A rice machine transplanting operation progress statistics method and system

Abstract

The application provides a rice machine transplanting operation progress statistical method and system, through obtaining operation data of a rice transplanter terminal device, and cleaning the operation data; according to the cleaned operation track, the area of an irregular land boundary is calculated, and the land boundary is output as vector data; according to topological analysis, the vector data is associated with corresponding attribute information to form a structured spatial database; according to the vector data in the structured spatial database, image and near-infrared band data of each part of the farmland are obtained, superimposed, then the superimposed results of each part of the farmland are fused and spliced to obtain a target farmland map; according to a deep learning algorithm, rice detection and identification are performed on the target farmland map to determine a current planting area; according to a historical rice planting area, the current planting area and a time period, the transplanting progress is counted and visualized, and the real-time counting of the rice transplanting operation progress is realized.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of rice machine transplanting operation progress statistics, and particularly relates to a rice machine transplanting operation progress statistics method and system. BACKGROUND

[0002] Rice is an important crop, and real-time acquisition of rice planting area and planting progress is very important for the next agricultural decision-making. The traditional rice planting area and progress statistics method relies on manual investigation and statistics, which is inefficient and prone to errors, and cannot meet the real-time and accuracy requirements of modern agriculture. SUMMARY

[0003] Therefore, the present application provides a rice machine transplanting operation progress statistics method and system, which aims to real-time statistics of rice transplanting operation progress.

[0004] The first aspect of the present application provides a rice machine transplanting operation progress statistics method, which is applied to a scenario with a rice transplanter terminal device, and the method comprises:

[0005] Obtaining operation data of the rice transplanter terminal device, the operation data at least including operation position, operation time and operation trajectory, and cleaning the operation data;

[0006] According to the cleaned operation trajectory, the area of irregular land boundary is calculated, and the land boundary is output as vector data;

[0007] According to the topological analysis, the vector data is associated with the corresponding attribute information to form a structured spatial database, wherein the attribute information at least includes administrative division information and land contractor information;

[0008] According to the vector data in the structured spatial database, the image and near-infrared band data of each part of the farmland are obtained and superimposed, and then the superimposed results of each part of the farmland are fused and spliced to obtain a target farmland map;

[0009] According to the deep learning algorithm, the target farmland map is detected and identified to determine the current planting area;

[0010] According to the historical rice planting area, the current planting area and the time period, the transplanting progress is counted and visualized.

[0011] Further, in the step of cleaning the operation trajectory, the operation trajectory is processed by point-by-point smoothing drift, and the broken part is filled to obtain the cleaned operation trajectory, so as to ensure the continuity of the operation trajectory.

[0012] Further, the step of performing point-by-point smoothing drift processing on the work trajectory comprises:

[0013] Obtaining a trajectory point of the work trajectory, and predicting a trajectory point at a current time according to a trajectory point at a previous time and a dynamic model;

[0014] Fusing the actual trajectory point with the predicted trajectory point at the current time, and calculating a Kalman gain according to a predicted state covariance matrix;

[0015] Calculating a difference between the actual trajectory point and the predicted trajectory point at the current time, and performing weighted correction on the predicted trajectory point at the current time by using the Kalman gain.

[0016] Further, the step of filling the break comprises:

[0017] Obtaining a trajectory point within a preset distance of the break, and determining whether it is a uniform motion scene;

[0018] If it is determined that it is a uniform motion scene, performing linear interpolation according to the trajectory point at the break to fill the break;

[0019] If it is determined that it is not a uniform motion scene, performing polynomial interpolation according to the trajectory point within the preset distance of the break to fill the break.

[0020] Further, the step of obtaining image and near-infrared band data of each part of the farmland according to vector data in the structured spatial database, performing superposition, then fusing and splicing the superposition results of each part of the farmland to obtain a target farmland map comprises:

[0021] Obtaining images of the same area, determining R, G and B channels in the images, and performing band synthesis on the R, G and B channels and the near-infrared band data in ENVI or QGIS to generate four-channel remote sensing images;

[0022] Applying an NDVI formula to the four-channel remote sensing images through a raster calculator to output a single-band image with a range of [-1, 1];

[0023] According to the NDVI result, performing density segmentation on the area and performing false color rendering to determine the superposition result;

[0024] Performing Brovey transformation on the superposition result and then splicing to obtain a target farmland map.

