined based on the corrected wetting response.

[0012] Obtain crop wetting characteristic reference information corresponding to the target crop. The crop wetting characteristic reference information is obtained by detecting crop reference leaves or pre-stored. It corresponds to the physical properties of the detected droplets, the application conditions and the predetermined leaf attitude reference, and includes the reference range and deviation judgment conditions of the wetting polarization response parameters of the target crop. The deviation judgment conditions specify the combination relationship between one or more of the wetting polarization response parameters and the corresponding reference range.

[0013] The wetting polarization response parameters of the plant leaf to be tested are compared with the crop wetting characteristic reference information. The weed determination result is obtained according to whether the plant leaf to be tested meets the deviation determination condition. When the deviation determination condition is met, the plant leaf to be tested is determined to be a weed object in the target crop planting area.

[0014] Further, determining the wetting response and the wetting polarization response parameters includes:

[0015] The detection droplet's influence area and its contact boundary are determined from the polarization image sequence. The contact boundary is the boundary of the contact area between the detection droplet and the plant leaf to be detected. According to the acquisition sequence, a geometric response sequence is formed based on the area change of the detection droplet's influence area, the displacement change of the contact boundary, or the change of the expansion direction.

[0016] A polarization change sequence is formed based on the change in polarization response within the region affected by the detected droplet relative to the baseline of the polarization response before application;

[0017] Based on the application time and the acquisition time of each polarization image, the geometric response sequence and the polarization change sequence are time-aligned to determine the time correspondence between the motion state change of the contact boundary and the polarization response change.

[0018] The geometric response sequence, the polarization change sequence, and the time correspondence are used as the wetting response, and the wetting polarization response parameters are determined based on the corrected wetting response. The wetting polarization response parameters include wetting geometric parameters determined by the geometric response sequence, polarization change parameters determined by the polarization change sequence, and coupling timing parameters determined by the time correspondence.

[0019] Furthermore, based on the area change of the affected area of ​​the detected droplet, the displacement change of the contact boundary, and the polarization change sequence, the expansion stage of the detected droplet is identified, and at least one subsequent stage is identified from the boundary pinning stage, the holding stage, the rolling stage, the shrinkage stage, and the leaf surface polarization response recovery stage.

[0020] The wetting polarization response parameters include at least one of the following: duration of the expansion phase, expansion rate, anisotropy of the expansion direction, duration of boundary pinning, duration of holding, roll-off initiation time, contraction rate, peak polarization response time, polarization response recovery time, and the time difference between the start of contact boundary state change and polarization response state change.

[0021] Furthermore, obtaining the leaf surface spatial attitude information includes:

[0022] Acquire depth data or multi-view images of the plant leaf before the application of the detection droplet, and perform spatial registration and acquisition timing association with the polarization image sequence;

[0023] Fit the local surface of the leaf surface based on the local spatial points of the predetermined detection area and its neighborhood, and determine the leaf surface normal;

[0024] The spatial attitude relationship of the predetermined detection area relative to the gravity direction, illumination direction and imaging direction is determined based on the leaf surface normal, and the leaf surface normal and the spatial attitude relationship are used as the leaf surface spatial attitude information.

[0025] Furthermore, the leaf surface spatial attitude correction includes:

[0026] Using the leaf surface normal as the third coordinate axis, and following the order of the illumination direction, the imaging direction, and the horizontal axis of the image coordinate system of the polarized image, the first projection with a projection modulus greater than the effective projection threshold on the tangent plane of the local surface of the leaf is selected, and its normalized direction is used as the first coordinate axis. The cross product direction of the third coordinate axis and the first axis is used as the second coordinate axis to establish a right-handed leaf surface intrinsic coordinate system.

[0027] The direction transformation relationship between the polarization reference direction used for polarization image acquisition and the projection direction of the first coordinate axis on a plane perpendicular to the imaging direction is determined, and the incident angle and observation angle of the predetermined detection area in the leaf surface inherent coordinate system are determined; the predetermined leaf surface attitude reference includes the predetermined reference incident angle, reference observation angle and leaf surface inherent coordinate direction;

[0028] The polarization direction component in the wetting response is transformed according to the direction transformation relationship to determine the detection droplet influence area and its contact boundary in the polarization image, and the detection droplet influence area and its contact boundary are projected onto the tangent plane; the wetting response after coordinate transformation and projection is transformed to the predetermined leaf attitude reference according to the incident angle, the observation angle and the pre-calibrated attitude response mapping relationship.

[0029] Furthermore, multiple polarization images with different polarization directions are acquired in each polarization sampling period, and the total intensity Stokes component is determined based on the intensity of each polarization image. First linearly polarized Stokes component Second linearly polarized Stokes component ;

[0030] The angle by which the polarization reference direction is rotated in a plane perpendicular to the imaging direction to the projection direction of the first coordinate axis on that plane is denoted as the rotation angle. The rotation angle is defined as the counterclockwise direction when viewed along the direction of the imaging light propagation. The positive direction, and the first linearly polarized Stokes component according to the following formula. and the second linearly polarized Stokes component Perform coordinate transformation:

[0031] ;

[0032] ;

[0033] Determine the linear polarization intensity based on the converted Stokes component. Linear polarization degree and polarization direction ,in:

[0034] ;

[0035] ;

[0036] ;

[0037] respectively and As a measure of specular reflection and diffuse reflection, and based on the application of the detection droplet before and after... , , and The change determines the wetting polarization response parameters.

[0038] Furthermore, the crop wetting characteristic reference information is divided into multiple attitude reference partitions according to the leaf normal, leaf tilt angle, incident angle or observation angle of the crop reference leaf before leaf spatial attitude correction, and the wetting polarization response parameters in each attitude reference partition are converted to the same predetermined leaf attitude reference.

[0039] The corresponding attitude reference partition is determined based on the leaf surface spatial attitude information of the plant leaf to be tested, and the wetting polarization response parameters of the plant leaf to be tested are compared with the crop wetting characteristic reference information in the attitude reference partition.

[0040] Furthermore, the crop wetting characteristic reference information is established through the following steps:

[0041] Select a crop reference leaf that is the same as the target crop type and growth stage. The tested leaf surface of the crop reference leaf and the tested leaf surface of the plant to be tested are both paraxial or both abaxial.

[0042] The crop reference leaf was tested using the same detection droplets with the same physical properties as those used when testing the plant leaf to be tested, and with the same application control method and polarization image acquisition conditions.

[0043] Leaf spatial attitude correction is performed on the wetting response of the crop reference leaf. According to the attitude reference partition, the corresponding reference range is determined based on the distribution of wetting polarization response parameters in each partition. The deviation judgment condition is established based on the combination relationship between one or more wetting polarization response parameters and the corresponding reference range.

[0044] Furthermore, the wetting polarization response parameters are divided into a first response parameter group and a second response parameter group;

[0045] The first set of response parameters includes parameters characterizing the processes of droplet expansion, contact boundary displacement, boundary pinning, holding, rolling off, or contraction. The second set of response parameters includes parameters characterizing polarization intensity, degree of linear polarization, polarization direction, specular reflection characterization, diffuse reflection characterization, or leaf surface polarization response recovery process.

[0046] The crop wetting characteristic reference information includes the reference range corresponding to each parameter in the first response parameter group and the second response parameter group;

[0047] The deviation determination condition is: there is at least one parameter in the first response parameter group that exceeds the corresponding reference range, and there is at least one parameter in the second response parameter group that exceeds the corresponding reference range;

[0048] When the deviation determination condition is met, a weed determination result is obtained indicating that the leaf of the plant to be detected is a weed; otherwise, a weed determination result is obtained indicating that the leaf of the plant to be detected does not meet the deviation determination condition.

[0049] This invention also provides a system for detecting the leaf wettability characteristics of plants in the field and identifying weeds, used to determine the leaves of plants to be tested in a field where the target crop is grown, including:

[0050] A detection droplet application device includes a liquid storage section, a quantitative output section, and an application end, for applying a detection droplet with known physical properties to a predetermined detection area of ​​the leaf of the plant to be tested at a determined application time;

[0051] A polarization image acquisition device, comprising an illumination unit, a polarization optical component, and an electronic image sensor, is used to acquire a sequence of polarization images during the leaf wetting response process before and from the moment of application of the detection droplet.

[0052] A leaf spatial attitude acquisition device, including a depth sensor or a multi-view image acquisition component, is used to acquire leaf spatial attitude information corresponding to the polarization image sequence;

[0053] A synchronous control device is connected to the droplet detection application device, the polarization image acquisition device, and the leaf surface spatial attitude acquisition device, respectively, and is used to control the application time and synchronously control the acquisition of the polarization image sequence and the leaf surface spatial attitude information with the application time as a time reference.

[0054] A data processing device, connected to the polarization image acquisition device, the leaf spatial attitude acquisition device, and the synchronization control device, includes a processor and a memory. The memory stores crop wetting characteristic reference information corresponding to the physical properties, application conditions, and predetermined leaf attitude reference of the target crop and the detection droplet. The crop wetting characteristic reference information includes the reference range and deviation judgment conditions of the wetting polarization response parameters. The processor is used for:

[0055] The pre-application polarization response baseline is determined based on the pre-application polarization image, and the wetting response is determined based on the temporal variation of the post-application polarization image relative to the pre-application polarization response baseline.