[0025] Further, the step of obtaining image and near-infrared band data of each part of the farmland according to vector data in the structured spatial database, performing superposition, then fusing and splicing the superposition results of each part of the farmland to obtain a target farmland map comprises:

[0026] Obtain the target farmland map and the NDVI result, set an initial segmentation scale parameter, and generate image objects of different scales in combination with spectral heterogeneity and spatial heterogeneity;

[0027] Calculate spectral features, geometric features, and texture features for each image object, and extract a target farmland region according to a preset rule set;

[0028] Vectorize the target farmland region to generate a two-dimensional boundary;

[0029] Convert the two-dimensional boundary into a three-dimensional boundary according to elevation data, and optimize the three-dimensional boundary according to terrain features.

[0030] Further, the deep learning algorithm is an improved YOLOv5 algorithm, which is optimized by transfer learning for rice seedling morphological features, wherein the training set contains labeled samples under different growth stages and different light conditions.

[0031] A second aspect of the embodiment of the application provides a rice machine transplanting operation progress statistical system for implementing the rice machine transplanting operation progress statistical method of the first aspect, and the system comprises:

[0032] An acquisition module is configured to acquire operation data of a rice transplanter terminal device, wherein the operation data at least includes an operation position, an operation time, and an operation trajectory, and the operation data is cleaned;

[0033] A calculation module is configured to calculate an area of an irregular land boundary according to the cleaned operation trajectory, and output a land boundary as vector data;

[0034] An association module is configured to associate the vector data with corresponding attribute information to form a structured spatial database according to topological analysis, wherein the attribute information at least includes administrative division information and land contractor information;

[0035] A superposition module is configured to acquire image and near-infrared band data of each part of a farmland according to the vector data in the structured spatial database, superimpose the data, and then fuse and splice superimposition results of each part of the farmland to obtain a target farmland map;

[0036] An identification module is configured to perform rice detection and identification on the target farmland map according to a deep learning algorithm to determine a current planting area;

[0037] A statistical module is configured to statistically analyze a transplanting progress according to a historical rice planting area, the current planting area, and a time period, and visualize the progress.

[0038] The third aspect of the embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the program is executed by a processor to realize the rice machine transplanting operation progress statistical method provided in the first aspect.

[0039] The fourth aspect of the embodiment of the present application provides an electronic device, which comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor executes the program to realize the rice machine transplanting operation progress statistical method provided in the first aspect.

[0040] The rice machine transplanting operation progress statistical method and system provided in the embodiment of the present application, by acquiring the operation data of the rice transplanter terminal device, the operation data at least comprising the operation position, the operation time and the operation trajectory, and cleaning the operation data; according to the cleaned operation trajectory, the area of the irregular land boundary is calculated, and the land boundary is output as vector data; according to the topological analysis, the vector data is associated with the corresponding attribute information to form a structured spatial database, wherein the attribute information at least comprises administrative division information and land contractor information; according to the vector data in the structured spatial database, the image and near-infrared band data of each part of the farmland are acquired and superimposed, and then the superimposed results of each part of the farmland are fused and spliced to obtain a target farmland map; according to the deep learning algorithm, the target farmland map is detected and recognized to determine the current planting area; according to the historical rice planting area, the current planting area and the time period, the transplanting progress is counted and visualized, realizing the real-time counting of the rice transplanting operation progress, and providing accurate data support for agricultural decision-making. BRIEF DESCRIPTION OF DRAWINGS

[0041] Figure 1 The implementation flowchart of the rice machine transplanting operation progress statistical method provided for the embodiment one of the present application;

[0042] Figure 2 The structural block diagram of the rice machine transplanting operation progress statistical system provided for the embodiment two of the present application;

[0043] Figure 3 The structural block diagram of the electronic device provided for the embodiment three of the present application. DETAILED DESCRIPTION

[0044] In order to facilitate the understanding of the present application, the present application will be described more fully below with reference to the related drawings. The drawings show several embodiments of the present application. However, the present application can be realized in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present application more thorough and comprehensive.

[0045] It is to be understood that where an element such as a layer, region or substrate is described as being "on" another element, it can be directly on the other element or intervening elements can also be present. Where an element is described as being "connected" or "coupled" to another element, it can be directly connected or coupled or intervening elements can be present. As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0046] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the description herein is for describing particular embodiments only and is not intended to be limiting of the application. As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0047] Embodiment one

[0048] According to the embodiment of the present application, a rice machine transplanting operation progress statistical method embodiment is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that shown herein.

[0049] In this embodiment one, a rice machine transplanting operation progress statistical method is provided, which is applied to a scenario with a rice transplanter terminal device and can be used in an electronic device such as a computer. Please refer to Figure 1 , Figure 1 The implementation flowchart of the rice machine transplanting operation progress statistical method provided by the embodiment one of the present application is shown, which specifically includes steps S01 to S06.

[0050] In step S01, the operation data of the rice transplanter terminal device is acquired, and the operation data at least includes operation position, operation time and operation trajectory, and the operation data is cleaned.

[0051] Specifically, the rice transplanter terminal device can collect operation data in real time. In addition, the rice transplanter terminal device can collect GPS coordinates at a second level interval, and transmit the data to the cloud server in real time through 4G / 5G network or LoRa Internet of Things protocol. Further, the collected operation position, operation time and operation trajectory and other data are cleaned to remove abnormal and error data, so as to ensure the accuracy and reliability of the data.

[0052] For the operation trajectory, point-by-point smoothing drift processing is performed, and through filling the broken processing, the cleaned operation trajectory is obtained to ensure the continuity of the operation trajectory. It should be noted that the trajectory point of the operation trajectory is obtained, and the trajectory point at the current time is predicted according to the trajectory point at the last time and the dynamic model, which can be represented as:

[0053] ;

[0054] wherein, is the optimal estimation state at k-1 time, that is, the observation data at the previous time is fused, is the predicted state at the current time (k time) based on the state at the previous time (k-1 time), F is a state transition matrix, which encodes the motion law of the target (such as uniform speed, uniform acceleration), B is a control matrix, is an external control amount;

[0055] The actual trajectory point and the predicted trajectory point at the current time are fused, and the Kalman gain is calculated according to the predicted state covariance matrix, which can be represented as:

[0056] ;

[0057] ;

[0058] wherein, is the residual, that is, the difference between the actual trajectory point and the predicted trajectory point at the current time, is the GPS observation value, is an observation matrix, is the transpose of the observation matrix, is the Kalman gain, is an observation noise covariance matrix, is the error covariance of the predicted state value at the current time (k time) based on the state at the previous time (k-1 time);

[0059] The difference between the actual trajectory point and the predicted trajectory point at the current time is calculated, and the predicted trajectory point at the current time is weighted and corrected by using the Kalman gain, which can be represented as:

[0060] ;

[0061] ;

[0062] wherein, is an identity matrix, is the state estimation value after correction at the current time (k time), is the error covariance of the state estimation value after correction at the current time (k time), Error covariance of a predicted state value of a current time (k time) based on a state of a previous time (k-1 time).

[0063] It can be understood that the above process is repeated to filter each time point of the trajectory point, and smooth the drifting signal (such as an abnormally jumping coordinate point).

[0064] Further, the step of filling the broken part comprises:

[0065] Obtaining the trajectory points within a preset distance of the broken part, and judging whether it is a uniform motion scene. It can be understood that according to the acquisition time and interval distance of the trajectory points, the scene can be judged.

[0066] If it is judged to be a uniform motion scene, then linear interpolation is performed according to the trajectory points of the broken part to fill the broken part. Specifically, the missing point time is t m , the front and rear effective points are (x1, y1, t1) and (x2, y2, t2), and then:

[0067] ;

[0068] ;

[0069] Obtaining the trajectory points (x m , y m , t m ) of the broken part;

[0070] If it is judged to be a non-uniform motion scene, then polynomial interpolation is performed according to the trajectory points within a preset distance of the broken part to fill the broken part. In this way, a polynomial function is used to fit the trajectory curve, a polynomial equation is constructed through multiple adjacent points, and the missing point is solved. For example, a cubic polynomial interpolation requires at least 4 effective points to construct a cubic function , and the missing point coordinates are calculated after fitting the curve by the least square method. a, b, c, and d are coefficients of the cubic function.

[0071] In some other embodiments of the application, the time interval and the historical speed mode can also be combined to assign weights to adjacent points, so as to more accurately estimate the missing point. For example, the weight of the adjacent point is higher, or the weight is adjusted according to the historical speed variance, such as:

[0072] ;

[0073] wherein, is the speed variance, reflecting the motion stability, w i is the weight of the i-th adjacent point, t m is the missing point time, t i is the time of the i-th adjacent point.

[0074] Or, if the trajectory matches the electronic map, the road topological structure constraint can be used to interpolate the path to avoid generating unreasonable "through the building" trajectory. It can be understood that if the electronic map road of rice seedling transplanting is set in advance, the effective points are matched to the map road, the missing points are interpolated along the road center line, and the trajectory is ensured to comply with the actual traffic rules.

[0075] In step S02, the area of the irregular land boundary is calculated according to the cleaned operation trajectory, and the land boundary is output as vector data.