[0056] The wetting response is corrected based on the leaf spatial attitude information and converted to the predetermined leaf attitude reference. The wetting polarization response parameters are then determined based on the corrected wetting response.

[0057] The wetting polarization response parameters are compared with the crop wetting characteristic reference information. The weed determination result is obtained based on whether the leaf of the plant to be detected meets the deviation determination condition. If the deviation determination condition is met, the leaf of the plant to be detected is determined to be a weed.

[0058] The beneficial effects of this invention are:

[0059] 1. This invention involves the controlled application of a detection droplet with known physical properties, and the formation of a geometric response sequence and a polarization change sequence using the pre-application polarization response baseline as a comparison benchmark. This allows the extraction of dynamic wetting changes caused by the action of the detection droplet from the original polarization state of the leaf surface, reducing the interference of inherent leaf surface reflection or polarization differences on the wetting polarization response parameters. This provides a more targeted detection basis for comparing the wetting characteristics of the plant leaf under test with the target crop.

[0060] 2. This invention acquires leaf surface spatial attitude information and converts the wetting response to a predetermined leaf surface attitude reference, so that the wetting polarization response parameters obtained under different leaf surface normals, incident angles and observation angles can be compared under a unified attitude reference, which helps to reduce the influence of differences in leaf surface spatial attitude on the parameter comparison results.

[0061] 3. This invention divides the wetting polarization response parameters into a first response parameter group characterizing dynamic geometric changes and a second response parameter group characterizing polarization changes, and performs joint deviation judgment with the crop wetting characteristic reference information corresponding to the target crop. This allows weed judgment to be based on both dynamic geometric response and polarization change, avoiding the formation of weed judgment results solely due to a single response parameter exceeding the reference range. Attached Figure Description

[0062] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0063] Figure 1 This is an overall flowchart of a method for detecting the leaf wettability characteristics of plants and identifying weeds in the field, according to the present invention.

[0064] Figure 2 This is a flowchart illustrating the determination of wetting response and wetting polarization response parameters in a field plant leaf wetting characteristic detection and weed identification method of the present invention.

[0065] Figure 3 This is a flowchart illustrating the acquisition and correction of leaf spatial posture information in a method for detecting the wetting characteristics of plant leaves and identifying weeds in the field, according to the present invention.

[0066] Figure 4 This is a flowchart illustrating the Stokes component coordinate transformation and wetting polarization response parameter determination process for a field plant leaf wetting characteristic detection and weed identification method of the present invention.

[0067] Figure 5 This is a flowchart illustrating the establishment of crop wetting characteristic reference information and deviation determination in a field plant leaf wetting characteristic detection and weed identification method of the present invention. Detailed Implementation

[0068] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0069] Example 1

[0070] Reference Figure 1 , Figure 2 and Figure 5 This is the first embodiment of the present invention, providing a method for detecting the leaf wettability characteristics of plants in the field and identifying weeds. The method is used to detect the deviation of the leaf wettability characteristics of the target plant relative to the target crop in a field where the target crop is planted, and to obtain a weed identification result based on the detection results. The method includes the following steps:

[0071] S1: Determine the target crop type and apply detection droplets in a controlled manner.

[0072] Specifically, the target crop type for the current planting area is determined, and the physical properties of the detection droplets are obtained. These physical properties include one or more of the following: droplet composition, volume, surface tension, dynamic viscosity, density, and temperature. Specific parameters are selected based on the influence of the corresponding physical properties on the droplet's expansion, retention, rolling, shrinkage, and leaf surface polarization response recovery process.

[0073] Physical property information can be obtained through actual measurements or read from verified liquid batch information. To ensure comparability of subsequent parameter comparisons, test droplets with the same physical properties or those confirmed by measurement are used for the plant leaves to be tested and the crop reference leaves used to establish crop wetting characteristics.

[0074] Subsequently, a predetermined detection area is determined on the leaf of the plant to be tested. Preferably, the predetermined detection area avoids the leaf margin, obviously damaged areas, and areas with large areas of liquid adhering to them, in order to reduce the impact of local abnormalities on the wetting response.

[0075] After determining the predetermined detection area, the detection droplet application device is controlled to apply a detection droplet to the predetermined detection area, and the trigger time of the detection droplet application device is recorded. The moment when the detection droplet first makes contact with the predetermined detection area is determined as the application moment. This application moment can be determined based on the acquisition time corresponding to the polarization image of the first appearance of the detection droplet contact area in the polarization image sequence, or based on the droplet contact detection signal generated by the detection droplet application device or near the predetermined detection area. The action trigger time is used to initiate synchronous acquisition related to droplet application and does not replace the application moment.

[0076] To ensure that the plant leaf under test and the crop reference leaf are under corresponding testing conditions, both are subjected to the same or calibrated equivalent droplet output method, application direction, and application control process. The specific volume, application distance, and application speed of the detection droplet are determined based on the calibration results of the detection droplet application device and the imaging resolution of the predetermined detection area, without pre-setting unverified fixed values.

[0077] Therefore, by controlling the physical properties and application conditions of the detection droplets, and using the actual application time of the detection droplets in contact with the leaf surface as the time reference, a unified detection basis is provided for subsequently establishing a polarization response baseline before application, determining the wetting response, and comparing wetting polarization response parameters.

[0078] S2: Obtain polarization image sequence and leaf spatial attitude information.

[0079] Specifically, using the application time as a time reference, polarization images of the plant leaves to be tested are acquired before the application of the detection droplet, and polarization images are acquired continuously or in a predetermined acquisition sequence during the leaf wetting response process starting from the application time, forming a polarization image sequence.

[0080] Associate each polarization image in the polarization image sequence with its acquisition time. Let the first polarization image be... The acquisition time of the polarization image corresponding to each valid acquisition time is: The application time is Then the relative acquisition time of the polarization image with respect to the application time is: ;in, The polarization image is the image before polarization is applied. The polarization image is the polarization image after polarization is applied.

[0081] Simultaneously, leaf spatial attitude information corresponding to the polarization image sequence is acquired, and based on the acquisition time and spatial registration relationship, the leaf spatial attitude information is associated with the predetermined detection area and the polarization image sequence. The leaf spatial attitude information can at least characterize the spatial attitude of the predetermined detection area relative to the imaging direction, illumination direction, or gravity direction.

[0082] In one specific implementation, before determining the pre-application polarization response baseline, based on the spatial correspondence characterized by leaf spatial attitude information, or based on a stable leaf surface region within a predetermined detection area unaffected by the detected droplets, each polarization image in the polarization image sequence is spatially aligned with the reference pre-application polarization image, so that the same image position at different acquisition times corresponds to the same leaf surface position on the leaf of the plant to be detected. Subsequent positions All represent the physical position of the same leaf surface after spatial alignment. When the spatial alignment residual of the polarization image exceeds the allowable range determined by the repetition alignment error of multiple pre-applied polarization images or a pre-calibrated spatial alignment error, the corresponding polarization image is marked as an invalid image and is not used to determine the wetting response.

[0083] It should be noted that the specific methods for acquiring leaf spatial attitude information, determining leaf normal, and calculating spatial attitude relationships are described in Example 2. This example mainly utilizes leaf spatial attitude information to establish a data conversion relationship between polarization image sequences and predetermined leaf attitude references.

[0084] S3: Determine the wetting response and establish the calculation relationship of the wetting polarization response parameters.

[0085] This step uses the pre-application polarization response as a benchmark to extract the dynamic geometric and polarization changes of the detected droplet from the polarization image sequence. It then correlates these two changes temporally according to the application time to form a wetting response for leaf spatial attitude correction and establishes the calculation relationship for the wetting polarization response parameters. Specifically, it includes the following sub-steps:

[0086] S301: Determine the baseline for applying the pre-polarization response.

[0087] Specifically, the effective pre-polarization image is selected from a spatially aligned sequence of polarization images. For the location within the predetermined detection region... Passing the exam The polarization response component is denoted as . The polarization response component corresponding to the pre-polarization image is The median value of the same polarization response component at the same leaf surface position in each pre-polarization image is taken to obtain the pre-polarization response baseline of that polarization response component. .

[0088] The baseline values ​​of each polarization response component together constitute the polarization response baseline before application. The polarization response components may include the image intensity corresponding to different polarization directions or the polarization physical quantities determined by them. The specific polarization physical quantities and their calculation methods are described in Example 2.

[0089] When there is only one valid pre-polarization image, the values ​​of each polarization response component in the polarization image can be used as the corresponding pre-polarization response baseline, and the baseline fluctuation range can be determined according to the noise calibration result of the polarization image acquisition device or the pre-observation results obtained under the same acquisition conditions. When there are multiple valid pre-polarization images, it is preferable to use the median value of each polarization response component as the baseline value, and the baseline fluctuation range can be determined according to the multiple pre-polarization images.

[0090] Therefore, the pre-application polarization response baseline can characterize the polarization state of the predetermined detection area before the droplet is applied, providing a comparative benchmark for distinguishing the original polarization characteristics of the leaf surface from the polarization changes caused by the droplet.

[0091] S302: Determine the area of ​​influence and contact boundary of the detection droplet.

[0092] Based on the pre-application polarization response baseline, the polarization response changes at the same leaf surface position in each post-application polarization image were determined.

[0093] For the The position of the applied polarization image corresponding to each valid acquisition time is calculated. Each effective polarization response component relative to the corresponding baseline value The component differences are determined as follows: For polarization intensity and linear polarization degree, the component differences are determined by the absolute value of the difference between the applied value and the corresponding baseline value; for polarization direction, the component differences are determined by the minimum directional difference considering the periodicity of the polarization direction. The method for calculating the periodic difference of the polarization direction is further explained in Example 2.

[0094] The component difference of each polarization response component is compared with the region determination threshold corresponding to that component to obtain the corresponding candidate change region. The region determination threshold of each component is determined based on the pre-application fluctuation of the corresponding polarization response component, the measurement error of the imaging equipment, and the region repeated extraction error. The polarization response components of different dimensions are not directly summed or the unnormalized unified vector norm is not calculated.

[0095] Furthermore, from the union of all candidate change regions, a connected region containing the initial contact position of the detected droplet and maintaining spatial continuity at adjacent sampling times is selected and determined as the droplet influence region. For isolated areas of change caused by momentary specular brightness, local occlusion, or noise, they are excluded based on their connectivity with the initial contact position of the detected droplet and their persistence in adjacent sampling times.

[0096] During the period when the detection droplet is in contact with the leaf of the plant to be tested, the area affected by the detection droplet is the continuous wetting area formed after the detection droplet comes into contact with the leaf of the plant to be tested. The outer boundary of the part of the continuous wetting area in which the detection droplet contacts the leaf of the plant to be tested is defined as the contact boundary.

[0097] After the detection droplet rolls away from the predetermined detection area or the main body of the detection droplet disappears, the region where polarization response changes still occur is defined as the residual polarization response region. The residual polarization response region is used to determine the leaf surface polarization response recovery process, and its boundary is no longer considered as the contact boundary.

[0098] S303: Forms a geometric response sequence and a polarization change sequence.

[0099] After obtaining the affected area and contact boundary of the detected droplet, according to the acquisition sequence of the polarization image, the area of ​​the affected area, the center position of the area, the position of the contact boundary, the displacement of the contact boundary relative to the previous effective acquisition time, and the expansion distance of the detected droplet in different directions are determined for each effective acquisition time, forming a geometric response sequence: ;in, For the first The geometric response at each valid acquisition moment includes at least one of the following: the area of ​​the droplet-affected region, the displacement of the contact boundary, and the direction of expansion; during the wetting phase identification of S304, the geometric response... This includes at least detecting the area of ​​the droplet's influence region and the displacement of the contact boundary; and, when determining the anisotropy of the propagation direction, the geometric response. It also includes extended direction information; The acquisition time is relative to the application time; This represents the number of effective data collection moments during the leaf wetting response process.

[0100] Simultaneously, based on the changes in each polarization response component within the droplet's influence region relative to the pre-application polarization response baseline, the polarization change at each effective acquisition moment is determined. This forms a polarization change sequence: Among them, the polarization change The mean, median, extreme values, or spatial distribution of the polarization response differences within the affected region of the detected droplet can be used to represent the polarization changes. In one specific implementation, the median value of the polarization response differences is used to characterize the overall polarization change in the affected region of the detected droplet, thereby reducing the impact of a small number of abnormal pixels on the polarization change sequence.

[0101] Subsequently, the geometric response sequence and polarization change sequence are time-aligned based on the application time and the acquisition time of each polarization image. When the geometric response and polarization change are obtained from the same polarization image, they are directly assigned the same relative acquisition time; when they come from different acquisition channels, the temporally corresponding observations are selected based on the acquisition timestamps, or interpolation is performed between adjacent valid acquisition times. The allowable time difference is determined based on the polarization image acquisition period and the rate of change of the detected wetting process.

[0102] The time-aligned geometric response sequence, polarization change sequence, and the time correspondence between the contact boundary motion state change and the polarization response change are collectively used as the wetting response. By correlating and detecting the geometric and polarization changes of the droplet under the same time reference, the dynamic relationship between the contact boundary motion and the change in the leaf surface polarization state can be characterized.

[0103] S304: Identify different stages in the leaf wetting response process.

[0104] Based on the geometric response sequence and polarization change sequence, the expansion phase of the detected droplet is identified as follows, and at least one subsequent phase is identified from the boundary pinning phase, holding phase, rolling phase, contraction phase, and leaf surface polarization response recovery phase:

[0105] When the area affected by the detected droplet increases and the contact boundary moves outward beyond the boundary repeated extraction error, it is identified as the expansion stage;

[0106] When the contact boundary displacement does not exceed the allowable range of boundary displacement, but the polarization change sequence continues to change, it is identified as the boundary pinning stage;

[0107] When both the contact boundary displacement and polarization response fluctuations are within their respective allowable ranges, it is identified as the holding phase;

[0108] When the center position of the detected droplet's affected area is aligned with the contact boundary and the displacement exceeds the continuous displacement of the boundary repeated extraction error, it is identified as the rolling stage.

[0109] When the area affected by the detection droplet decreases and the contact boundary moves toward the interior of the area affected by the detection droplet, it is identified as the contraction phase.

[0110] When the detection droplet rolls away from the predetermined detection area or the main body of the detection droplet disappears, and the polarization response of the residual polarization response area gradually approaches the baseline of the polarization response before application, it is identified as the leaf surface polarization response recovery stage.

[0111] The allowable range of boundary displacement is determined based on the repeated extraction error of the contact boundary, and the allowable range of polarization response fluctuation is determined based on the fluctuation of the polarization response before application. The candidate start time and candidate end time for each stage are determined based on the effective acquisition time when the corresponding state first continuously satisfies and no longer satisfies the above identification conditions, respectively.

[0112] It should be noted that S304 first determines the candidate start and end times of each wetting stage based on the wetting response without leaf spatial attitude correction; when the leaf spatial attitude correction in S4 causes a change in the identification criteria of the corresponding stage, the start and end times of the corresponding stage are re-determined based on the corrected wetting response.

[0113] S305: Establish the calculation relationship of wetting polarization response parameters.

[0114] After determining the start and end times of each wetting stage, the calculation relationships of wetting geometric parameters, polarization change parameters, and coupling timing parameters are established. The final wetting polarization response parameters, involving the area of ​​the droplet's influence region, contact boundary displacement, expansion direction, and polarization direction components, are determined using the corrected wetting response obtained from S4 as follows:

[0115] The duration of the expansion phase is: ;in, and These are the start and end times of the expansion phase, respectively.

[0116] The expansion rate is determined by the ratio of the area change at the start and end of the expansion phase to the duration of the expansion phase: ;in, and These represent the areas of the droplet-affected region after leaf surface attitude correction at the start and end times of the expansion phase, respectively; only in... Calculate the expansion rate at any time.

[0117] The contraction rate is determined based on the ratio of the area reduction at the start and end of the contraction phase to the duration of the contraction phase: ;in, and These represent the areas of the droplet-affected region after leaf surface spatial attitude correction at the start and end times of the contraction phase, respectively. and These are the corresponding times, and .

[0118] Anisotropy in the direction of expansion is determined by the maximum expansion distance from the center of the droplet's influence region to the contact boundary at the same effective acquisition time. and minimum expansion distance Sure: ;in, and This represents the extended distance after blade spatial attitude correction, expressed under a predetermined blade attitude reference. Only... The anisotropy of the expansion direction is calculated in real time; if this condition is not met, the parameter is marked as invalid. In a specific implementation, the time corresponding to the termination of the expansion phase is... The anisotropy of the extended direction is determined for parameter comparison; the plant leaf to be tested and the crop reference leaf are subjected to the same value taking time and calculation rules.

[0119] The duration of boundary pinning and the duration of holding are determined based on the difference between the end time and the start time of the corresponding stage, respectively; the roll-off start time is determined as the effective acquisition time corresponding to the center position of the detection droplet's influence area and the first continuous displacement in the same direction as the contact boundary.

[0120] The peak time of the polarization response is determined based on the polarization change sequence. In one specific implementation, the effective acquisition time when the absolute value of the polarization change reaches its maximum value after the droplet is applied is determined as the peak time of the polarization response; if multiple identical maximum values ​​exist, the earliest effective acquisition time is used.

[0121] The polarization response recovery time is defined as the time interval from the moment the detection droplet begins to roll away from the predetermined detection area, the main body of the detection droplet disappears, or the residual polarization response area begins to form, until the polarization response within the residual polarization response area re-enters the baseline allowable range determined by the pre-application polarization response fluctuations and remains continuously stable. If the polarization response has not yet entered the baseline allowable range at the end of the acquisition, the end time of acquisition is not considered the polarization response recovery time, but rather the parameter is marked as incomplete observation.