[0076] It should be noted that the operation trajectory is obtained by obtaining the historical operation data of the rice transplanter terminal device, and the overall operation area and the operation area can be determined according to the operation trajectory and the operation area of the rice transplanter terminal device per unit time. Finally, the preliminary land boundary is determined according to the operation area, so as to provide a basis for the subsequent superposition operation of the image and near-infrared band data of each part of the farmland.

[0077] In step S03, the vector data is associated with corresponding attribute information according to topological analysis to form a structured spatial database, wherein the attribute information at least includes administrative division information and land contractor information.

[0078] In the embodiment of the application, the topological tool of the GIS software (such as ArcGIS, QGIS) is used to verify the land boundary vector data. For example, whether the polygon is closed, whether there is a hanging node or a pseudo node is checked; overlapping areas (such as the intersection of adjacent land boundaries) or gaps (such as the existence of no data area between lands) are identified and repaired. Then, the land vector data is associated with the administrative boundary (such as the village-level administrative boundary) through spatial connection (SpatialJoin).

[0079] In step S04, the image and near-infrared band data of each part of the farmland are obtained according to the vector data in the structured spatial database, and are superimposed, and then the superposition results of each part of the farmland are fused and spliced to obtain a target farmland map.

[0080] Specifically, when the rice seedling progress needs to be counted, the image of the same area is obtained according to the vector data in the structured spatial database, i.e. according to the preliminarily determined land boundary, the R, G and B channels in the image are determined, and the R, G and B channels are combined with the near-infrared band data in ENVI (professional remote sensing software) or QGIS (open source GIS tool) to generate a four-channel remote sensing image, which provides a data basis for NDVI calculation.

[0081] The four-channel remote sensing image is applied with the NDVI formula through the grid calculator, and a single-band image with a range of [-1, 1] is output, wherein, based on the synthesized four-channel data, the NDVI formula is used to amplify the spectral difference between vegetation and non-vegetation (the vegetation has high reflectivity in the near-infrared band and low reflectivity in the red band), and the spectral information is converted into a numerical vegetation index, which provides a basis for threshold segmentation, that is, density segmentation. Specifically, the NDVI formula is:

[0082]

[0083] NIR is the reflectivity in the near-infrared band, and Red is the reflectivity in the red band. It can be understood that when the NDVI value is close to 1, it indicates that the vegetation is dense (such as farmland crops); when the NDVI value is close to 0, it indicates non-vegetation (such as bare land, buildings) or low vegetation coverage; and when the NDVI value is negative, it indicates water or no data area (which needs to be removed by a mask);

[0084] According to the NDVI result, the region is subjected to density segmentation, and false color rendering is performed to determine the superimposed result. It should be noted that based on the NDVI statistical histogram, the threshold is set in combination with the actual situation of the study area. For example, the NDVI of farmland crops (such as rice and wheat) during the vigorous growth period is usually greater than or equal to 0.3, which can be set as a high vegetation area; the NDVI of bare land, roads and the like is usually less than or equal to 0.1, which can be set as a non-vegetation area; and the intermediate interval (0.1-0.2) can be regarded as a low vegetation or transition area, which needs to be verified manually. In addition, in the false color rendering, the high sensitivity of the near-infrared band to vegetation is utilized to map it to the red channel, thereby enhancing the visual performance of the vegetation. Specifically, in ENVI, the bands are combined, that is, the red channel is combined with the near-infrared band, the green channel is combined with the red band, and the blue channel is combined with the green band. Linear stretching (such as 2% linear stretching) is applied to enhance the contrast, so that the vegetation is displayed as bright red, and the non-vegetation (such as roads and buildings) is displayed as blue / cyan;

[0085] The superimposed result is subjected to Brovey transformation and then spliced to obtain a target farmland map, wherein the Brovey transformation is to select the “image fusion” tool in Global Mapper, to perform weighted fusion of the near-infrared band as the brightness component and the visible light three channels, and to improve the definition of the vegetation boundary.

[0086] In some other embodiments of the present application, in order to avoid the area calculation deviation caused by the traditional two-dimensional image splicing, specifically, the target farmland map and the NDVI result are obtained, an initial segmentation scale parameter (such as 20-30 pixels for terrace areas and 10-15 pixels for gully areas) is set, and the spectral heterogeneity (such as the NDVI value difference) and the spatial heterogeneity (such as the texture roughness) are combined to generate image objects of different scales. ​

[0087] Calculate the spectral features, geometric features and texture features for each image object, and extract the target farmland region according to the preset rule set, wherein the spectral features include mean and variance, the geometric features include area, perimeter and compactness, and the texture features are GLCM gray level co-occurrence matrix, and the preset rule set can be NDVI≥0.2, compactness≥0.6 and area≥50 square meters;

[0088] Vectorize the target farmland region to generate a two-dimensional boundary, specifically, manually or automatically vectorize the segmented farmland region in Global Mapper to generate a two-dimensional plot boundary (only containing X and Y coordinates);

[0089] According to the elevation data, the two-dimensional boundary is converted into a three-dimensional boundary, specifically, the two-dimensional boundary vector is imported into QGIS, the elevation value (Z coordinate) of each boundary point is obtained from the DEM through the "value extraction by position" tool, a three-dimensional vector (X, Y, Z) is generated, and the three-dimensional boundary is optimized according to the terrain characteristics, it should be noted that for terrace area, the boundary is smoothed along the contour direction (Z value is similar) to avoid "staircase" error caused by two-dimensional splicing; for gully area, use DEM slope analysis (Slope Analysis) to encrypt boundary control points at slope mutation places (such as ≥20°) to ensure that the boundary fits the actual terrain.

[0090] Step S05, according to the deep learning algorithm, the target farmland map is detected and identified, and the current planting area is determined.

[0091] Among them, the improved YOLOv5 algorithm is used, the morphological characteristics of rice seedlings are optimized by transfer learning, and the training set contains labeled samples under different growth stages (3-5 leaf stage) and different light conditions. Specifically, the model will identify each rice target in the spliced image and frame its range with a rectangular box. In the identification process, a dynamic threshold adjustment mechanism is introduced: according to the agronomic requirements of rice varieties (such as hybrid rice 30cm×20cm, conventional rice 25cm×16cm), the rectangular box spacing threshold is automatically adjusted to realize accurate determination of planting density.

[0092] In the embodiments of the present application, the input of the improved YOLOv5 algorithm is a four-channel remote sensing image and an NDVI feature map; the backbone network is CSPDarknet+CBAM attention, CBAM (Convolutional Block Attention Module) is embedded in the CSP module to enhance the feature attention of the texture of rice leaves and the structure of stems, and to suppress noise such as water field reflection and weeds; the neck network adopts FPN+PAN+ASPP multi-scale fusion, on the basis of FPN (top-down path), the fusion of bottom details (such as seedling edges) and high-level semantics (such as “rice” category) is enhanced through PAN (bottom-up path), and the detection accuracy of dense small targets is improved, ASPP (Atrous Spatial Pyramid Pooling) is introduced in the PAN stage, and features with different dilution rates are sampled in parallel to adapt to the scale changes of rice plants at different growth stages (such as small plants at the seedling stage and large plants at the mature stage); the head network includes a multi-task detection head, which is a target classification branch, a growth stage regression branch and a density confidence branch, the target classification branch is used to distinguish rice and non-rice (such as weeds and sundries), the growth stage regression branch is used to predict the leaf period (3-5 leaves) of seedlings, and the density confidence branch is used to output the plant density level (high / medium / low) to assist subsequent planting spacing analysis; a joint loss function is designed in the training process, which is represented as:

[0093] ;

[0094] wherein, is a classification loss (cross entropy) for distinguishing rice and background, is a bounding box regression loss (CIoU loss) for optimizing the positioning accuracy of the rectangular frame, is a growth stage regression loss (MSE loss) for predicting the leaf period continuous value, is a density classification loss (FocalLoss), λ1, λ2 and λ3 are corresponding weights, and the imbalance problem of positive and negative samples in dense areas is alleviated.

[0095] At the same time, according to a certain threshold (such as an interval of 5cm*5cm), the gap area between the rectangular frames of rice is calculated, if there is a gap, it means that the distribution of rice planting is unreasonable, and the land planting area is wasted, combined with the farmland boundary area provided by the previous algorithm calculation, the operation progress and the standard rate can be further calculated. The position coordinates of the continuous gap area can also be marked to generate a “to-be-supplemented area heat map” to provide navigation basis for subsequent artificial seedling supplement. For example, the distribution of rice plants can be identified to calculate the area of the inserted seedlings and the area of the non-inserted seedlings, and then the progress and standard rate of the seedling insertion operation are obtained.

[0096] Step S06, according to the historical rice planting area, the current planting area and the time period, the transplanting progress is counted and visualized.

[0097] A transplanting progress statistical model is constructed, which compares the current planting area and the time period with the historical rice planting area to count the transplanting progress. In addition, through GIS technology, the system can visualize the transplanting progress data, so that the agricultural decision maker can intuitively understand the progress of rice planting.