[0122] The initial time difference between the change in the contact boundary state and the change in the polarization response state is expressed as: ;in, The starting point at which the contact boundary begins to expand, pin, roll off, or contract. This refers to the moment when the polarization response state begins to change corresponding to the change in the contact boundary state. Both starting moments are determined based on the first valid acquisition moment when the corresponding change exceeds the baseline fluctuation or repeated extraction error.

[0123] Through the above processing, S3 forms the wetting response without leaf spatial attitude correction, the candidate start and end times of each wetting stage, and the calculation relationship of wetting polarization response parameters; S4 further performs leaf spatial attitude correction on the wetting response and determines the final wetting polarization response parameters accordingly.

[0124] S4: Perform leaf surface spatial attitude correction on the wetting response and determine the wetting polarization response parameters.

[0125] This step is used to convert the wetting response obtained under different leaf spatial attitudes to a unified predetermined leaf attitude reference, so as to reduce the influence of differences in leaf tilt, illumination direction and imaging direction on the wetting response, and obtain wetting polarization response parameters that can be compared with crop wetting characteristic reference information.

[0126] Specifically, the wetting response obtained in S3 and the candidate start and end times of each wetting stage are correlated with the corresponding leaf spatial attitude information. Based on the leaf spatial attitude information, the conversion relationship between the leaf attitude during detection and the predetermined leaf attitude reference is determined. The predetermined leaf attitude reference is used to uniformly represent the wetting response obtained under different leaf spatial attitudes, including the predetermined reference incident angle, reference observation angle, and leaf inherent coordinate direction.

[0127] Subsequently, based on the conversion relationship, the area of ​​the detection droplet's influence region, the displacement and direction of the contact boundary, the expansion direction, and the polarization response amplitude and polarization response direction components affected by the leaf incident angle, observation angle, and the inherent coordinate direction of the leaf surface in the wetting response are converted to obtain the corrected wetting response represented relative to the predetermined leaf attitude reference.

[0128] After completing the leaf spatial attitude correction, based on the corrected wetting response and the candidate start and end times of each wetting stage identified by S3, the wetting geometric parameters, polarization change parameters, and coupling timing parameters are determined according to the calculation relationship established by S3, forming the wetting polarization response parameters of the plant leaf to be tested. When the area change, contact boundary displacement change, or polarization change characterized by the corrected wetting response is inconsistent with the stage identification results before correction, the start and end times of the corresponding wetting stage are re-determined based on the corrected wetting response, and the corresponding wetting polarization response parameters are determined accordingly.

[0129] The establishment of the leaf's intrinsic coordinate system, polarization direction transformation, and attitude response mapping relationship used in leaf spatial attitude correction are specifically described in Example 2. The wetting polarization response parameters obtained in this step are used for crop wetting characteristic reference information matching in S5 and deviation determination in S6.

[0130] S5: Establish or obtain reference information on crop wetting characteristics corresponding to the target crop.

[0131] This step is used to obtain the reference range and deviation judgment conditions corresponding to the target crop type, the physical properties of the detection droplets, the application conditions, and the predetermined leaf attitude benchmark, so as to provide a basis for S6 to compare the wetting polarization response parameters of the plant leaves to be tested and the target crop under the same detection basis.

[0132] When the pre-stored crop wetting characteristic reference information corresponds to the current target crop type and detection conditions, the corresponding crop wetting characteristic reference information is read directly; when there is no corresponding crop wetting characteristic reference information, it is established through the following steps.

[0133] S501: Select and test crop reference leaves.

[0134] Select a crop reference leaf that is the same as the target crop type and growth stage, and ensure that the tested leaf surface of the crop reference leaf and the tested leaf of the plant to be tested are both paraxial or both abaxial.

[0135] Using detection droplets with identical physical properties, and applying the same or calibrated equivalent control methods and polarization image acquisition conditions, crop reference leaves were detected. The wetting response of each crop reference leaf was determined according to S1-S4, and leaf surface spatial attitude correction was performed to obtain wetting polarization response parameters relative to the same predetermined leaf surface attitude reference.

[0136] S502: Divide the attitude reference zones and generate reference observation results.

[0137] Based on one or more of the following parameters of the crop reference leaf before leaf spatial attitude correction: leaf normal, leaf tilt angle, incident angle, or observation angle, the observation results of the crop reference leaf are divided into corresponding attitude reference zones. The boundaries of the attitude reference zones are determined based on the correction residuals of the attitude response mapping relationship, the measurement uncertainty of the leaf spatial attitude information, and the attitude coverage of the crop reference samples, without pre-setting unverified fixed angle ranges.

[0138] The uncorrected leaf spatial attitude information is used to select the attitude reference zone, and the wetting polarization response parameters after leaf spatial attitude correction are used to establish a reference range and compare execution parameters within the corresponding attitude reference zone.

[0139] The reference observations used to determine the parameter distribution are preferably derived from different crop reference leaves. Repeated detection results of the same crop reference leaf are used to evaluate detection repeatability and determine the representative parameter value for that crop reference leaf, and are not counted as independent crop reference samples. For example, the median of multiple valid detection results for the same crop reference leaf can be used as its representative parameter value.

[0140] S503: Determine the parameter reference range and deviation judgment conditions.

[0141] The candidate wetting polarization response parameters are divided into a first response parameter group and a second response parameter group. The first response parameter group includes parameters characterizing the processes of droplet expansion, contact boundary displacement, boundary pinning, holding, rolling off, or contraction. The second response parameter group includes parameters characterizing polarization intensity, degree of linear polarization, polarization direction, specular reflection, diffuse reflection, or leaf surface polarization response recovery process.

[0142] Before determining the final parameters for weed identification, the measurability and repeatability of candidate parameters can be evaluated using crop reference samples, and known weed samples can be used to test whether the candidate parameters can form repeatable deviations relative to the target crop reference range. Only parameters that are readily obtainable, have stable calculation results, and have complementary physical meanings between the two response parameter sets are used to establish deviation identification criteria. Known weed samples are only used to determine parameter combinations and are not used as a comparison reference for online identification of the plant leaves to be tested.

[0143] For the final wetting polarization response parameters, within each attitude reference zone, the corresponding reference range is determined using an empirical distribution interval or a statistically permissible interval based on the parameter distribution of multiple valid and independent crop reference observations. The coverage level of the reference range is determined based on the distribution stability of the crop reference samples, sample coverage, and pre-determined judgment tolerances, without pre-setting unverified empirical quantiles. When the valid crop reference observations within an attitude reference zone are insufficient to establish a reference range, that attitude reference zone is not marked as having a valid reference range.

[0144] In one specific implementation, the representative values ​​of parameters corresponding to the reference observation results of each independent crop within the same attitude reference zone are arranged in numerical order, and the lower empirical quantile value and the upper empirical quantile value are selected according to the predetermined coverage requirements, which are respectively used as the lower limit and upper limit of the corresponding parameter reference range.

[0145] Deviation judgment conditions are established based on the reference ranges corresponding to each parameter in the two response parameter groups: when at least one parameter in the first response parameter group exceeds the corresponding reference range, and at least one parameter in the second response parameter group exceeds the corresponding reference range, the plant leaf to be tested is determined to meet the deviation judgment conditions.

[0146] S504: Provides reference information on crop wetting characteristics.

[0147] The target crop type, growth stage, leaf surface being tested, physical properties of the test droplets, application conditions, predetermined leaf surface attitude benchmark, attitude reference zone, final parameter combination, reference range of each parameter, and deviation judgment conditions are associated and stored to form crop wetting characteristic reference information, which is then provided to S6 for parameter comparison and weed judgment.

[0148] S6: Compare the wetting polarization response parameters and obtain the weed determination results.

[0149] This step is used to match the crop wetting characteristics reference information corresponding to the detection conditions for the leaves of the plant to be tested, and, under the condition that the detection data and reference information are valid, to obtain the weed judgment result based on the deviation judgment condition.

[0150] Specifically, based on the leaf spatial attitude information of the plant leaf before leaf spatial attitude correction, the corresponding attitude reference zone is determined. The parameter reference range and deviation judgment conditions corresponding to the current target crop type, target crop growth stage, the detected leaf surface, the physical properties of the detection droplet, application conditions, and the predetermined leaf attitude benchmark are read from the crop wetting characteristic reference information. If there is no corresponding crop wetting characteristic reference information, or if a valid reference range has not been established for the corresponding attitude reference zone, "No valid weed judgment result formed" is output, and deviation judgment is not performed.

[0151] Before determining deviations, the detection data and the wetting polarization response parameters used in the determination are checked for validity. Validity checks include: whether the polarization image sequence contains valid pre-application polarization images and wetting response process images; whether each polarization image meets spatial alignment requirements; whether the droplet-affected area and the contact boundary during droplet contact can be extracted according to the corresponding wetting stage; whether the leaf surface spatial attitude information corresponds to the polarization image sequence; and whether the wetting polarization response parameters used in the determination have complete calculation inputs, valid start and end times, and corresponding reference ranges. When it is necessary to determine the polarization response recovery time, it is also checked whether the residual polarization response region can be effectively extracted.