[0098] After matching, the results are reconciled through field verification and related data provided by a third party to ensure the accuracy of the statistical results. According to the verification results, the parameters are optimized, for example, by adjusting the parameters of the transplanting progress statistical model, optimizing the data storage and retrieval strategy, to ensure that it can adapt to different scales and complexities of the transplanting operation data. This optimization process not only improves the statistical accuracy, but also further improves the overall efficiency of the system.

[0099] In summary, the rice machine transplanting operation progress statistical method in the above embodiments of the present application, the method obtains the operation data of the rice transplanter terminal device, the operation data at least includes operation position, operation time and operation trajectory, and the operation data is cleaned; according to the cleaned operation trajectory, the area of the irregular land boundary is calculated, and the land boundary is output as vector data; according to the topological analysis, the vector data is associated with the corresponding attribute information to form a structured spatial database, wherein the attribute information at least includes administrative division information and land contractor information; according to the vector data in the structured spatial database, the image and near-infrared band data of each part of the farmland are obtained, and are superimposed, then the superimposed results of each part of the farmland are fused and spliced to obtain a target farmland map; according to the deep learning algorithm, the target farmland map is detected and recognized to determine the current planting area; according to the historical rice planting area, the current planting area and the time period, the transplanting progress is counted and visualized, realizing the real-time statistics of the rice transplanting operation progress, and providing accurate data support for agricultural decision-making.

[0100] Embodiment two

[0101] Please refer to Figure 2 , Figure 2 is a structural block diagram of a rice machine transplanting operation progress statistical system provided by the embodiment two of the present application, the rice machine transplanting operation progress statistical system 200 is used to realize the above embodiments and preferred embodiments, and the repeated description is not described. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, hardware, or a combination of software and hardware implementation is also possible and conceived.

[0102] Specifically, the rice machine transplanting operation progress statistical system 200 comprises an acquisition module 21, a calculation module 22, an association module 23, a superposition module 24, an identification module 25 and a statistical module 26, wherein:

[0103] The acquisition module 21 is configured to acquire operation data of a rice transplanter terminal device, wherein the operation data at least comprises an operation position, an operation time and an operation trajectory, and clean the operation data; in the step of cleaning the operation trajectory, the operation trajectory is subjected to a point-by-point smoothing drift processing and a broken gap filling processing to obtain cleaned operation trajectory, so as to ensure the continuity of the operation trajectory.

[0104] The calculation module 22 is configured to calculate the area of an irregular land boundary according to the cleaned operation trajectory, and output a land boundary as vector data.

[0105] The association module 23 is configured to associate the vector data with corresponding attribute information according to a topological analysis to form a structured spatial database, wherein the attribute information at least comprises administrative division information and land contractor information.

[0106] The superposition module 24 is configured to acquire image and near-infrared band data of each part of a farmland according to the vector data in the structured spatial database, and perform superposition, and then fuse and splice the superposition results of each part of the farmland to obtain a target farmland map.

[0107] The identification module 25 is configured to perform rice detection and identification on the target farmland map according to a deep learning algorithm to determine a current planting area, wherein the deep learning algorithm is an improved YOLOv5 algorithm, and the training set contains labeled samples under different growth stages and different light conditions.

[0108] The statistical module 26 is configured to statistically analyze the transplanting progress according to a historical rice planting area, a current planting area and a time period, and perform visualization.

[0109] Further, in some optional embodiments of the present application, the acquisition module 21 comprises:

[0110] An acquisition unit is configured to acquire a trajectory point of the operation trajectory, and predict a trajectory point at a current time according to a trajectory point at a previous time and a dynamic model.

[0111] A fusion unit is configured to fuse an actual trajectory point with the predicted trajectory point at the current time, and calculate a Kalman gain according to a predicted state covariance matrix.

[0112] The computing unit is configured to calculate a difference between the actual trajectory point and a predicted trajectory point at the current time, and correct the predicted trajectory point at the current time by using a Kalman gain.

[0113] Further, in some optional embodiments of the present application, the acquisition module 21 further comprises:

[0114] The judging unit is configured to acquire the trajectory points within the preset distance of the fracture, and judge whether it is a uniform motion scene.

[0115] The first filling unit is configured to, if it is judged as the uniform motion scene, perform linear interpolation according to the trajectory points at the fracture to fill the fracture.

[0116] The second filling unit is configured to, if it is not judged as the uniform motion scene, perform polynomial interpolation according to the trajectory points within the preset distance of the fracture to fill the fracture.

[0117] Further, in some optional embodiments of the present application, the superimposition module 24 comprises:

[0118] The synthesizing unit is configured to acquire the image of the same region, determine R, G and B three channels in the image, and perform band synthesis on the R, G and B three channels and the near-infrared band data in ENVI or QGIS to generate a four-channel remote sensing image.

[0119] The output unit is configured to apply an NDVI formula to the four-channel remote sensing image through a raster calculator, and output a single-band image with a range of [-1, 1].

[0120] The segmentation unit is configured to perform density segmentation on the region according to the NDVI result, perform false color rendering, and determine a superimposition result.

[0121] The transforming unit is configured to perform Brovey transformation on the superimposition result and then splice the superimposition result to obtain a target farmland map.

[0122] Further, in some optional embodiments of the present application, the rice machine transplanting operation progress statistical system 200 further comprises:

[0123] The generating module is configured to acquire the target farmland map and the NDVI result, set an initial segmentation scale parameter, and generate image objects of different scales in combination with spectral heterogeneity and spatial heterogeneity.

[0124] The extracting module is configured to calculate spectral features, geometric features and texture features for each image object, and extract a target farmland region according to a preset rule set.

[0125] The vectorization module is configured to vectorize the target farmland region to generate a two-dimensional boundary.

[0126] a conversion module, configured to convert the two-dimensional boundary into a three-dimensional boundary according to the height data, and optimize the three-dimensional boundary according to the terrain feature.

[0127] Embodiment Three

[0128] Another aspect of the present application provides an electronic device, which is shown in Figure 3 , which is an electronic device in the embodiment three of the present application, comprising a memory 20, a processor 10, and a computer program 30 stored in the memory and capable of running on the processor, wherein the processor 10 implements the rice machine transplanting operation progress statistical method as described above when running the computer program 30.

[0129] In some embodiments, the processor 10 can be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip, which is configured to run program codes or process data stored in the memory 20, such as executing access restriction programs.

[0130] In some embodiments, the memory 20 can be an internal storage unit of the electronic device, such as a hard disk of the electronic device. In other embodiments, the memory 20 can also be an external storage device of the electronic device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. Further, the memory 20 can include both the internal storage unit and the external storage device. The memory 20 can be used not only to store application software and various data of the electronic device, but also to temporarily store data that has been output or will be output.

[0131] It should be noted that, Figure 3 The structures shown do not constitute a limitation on the electronic device, and in other embodiments, the electronic device can include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0132] The embodiments of the present application also provide a computer readable storage medium, which has a computer program stored thereon, and the program is executed by a processor to implement the rice machine transplanting operation progress statistical method as described above.

[0133] Those skilled in the art will appreciate that the logic and / or steps represented in the flow diagrams, or otherwise described herein, can be embodied in

[0134] More specific examples (a non-exhaustive list) of the computer readable medium include the following: a portable computer diskette (magnetic device); a random access memory (RAM); a read-only memory (ROM); an erasable programmable read-only memory (EPROM or Flash memory); an optical fiber device; and a portable compact disc read-only memory (CDROM). Additionally, the computer readable medium can also be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, for example, via optical scanning of the paper or other medium, then compiled, interpreted, or otherwise processed in a suitable manner, if necessary, and then stored in a computer memory.

[0135] It should be understood that aspects of the application can be implemented in hardware, software, firmware or combinations thereof. In the above embodiments, various steps or methods can be implemented in software or firmware that is stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any of the following technologies, known in the art, can be used: a combination of discrete logic circuits having logic gates for implementing logic functions upon an application of data signals; application specific integrated circuits having logic gates, which are programmed configuration logic gates; programmable gate arrays (PGA), field programmable gate arrays (FPGA), and the like.

[0136] In the description of the specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" and the like means that the specific feature, structure, material or characteristic being described in connection with this embodiment or example is included in at least one embodiment or example of the present application. Descriptive terms of the above-mentioned terms in the specification do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any suitable manner in one or more embodiments or examples.