[0152] If the polarization image sequence, leaf surface spatial attitude information, or wetting polarization response parameters involved in deviation determination do not meet the above validity requirements, resulting in the deviation determination conditions not being fully executed, the output "Invalid detection data" will be displayed, and the deviation determination will not be executed.

[0153] When both the detection data and crop wetting characteristic reference information are valid, the wetting polarization response parameters involved in the determination are compared with their corresponding reference ranges. If at least one parameter in the first response parameter group exceeds the corresponding reference range, and at least one parameter in the second response parameter group also exceeds the corresponding reference range, the plant leaf to be detected is determined to meet the deviation determination condition, and a weed determination result is obtained indicating that the plant leaf to be detected is a weed object in the target crop planting area.

[0154] When the parameters in both response parameter groups can be effectively compared, but the above deviation requirements are not simultaneously met, a weed determination result indicating "the current detection result does not meet the deviation judgment condition" is obtained. This result only indicates that the current detection did not meet the weed judgment condition, and does not mean that all the wetting characteristics of the tested plant leaves are the same as those of the target crop.

[0155] "Invalid detection data" indicates that the currently collected data or parameter calculation results cannot support the deviation judgment; "No valid weed judgment result formed" indicates that there is a lack of valid crop wetting characteristic reference information that matches the current detection conditions. Neither of these indicates that the leaves of the plant under test have not deviated from the wetting characteristics of the target crop.

[0156] By combining the above-mentioned deviation determination, weed determination can be based on both dynamic geometric response and polarization change, reducing the possibility of weed determination results being formed by fluctuations of a single parameter.

[0157] In summary, this embodiment, by applying detection droplets in a controlled manner, combining wetting dynamic geometric response, polarization change, and leaf spatial attitude correction, and obtaining weed determination results based on crop wetting characteristic reference information corresponding to the target crop, helps to reduce the impact of leaf spatial attitude differences and fluctuations in single wetting polarization response parameters on field weed determination results.

[0158] Example 2

[0159] Reference Figure 3 and Figure 4 This is the second embodiment of the present invention. Based on embodiment 1, this embodiment further explains the acquisition of leaf spatial attitude information and the process of leaf spatial attitude correction.

[0160] This embodiment uses the polarization image sequence, predetermined detection area, acquisition time sequence, and wetting response obtained in Embodiment 1 as input. By acquiring the three-dimensional spatial information of the predetermined detection area, the local surface of the leaf and the leaf normal are determined, a leaf surface intrinsic coordinate system is established, and the geometric response and polarization response in the wetting response are transformed to obtain the corrected wetting response and wetting polarization response parameters relative to the predetermined leaf attitude reference. The obtained wetting polarization response parameters are provided to S5 and S6 of Embodiment 1 for crop wetting characteristic reference information matching and weed identification. This embodiment includes the following steps:

[0161] S201: Acquire three-dimensional spatial data and establish data correspondence.

[0162] This step is used to obtain the three-dimensional spatial information of the predetermined detection area and its neighborhood, and to establish the spatial correspondence and acquisition timing relationship between the three-dimensional spatial information and the polarization image sequence.

[0163] Specifically, before applying the detection droplet, depth data or multi-view images of the plant leaf to be detected are acquired. When using depth data, pixels in the depth image corresponding to the predetermined detection area and its neighborhood are converted into spatial points according to the internal calibration parameters of the depth measurement device. When using multi-view images, corresponding image points are determined according to the internal and external calibration parameters of each viewpoint, and corresponding local spatial points on the leaf surface are obtained through triangulation.

[0164] By establishing a spatial transformation relationship between a pre-calibrated depth data coordinate system or multi-view image coordinate system and a polarization image coordinate system, local spatial points on each leaf surface are projected onto the polarization image coordinate system, or a predetermined detection area in the polarization image is back-projected onto three-dimensional space, thereby establishing a correspondence between local spatial points on the leaf surface and pixels in the polarization image. The spatial transformation relationship can be determined by acquiring depth data, multi-view images, and polarization images of a common calibration object and based on common calibration points.

[0165] Furthermore, the leaf spatial attitude corresponding to the depth data or multi-view images acquired before the droplet application is used as the initial leaf spatial attitude, and it is associated with the polarization image before application based on the acquisition timestamp. For the polarization image after the droplet application, when the leaf spatial attitude of the predetermined detection area remains stable within the allowable range of variation, the initial leaf spatial attitude is associated with the corresponding polarization image.

[0166] When the plant leaf under test undergoes overall movement during the collection process, a spatial transformation relationship can be determined between the pre-application polarization image and each post-application polarization image based on the stable leaf surface region outside the predetermined detection area that is not affected by the detection droplets. This transformation is then used to update the representation of local spatial points on the leaf surface and the leaf surface normal in the imaging coordinate system. In one specific embodiment, a planar homography matrix is ​​determined based on non-collinear corresponding points in the stable leaf surface region. Combined with the internal calibration parameters of the imaging device and the local leaf surface before application, the rotational relationship of the local leaf surface relative to the initial spatial attitude of the leaf surface is determined.

[0167] When only two-dimensional translation can be obtained and three-dimensional attitude change cannot be uniquely determined, the two-dimensional image alignment result is not directly used as the leaf space attitude update result. In this case, the initial leaf space attitude is used only if the leaf space attitude change does not exceed the calibration allowable range; otherwise, the leaf space attitude information of the corresponding polarization image is marked as invalid.

[0168] When the spatial registration residual exceeds the allowable range determined based on the calibration repeatability error and the three-dimensional measurement uncertainty, or when the leaf of the plant to be tested undergoes local deformation that cannot be represented by spatial transformation relationships, the corresponding leaf spatial attitude information or polarization image is marked as invalid and is not used for leaf spatial attitude correction.

[0169] S202: Fit the local surface of the leaf and determine the leaf normal.

[0170] This step is used to determine the local surface of the leaf based on the local spatial points of the leaf surface in the predetermined detection area and its neighborhood, and to obtain the leaf surface normal with a unique orientation.

[0171] Let the local spatial points on the leaf surface used for fitting the local surface of the leaf be . ,in, , This represents the number of local spatial points on the leaf surface participating in the fitting. First, the center positions of these local spatial points are calculated:

[0172] ;

[0173] Then, the covariance matrix is ​​established:

[0174] ;

[0175] in, The center position of a local spatial point on the leaf surface. This is the covariance matrix characterizing the distribution of local spatial points on the leaf surface. Regarding the covariance matrix... Eigenvalue decomposition is performed, and the unit eigenvector corresponding to the smallest eigenvalue is determined as the initial normal of the local surface of the leaf.

[0176] The local spatial points on the leaf surface used for fitting should include at least non-collinear spatial points that can determine the local surface. When the local spatial points on the leaf surface tend to be collinear, or when the characteristic direction corresponding to the smallest eigenvalue of the covariance matrix cannot be distinguished from other characteristic directions based on the three-dimensional measurement uncertainty, this set of spatial points should not be used to determine the leaf surface normal.

[0177] To eliminate the problem of the initial normal having two directions, positive and negative, the imaging direction is defined as a unit direction pointing from the predetermined detection area to the imaging device. When the inner product of the initial normal and the imaging direction is less than zero, the initial normal is reversed; otherwise, the initial normal remains unchanged, and the leaf normal facing the imaging device is obtained. For implementations where the outer direction of the leaf surface to be detected has been determined, the leaf normal can also be unified according to the predetermined outer direction of the leaf surface, but the same normal orientation rule is used for the crop reference leaf and the leaf of the plant to be detected in the same detection task.

[0178] Preferably, the fitting results are verified based on the distance residuals from local spatial points on the leaf surface to the fitted local surface, the eigenvalue distribution of the covariance matrix, and the three-dimensional measurement uncertainty. When the fitting results cannot stably determine the local surface of the leaf surface, the corresponding leaf spatial attitude information is marked as invalid.

[0179] In this embodiment, the illumination direction is defined as the unit direction from the predetermined detection area to the illumination unit, the imaging direction is defined as the unit direction from the predetermined detection area to the imaging device, and the gravity direction is defined as the unit direction pointed to by gravitational acceleration. Specifically, the direction in which the illumination light propagates from the illumination unit to the leaf surface is opposite to the illumination direction, while the direction in which the imaging light propagates from the leaf surface to the imaging device is the same as the imaging direction.

[0180] Based on the vector relationships between the leaf surface normal and the directions of gravity, illumination, and imaging, the spatial attitude relationship of the predetermined detection area relative to these directions is determined. The leaf surface normal and the spatial attitude relationship are used as the leaf surface spatial attitude information.

[0181] S203: Establish the inherent coordinate system of the leaf surface.

[0182] This step is used to establish a right-handed blade surface inherent coordinate system with a defined orientation on a local surface of the blade, providing a unified orientation reference for the geometric response and polarization response transformation under different blade surface spatial attitudes.

[0183] Specifically, the third coordinate axis, which serves as the inherent coordinate system of the leaf surface, is denoted as: ;in, The leaf surface normal is determined for S202.