[0137] The above embodiments only express several implementation manners of the present application, the description is more specific and detailed, but it cannot be understood as the limitation of the patent scope of the present application. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the present application, several modifications and improvements can be made, which belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A method for counting progress of rice machine transplanting operation, characterized by, Applied to a scenario with a rice transplanter terminal device, the method comprises: Obtaining work data of the rice transplanter terminal device, the work data comprising at least work position, work time and work trajectory, and cleaning the work data; According to the cleaned work trajectory, calculating the area of the irregular land boundary, and outputting the land boundary as vector data; According to topological analysis, associating the vector data with corresponding attribute information to form a structured spatial database, wherein the attribute information comprises at least administrative division information and land contractor information; According to the vector data in the structured spatial database, obtaining image and near-infrared band data of each part of the farmland, and superimposing them, then fusing and splicing the superimposed results of each part of the farmland to obtain a target farmland map; According to a deep learning algorithm, performing rice detection and identification on the target farmland map to determine the current planting area; According to the historical rice planting area, the current planting area and the time period, the transplanter progress is counted and visualized. The step of obtaining image and near-infrared band data of each part of the farmland according to the vector data in the structured spatial database, and superimposing them, then fusing and splicing the superimposed results of each part of the farmland to obtain a target farmland map comprises: Obtain the image of the same area, determine the R, G, B three channels in the image, and perform band synthesis on the R, G, B three channels and the near-infrared band data in ENVI or QGIS to generate a four-channel remote sensing image; Apply the NDVI formula to the four-channel remote sensing image through the raster calculator to output a single-band image with a range of [-1, 1]; According to the NDVI result, perform density segmentation on the region and perform false color rendering to determine the superimposed result; After the Brovey transformation of the superimposed result, the target farmland map is obtained by splicing; The step of obtaining image and near-infrared band data of each part of the farmland according to the vector data in the structured spatial database, and superimposing them, then fusing and splicing the superimposed results of each part of the farmland to obtain a target farmland map comprises: Obtain the target farmland map and the NDVI result, set the initial segmentation scale parameter, and combine the spectral heterogeneity and spatial heterogeneity to generate image objects of different scales; Calculate the spectral features, geometric features and texture features for each image object, and extract the target farmland area according to the preset rule set; Vectorize the target farmland area to generate a two-dimensional boundary; According to the elevation data, convert the two-dimensional boundary to a three-dimensional boundary, and optimize the three-dimensional boundary according to the terrain features.

2. The method for counting progress of rice machine transplanting operation according to claim 1, wherein In the step of cleaning the work trajectory, the work trajectory is processed by point-by-point smoothing drift, and the broken part is filled to obtain the cleaned work trajectory, so as to ensure the continuity of the work trajectory.

3. The method for counting progress of rice machine transplanting operation according to claim 2, wherein The step of processing the work trajectory by point-by-point smoothing drift comprises: Obtain the trajectory point of the work trajectory, and predict the trajectory point at the current time according to the trajectory point at the last time and the dynamic model; Fuse the actual trajectory point with the predicted trajectory point at the current time, and calculate the Kalman gain according to the predicted state covariance matrix; Calculate the difference between the actual trajectory point and the predicted trajectory point at the current time, and correct the predicted trajectory point at the current time by weighting with the Kalman gain.

4. The method for counting progress of rice machine transplanting operation according to claim 3, characterized in that, The step of filling the fracture includes: Obtain the trajectory points within a preset distance of the fracture, and determine whether it is a uniform motion scene; If it is determined to be a uniform motion scene, linear interpolation is performed according to the trajectory points at the fracture to fill the fracture; If it is determined to be a non-uniform motion scene, polynomial interpolation is performed according to the trajectory points within a preset distance of the fracture to fill the fracture.

5. The method for counting progress of rice machine transplanting operation according to claim 1, wherein The deep learning algorithm is an improved YOLOv5 algorithm, which is optimized by transfer learning for rice seedling morphological features, wherein the training set contains labeled samples under different growth stages and different light conditions.

6. A rice transplanting machine progress statistics system, characterized in that, For implementing the rice machine transplanting operation progress statistical method according to any one of claims 1-5, the system comprises: An acquisition module configured to acquire operation data of a rice transplanter terminal device, the operation data comprising at least operation position, operation time and operation trajectory, and clean the operation data; A calculation module configured to calculate the area of an irregular land boundary according to the cleaned operation trajectory, and output the land boundary as vector data; An association module configured to associate the vector data with corresponding attribute information according to topological analysis, and form a structured spatial database, wherein the attribute information comprises at least administrative division information and land contractor information; A superposition module configured to acquire image and near-infrared band data of each part of a farmland according to the vector data in the structured spatial database, superpose the data, and then fuse and splice the superposition results of each part of the farmland to obtain a target farmland map; An identification module configured to perform rice detection and identification on the target farmland map according to a deep learning algorithm, and determine a current planting area; A statistical module configured to statistically analyze the rice machine transplanting progress according to historical rice planting areas, the current planting area and a time period, and visualize the progress.

7. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by a processor to implement the rice machine transplanting operation progress statistical method according to any one of claims 1-5.

8. An electronic device, comprising: The computer program is stored in the memory and executable on the processor, and the processor executes the program to implement the rice machine transplanting operation progress statistical method according to any one of claims 1-5.

Images