[0184] Following the order of illumination direction, imaging direction, and the horizontal axis of the image coordinate system of the polarized image, the corresponding three-dimensional directions are sequentially selected as candidate directions. In this process, the three-dimensional direction corresponding to the horizontal axis of the image coordinate system of the polarized image is transformed to the coordinate system of the local spatial point on the leaf surface according to the internal and external calibration parameters of the imaging device.

[0185] For any candidate direction Its projection on the local tangent plane of the leaf surface for: .

[0186] Following the order of the candidate directions, from the projections of each candidate direction onto the tangent plane, select the first projection whose projection magnitude is greater than the effective projection threshold, and use its normalized direction as the first coordinate axis: .

[0187] The effective projection threshold is determined based on the calibration error of the candidate direction, the measurement uncertainty of the blade normal, and the accuracy of numerical calculation. This avoids instability in the first coordinate axis direction caused by the tangent plane projection approaching zero when the candidate direction is close to the blade normal.

[0188] The cross product direction of the third coordinate axis and the first coordinate axis is used as the second coordinate axis: ;Depend on , and A right-handed blade surface intrinsic coordinate system is established. When the projection modulus of each candidate direction does not reach the effective projection threshold, the blade surface intrinsic coordinate system is not established, and the corresponding blade surface spatial attitude information is marked as invalid.

[0189] S204: Determine the spatial attitude transformation relationship and correct the geometric response.

[0190] This step is used to determine the conversion relationship between the spatial attitude of the leaf surface and the predetermined leaf surface attitude reference during detection, complete the geometric response correction in the wetting response, and establish the attitude response mapping relationship for S205 to correct the polarization response amplitude.

[0191] According to the leaf surface normal , lighting direction and imaging direction Determine the angle of incidence and observation angle :

[0192] ;

[0193] ;

[0194] in, Used to handle numerical errors that exceed the range The inner product is restricted to this interval; This indicates the angle between the leaf surface normal and the direction from the predetermined detection area to the illumination unit. This indicates the angle between the leaf surface normal and the imaging direction.

[0195] For geometric response correction, based on the calibration parameters of the local leaf surface and the imaging device, the imaging rays corresponding to each image point in the droplet-affected area and contact boundary are intersected with the local leaf surface to obtain the corresponding local spatial points on the leaf surface. For effective depth data that directly corresponds to pixels in the polarized image, the corresponding local spatial points on the leaf surface can also be read.

[0196] For local spatial points on the leaf surface The center position of the local spatial point on the leaf surface determined by S202 As the origin of the tangent plane coordinate system, the coordinates of a local spatial point on the leaf surface on the tangent plane are:

[0197] ;

[0198] ;

[0199] The boundary points corresponding to the influence regions of each detected droplet The coordinates form the influence area of ​​the detected droplet within the tangential plane, and the corrected area of ​​the influence area is determined based on this region. The corrected contact boundary displacement and its direction are determined based on the coordinate changes of the region center or contact boundary within the tangential plane at adjacent effective acquisition times; the corrected expansion direction and expansion distance are determined based on the directional distance from the region center to the contact boundary.

[0200] The area of ​​the droplet's influence region, the displacement of the contact boundary, and the direction of expansion obtained through the above tangent plane transformation have been represented using the leaf's inherent coordinates. For the geometric response where perspective scaling and image orientation differences have been eliminated, the same attitude amplitude correction will not be applied again. Only when calibration results indicate that there are still attitude-related residuals in the corresponding geometric response that can be repeatedly observed will the attitude response mapping relationship be used to correct the residuals.

[0201] Furthermore, attitude response mapping relationships are established in advance. For the geometric or equipment calibration of imaging rays, spatial coordinates, and response non-uniformity, calibration objects with stable dimensions and surface conditions can be used; for the calibration of the amplitude of the wetting polarization response of plant leaves under the action of droplets, a crop reference leaf corresponding to the target crop type and the leaf surface being detected is used.

[0202] When establishing the attitude response mapping relationship corresponding to the wetting polarization response, under the conditions that the physical properties of the detected droplets, the application conditions, the illumination conditions, and the polarization image acquisition conditions are the same or equivalent after calibration, different crop reference leaves or different non-overlapping detection areas of the same crop reference leaf are placed in multiple known leaf surface spatial attitudes, and the wetting polarization response under the corresponding attitude is obtained respectively. For the same detection area, in principle, only one valid detection result is used to establish the mapping relationship to reduce the influence of previously detected droplets on subsequent detection results.

[0203] Using the incident angle, observation angle, and inherent coordinate direction of the leaf surface under each known spatial attitude as attitude input, and the correspondence between the response quantity under the corresponding attitude and the response quantity under the predetermined leaf surface attitude reference as output, an attitude response mapping relationship is established.

[0204] In one specific implementation, for the response quantity that needs to be corrected Based on the results of multiple crop reference observations, representative values ​​of parameters under the predetermined leaf attitude baseline were determined. and the Representative parameter values ​​under a given attitude and determine the first Correction values ​​corresponding to each calibrated attitude: .

[0205] For the current blade space attitude within the effective range of the attitude response mapping relationship, the correction values ​​corresponding to adjacent calibrated attitudes are interpolated to obtain... The response quantity converted to the predetermined blade attitude reference is obtained according to the following formula: ;in, This represents the response quantity under the current leaf surface spatial attitude. This represents the response quantity relative to a predetermined blade attitude reference. For periodic directional quantities, a correction quantity is established using the minimum periodic directional difference, rather than directly performing a standard arithmetic mean on directional values ​​crossing periodic boundaries.

[0206] The calibration attitude interval, parameter representative value determination method, interpolation method, and effective range are determined based on the calibration residual, the measurement uncertainty of leaf spatial attitude information, and the distribution stability of crop reference observation results. When the value exceeds the effective range of the attitude response mapping relationship, no extrapolation correction is performed; instead, the corresponding corrected wetting response or wetting polarization response parameter is marked as invalid.

[0207] This step outputs the corrected geometric response relative to a predetermined blade attitude reference, the incident and observation angles corresponding to the current blade spatial attitude, and the attitude response mapping relationship for use by S205. The final conversion of the polarization response amplitude is performed after S205 obtains the corresponding polarization physical quantities. This processing reduces the influence of differences in the incident and observation angles on the comparison of response amplitudes under different detection conditions.

[0208] S205: Determine the polarization response and complete the polarization reference direction conversion.

[0209] This step is used to solve for the linearly polarized Stokes components based on polarization images with multiple different polarization directions, and to transform the polarization reference direction to the reference direction corresponding to the blade's inherent coordinate system. Then, it combines the attitude response mapping relationship established by S204 to form the corrected wetting response.

[0210] Specifically, multiple polarization images with different polarization directions at the wetting response time constitute a polarization sampling period. When using simultaneous acquisition, each polarization image uses the same acquisition timestamp.

[0211] When using a sequential acquisition method, a common acquisition time is determined for each polarization sampling period after spatial alignment of each polarization image. For the same polarization direction, temporal interpolation is performed based on adjacent valid observations before and after the common acquisition time to obtain the image intensity corresponding to the common acquisition time. If there are no valid observations that can encompass the common acquisition time, or if the interpolation residual exceeds the allowable range determined based on the pre-application polarization response fluctuation and wetting response change rate, the Stokes component is not solved using that polarization sampling period.

[0212] Before calculating the Stokes components, spatial registration is performed on polarization images with different polarization directions, and dark field response, gain difference, and response non-uniformity correction are performed based on the calibration results of polarization image acquisition conditions.

[0213] In one specific implementation, polarization images with polarization directions of 0 degrees, 45 degrees, 90 degrees, and 135 degrees are acquired respectively, and the corresponding corrected image intensities are denoted as... , , and The polarization directions mentioned above are all based on the polarization reference direction used during polarization image acquisition as the 0-degree direction, and are related to the following rotation angles. The same positive angle. The Stokes component of the total intensity is determined according to the following formula. First linearly polarized Stokes component Second linearly polarized Stokes component :

[0214] ;

[0215] ;

[0216] ;

[0217] The above combination of polarization directions is only one specific implementation. When using other numbers or polarization directions of polarized images, for the first... polarization direction Its image intensity satisfies:

[0218] ;

[0219] A system of linear equations is established based on the calibrated polarization directions and corresponding image intensities, and obtained by direct solution or least squares solution. , and .

[0220] The polarization reference direction lies in a plane perpendicular to the imaging direction. The first coordinate axis of the leaf's intrinsic coordinate system is... Projecting onto this plane, we get: .in, The unit imaging direction is from the predetermined detection area to the imaging device. The first coordinate axis Perpendicular to the imaging direction The projection vector on the plane, with the subscript " "Indicates the projection vector and imaging direction" Vertical. When When the modulus satisfies the effective threshold of the direction projection, it is normalized to obtain the transformed polarization reference direction: Using the polarization reference direction used during polarization image acquisition as the starting direction, the converted polarization reference direction... As the termination direction; in a plane perpendicular to the imaging direction, the counterclockwise direction when observing along the imaging light propagation direction is defined as the positive direction, thereby determining the directional rotation angle. .

[0221] The effective threshold for directional projection is determined based on the calibration error of the first coordinate axis and the imaging direction, the fluctuation of the projection modulus when repeatedly determining the projection vector, and the accuracy of numerical calculation, so that the direction obtained after normalizing the projection vector remains stable within the allowable directional error range; the effective threshold for polarization reference directional projection is not preset with a fixed value.

[0222] When the projection magnitude of the first coordinate axis on the plane approaches the effective threshold of the directional projection, the consistency of the projection direction of the blade's inherent coordinate system can be checked using the projection of the second coordinate axis on the plane. Since the mutually orthogonal three-dimensional first and second coordinate axes may not remain orthogonal after being projected onto a plane perpendicular to the imaging direction, the projection direction of the second coordinate axis cannot simply be rotated 90 degrees as the projection direction of the first coordinate axis.

[0223] The equivalent polarization reference direction corresponding to the first coordinate axis is determined only when the equivalent polarization reference direction can be uniquely determined based on the complete three-dimensional coordinate relationship and the projection relationship from the pre-calibrated three-dimensional direction to the imaging plane. The effective projection of the second coordinate axis is used to assist in determining the equivalent polarization reference direction. If the projection of the first coordinate axis is zero or the unique direction transformation relationship cannot be stably determined, the corresponding polarization sampling period is marked as invalid.

[0224] The first linearly polarized Stokes component is calculated according to the following formula. Second linearly polarized Stokes component Transform to the polarization reference direction corresponding to the blade's natural coordinate system:

[0225] ;

[0226] ;

[0227] in, and The Stokes component is the linearly polarized component, represented by the polarization reference direction used during polarization image acquisition. and This represents the linearly polarized Stokes component expressed in terms of the transformed polarization reference direction. The above formula indicates a rotation of the polarization reference coordinates, but does not imply a change in the physical polarization state of the blade surface. By transforming the polarization reference direction, the polarization direction reference difference caused by blade rotation can be reduced.

[0228] The linear polarization intensity is determined based on the converted Stokes component. Linear polarization degree and polarization direction :

[0229] ;

[0230] ;

[0231] ;

[0232] Only The degree of linear polarization is calculated when the intensity exceeds the effective lower limit determined based on the calibration results of dark field noise and polarization image acquisition conditions. If the polarization direction is insufficient to support stable polarization direction calculation, the polarization direction is marked as invalid. Polarization directions are represented and compared according to the period of linear polarization directions, and polarization directions that differ by an integer multiple of 180 degrees are treated as the same direction.

[0233] Under ideal measurement conditions, the linear polarization intensity does not exceed the total intensity. This is in contrast to the effects of measurement noise or channel response errors. When the deviation is within the allowable range of equipment calibration error, it is corrected according to the equipment calibration result; when it exceeds the allowable range, the corresponding polarization response is marked as invalid and no linear polarization degree and polarization direction are generated.

[0234] Will As a measure of specular reflection, As a measure of diffuse reflection, specular reflection is used to characterize the relative change of the linearly polarized component, while diffuse reflection is used to characterize the relative change of the remaining intensity after subtracting the linearly polarized intensity from the total intensity. Since this embodiment does not utilize the circularly polarized Stokes components for complete polarization decomposition, the two measures are not equivalent to the strictly physical separation results of real specularly reflected light and real diffusely reflected light, respectively.

[0235] After completing the polarization reference direction conversion, based on the attitude response mapping relationship established in S204, the effects of the incident angle and observation angle are considered. , , and The response amplitude is converted; the polarization direction is preferentially unified to the blade's inherent coordinate direction through the aforementioned Stokes component coordinate conversion to avoid repeated correction of differences in the same direction.

[0236] Before and after the application of the detection droplet, the illumination conditions and polarization image acquisition conditions were kept consistent, and the results were based on the conversion to a predetermined leaf attitude reference. , , and The corrected wetting response is formed by the temporal variation relative to the pre-application polarization response baseline and combined with the corrected geometric response obtained in S204. The wetting polarization response parameters are determined according to the calculation relationship in Example 1.

[0237] In summary, this embodiment determines the leaf surface normal, establishes the leaf surface inherent coordinate system, and performs leaf surface spatial attitude correction on the geometric and polarization responses in the wetting response. It transforms the wetting responses obtained under different leaf surface spatial attitudes to a predetermined leaf surface attitude reference, thereby obtaining wetting polarization response parameters. This provides a unified parameter basis for comparing the wetting characteristics between the plant leaf to be tested and the crop reference leaf, and helps to reduce the influence of differences in leaf surface spatial attitude on the parameter comparison results.

[0238] Example 3

[0239] In a third embodiment of the present invention, a system is provided for the field plant leaf wettability detection and weed identification method, used to execute the field plant leaf wettability detection and weed identification methods of Embodiments 1 and 2. The system includes a detection droplet application device, a polarization image acquisition device, a leaf surface spatial attitude acquisition device, a synchronization control device, and a data processing device.

[0240] A detection droplet application device includes a liquid storage section, a quantitative output section, and an application end, for applying a detection droplet with known physical properties to a predetermined detection area of ​​a plant leaf to be tested at a determined application time;

[0241] A polarization image acquisition device, including an illumination unit, a polarization optical component and an electronic image sensor, is used to acquire polarization image sequences during the leaf wetting response process before and from the moment of application of the detection droplet.

[0242] A leaf spatial attitude acquisition device, including a depth sensor or a multi-view image acquisition component, is used to acquire leaf spatial attitude information corresponding to a polarized image sequence;

[0243] The synchronous control device is connected to the droplet application device, the polarization image acquisition device, and the leaf spatial attitude acquisition device, respectively. It is used to control the application time and synchronously control the acquisition of polarization image sequence and leaf spatial attitude information based on the application time.

[0244] The data processing device, connected to the polarization image acquisition device, the leaf spatial attitude acquisition device, and the synchronization control device, includes a processor and a memory. The memory stores crop wetting characteristic reference information corresponding to the physical properties of the target crop, the detected droplets, the application conditions, and the predetermined leaf attitude reference. This crop wetting characteristic reference information includes the reference range and deviation judgment conditions for the wetting polarization response parameters. The processor is used for:

[0245] The baseline of the pre-polarization response is determined based on the pre-polarization image, and the wetting response is determined based on the temporal variation of the post-polarization image relative to the pre-polarization response baseline.

[0246] Based on the leaf spatial attitude information, the wetting response is corrected to the leaf spatial attitude reference and converted to a predetermined leaf attitude reference. The wetting polarization response parameters are then determined based on the corrected wetting response.

[0247] By comparing the wetting polarization response parameters with crop wetting characteristic reference information, we can obtain the weed judgment result based on whether the leaves of the plant to be tested meet the deviation judgment condition, and the leaves of the plant to be tested are judged as weeds when the deviation judgment condition is met.

[0248] In summary, this embodiment, through the aforementioned physical device and its connection and control relationship, realizes the controlled application of the detected droplet, the synchronous acquisition of polarization images and leaf surface spatial attitude information, and the processing and comparison of wetting polarization response parameters, providing a corresponding system implementation basis for the methods of Embodiment 1 and Embodiment 2.

[0249] It should be noted that 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 preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for detecting the leaf wettability characteristics of plants and identifying weeds in a field, used to determine the characteristics of leaves of plants to be tested in a field where a target crop is grown, characterized in that... Includes the following steps: The target crop type is determined and the physical properties of the detection droplets are obtained. The detection droplet application device is controlled to apply the detection droplets to the predetermined detection area of ​​the leaf of the plant to be tested at a determined application time. Using the application time as a time reference, the polarization image sequence and corresponding leaf spatial attitude information of the plant leaf before the application of the detection droplet and from the application time are obtained according to the acquisition time sequence. The pre-application polarization response baseline is determined based on the pre-application polarization image in the polarization image sequence, and the wetting response is determined based on the temporal change of the post-application polarization image relative to the pre-application polarization response baseline. The wetting response is corrected based on the leaf spatial attitude information, and converted to a predetermined leaf attitude reference. The wetting polarization response parameters are then determined based on the corrected wetting response. Obtain crop wetting characteristic reference information corresponding to the target crop. The crop wetting characteristic reference information is obtained by detecting crop reference leaves or pre-stored. It corresponds to the physical properties of the detected droplets, the application conditions and the predetermined leaf attitude reference, and includes the reference range and deviation judgment conditions of the wetting polarization response parameters of the target crop. The deviation judgment conditions specify the combination relationship between one or more of the wetting polarization response parameters and the corresponding reference range. The wetting polarization response parameters of the plant leaves to be tested are compared with the crop wetting characteristic reference information. The weed determination result is obtained according to whether the plant leaves to be tested meet the deviation determination condition. When the deviation determination condition is met, the plant leaves to be tested are determined to be weeds in the target crop planting area.

2. The method for detecting leaf wettability characteristics of plants and identifying weeds in the field according to claim 1, characterized in that: Determining the wetting response and the wetting polarization response parameters includes: The detection droplet's influence area and its contact boundary are determined from the polarization image sequence. The contact boundary is the boundary of the contact area between the detection droplet and the plant leaf to be detected. According to the acquisition sequence, a geometric response sequence is formed based on the area change of the detection droplet's influence area, the displacement change of the contact boundary, or the change of the expansion direction. A polarization change sequence is formed based on the change in polarization response within the region affected by the detected droplet relative to the baseline of the polarization response before application; Based on the application time and the acquisition time of each polarization image, the geometric response sequence and the polarization change sequence are time-aligned to determine the time correspondence between the motion state change of the contact boundary and the polarization response change. The geometric response sequence, the polarization change sequence, and the time correspondence are used as the wetting response, and the wetting polarization response parameters are determined based on the corrected wetting response. The wetting polarization response parameters include wetting geometric parameters determined by the geometric response sequence, polarization change parameters determined by the polarization change sequence, and coupling timing parameters determined by the time correspondence.

3. The method for detecting leaf wettability characteristics of plants and identifying weeds in the field according to claim 2, characterized in that: Based on the area change of the affected area of ​​the detected droplet, the displacement change of the contact boundary, and the polarization change sequence, the expansion stage of the detected droplet is identified, and at least one subsequent stage is identified from the boundary pinning stage, holding stage, rolling stage, shrinkage stage, and leaf surface polarization response recovery stage. The wetting polarization response parameters include at least one of the following: duration of the expansion phase, expansion rate, anisotropy of the expansion direction, duration of boundary pinning, duration of holding, roll-off initiation time, contraction rate, peak polarization response time, polarization response recovery time, and the time difference between the start of contact boundary state change and polarization response state change.

4. The method for detecting leaf wettability characteristics of plants and identifying weeds in the field according to claim 1, characterized in that: Obtaining the leaf surface spatial attitude information includes: Acquire depth data or multi-view images of the plant leaf before the application of the detection droplet, and perform spatial registration and acquisition timing association with the polarization image sequence; Fit the local surface of the leaf surface based on the local spatial points of the predetermined detection area and its neighborhood, and determine the leaf surface normal. The spatial attitude relationship of the predetermined detection area relative to the gravity direction, illumination direction and imaging direction is determined based on the leaf surface normal, and the leaf surface normal and the spatial attitude relationship are used as the leaf surface spatial attitude information.

5. The method for detecting leaf wettability characteristics of plants and identifying weeds in the field according to claim 4, characterized in that: The leaf surface spatial attitude correction includes: Using the leaf surface normal as the third coordinate axis, and following the order of the illumination direction, the imaging direction, and the horizontal axis of the image coordinate system of the polarized image, the first projection with a projection modulus greater than the effective projection threshold on the tangent plane of the local surface of the leaf is selected, and its normalized direction is used as the first coordinate axis. The cross product direction of the third coordinate axis and the first axis is used as the second coordinate axis to establish a right-handed leaf surface intrinsic coordinate system. The direction transformation relationship between the polarization reference direction used for polarization image acquisition and the projection direction of the first coordinate axis on a plane perpendicular to the imaging direction is determined, and the incident angle and observation angle of the predetermined detection area in the leaf surface inherent coordinate system are determined; the predetermined leaf surface attitude reference includes the predetermined reference incident angle, reference observation angle and leaf surface inherent coordinate direction; The polarization direction component in the wetting response is transformed according to the direction transformation relationship to determine the detection droplet influence area and its contact boundary in the polarization image, and the detection droplet influence area and its contact boundary are projected onto the tangent plane; the wetting response after coordinate transformation and projection is transformed to the predetermined leaf attitude reference according to the incident angle, the observation angle and the pre-calibrated attitude response mapping relationship.

6. The method for detecting leaf wettability characteristics of plants and identifying weeds in the field according to claim 5, characterized in that: Multiple polarization images with different polarization directions are acquired in each polarization sampling period, and the total intensity Stokes component is determined based on the intensity of each polarization image. First linearly polarized Stokes component Second linearly polarized Stokes component ; The angle by which the polarization reference direction is rotated in a plane perpendicular to the imaging direction to the projection direction of the first coordinate axis on that plane is denoted as the rotation angle. The rotation angle is defined as the counterclockwise direction when viewed along the direction of the imaging light propagation. The positive direction, and the first linearly polarized Stokes component according to the following formula. and the second linearly polarized Stokes component Perform coordinate transformation: ； ； Determine the linear polarization intensity based on the converted Stokes component. Linear polarization degree and polarization direction ,in: ； ； ； respectively and As a measure of specular reflection and diffuse reflection, and based on the application of the detection droplet before and after... , , and The change determines the wetting polarization response parameters.

7. The method for detecting leaf wettability characteristics of plants and identifying weeds in the field according to claim 5, characterized in that: The crop wetting characteristic reference information is divided into multiple attitude reference zones according to the leaf surface normal, leaf surface tilt angle, incident angle or observation angle of the crop reference leaf before leaf surface spatial attitude correction. The wetting polarization response parameters in each attitude reference zone are converted to the same predetermined leaf surface attitude reference. The corresponding attitude reference partition is determined based on the leaf surface spatial attitude information of the plant leaf to be tested, and the wetting polarization response parameters of the plant leaf to be tested are compared with the crop wetting characteristic reference information in the attitude reference partition.

8. The method for detecting leaf wettability characteristics of plants and identifying weeds in the field according to claim 7, characterized in that: The crop wetting characteristic reference information is established through the following steps: Select a crop reference leaf that is the same as the target crop type and growth stage. The tested leaf surface of the crop reference leaf and the tested leaf surface of the plant to be tested are both paraxial or both abaxial. The crop reference leaf was tested using the same detection droplets with the same physical properties as those used when testing the plant leaf to be tested, and with the same application control method and polarization image acquisition conditions. Leaf spatial attitude correction is performed on the wetting response of the crop reference leaf. According to the attitude reference partition, the corresponding reference range is determined based on the distribution of wetting polarization response parameters in each partition. The deviation judgment condition is established based on the combination relationship between one or more wetting polarization response parameters and the corresponding reference range.

9. The method for detecting leaf wettability characteristics of plants and identifying weeds in the field according to claim 1, characterized in that: The wetting polarization response parameters are divided into a first response parameter group and a second response parameter group; The first set of response parameters includes parameters characterizing the processes of droplet expansion, contact boundary displacement, boundary pinning, holding, rolling off, or contraction. The second set of response parameters includes parameters characterizing polarization intensity, degree of linear polarization, polarization direction, specular reflection characterization, diffuse reflection characterization, or leaf surface polarization response recovery process. The crop wetting characteristic reference information includes the reference range corresponding to each parameter in the first response parameter group and the second response parameter group; The deviation determination condition is: there is at least one parameter in the first response parameter group that exceeds the corresponding reference range, and there is at least one parameter in the second response parameter group that exceeds the corresponding reference range; When the deviation determination condition is met, a weed determination result is obtained indicating that the leaf of the plant to be detected is a weed; otherwise, a weed determination result is obtained indicating that the leaf of the plant to be detected does not meet the deviation determination condition.

10. The system used in the field plant leaf wettability detection and weed identification method according to any one of claims 1 to 9, for identifying the leaves of the plant to be tested in a field where the target crop is planted, characterized in that, include: A detection droplet application device includes a liquid storage section, a quantitative output section, and an application end, for applying a detection droplet with known physical properties to a predetermined detection area of ​​the leaf of the plant to be tested at a determined application time; A polarization image acquisition device, comprising an illumination unit, a polarization optical component, and an electronic image sensor, is used to acquire a sequence of polarization images during the leaf wetting response process before and from the moment of application of the detection droplet. A leaf spatial attitude acquisition device, including a depth sensor or a multi-view image acquisition component, is used to acquire leaf spatial attitude information corresponding to the polarization image sequence; A synchronous control device is connected to the droplet detection application device, the polarization image acquisition device, and the leaf surface spatial attitude acquisition device, respectively, and is used to control the application time and synchronously control the acquisition of the polarization image sequence and the leaf surface spatial attitude information with the application time as a time reference. A data processing device, connected to the polarization image acquisition device, the leaf spatial attitude acquisition device, and the synchronization control device, includes a processor and a memory. The memory stores crop wetting characteristic reference information corresponding to the physical properties, application conditions, and predetermined leaf attitude reference of the target crop and the detected droplet. The crop wetting characteristic reference information includes the reference range and deviation judgment conditions of the wetting polarization response parameters. The processor is used for: The pre-application polarization response baseline is determined based on the pre-application polarization image, and the wetting response is determined based on the temporal variation of the post-application polarization image relative to the pre-application polarization response baseline. The wetting response is corrected based on the leaf spatial attitude information, and then converted to the predetermined leaf attitude reference. The wetting polarization response parameters are determined based on the corrected wetting response. The wetting polarization response parameters are compared with the crop wetting characteristic reference information. The weed determination result is obtained based on whether the leaf of the plant to be detected meets the deviation determination condition. If the deviation determination condition is met, the leaf of the plant to be detected is determined to be a weed.

Images