Intelligent seeding system and method based on vehicle and road cloud collaborative perception

Abstract

The invention provides an intelligent seeding system and method based on vehicle-road cloud collaborative perception. The method comprises the following steps: a vehicle end sensing module acquires surrounding environment data of a seeding machine; the navigation module acquires positioning data of the seeder; the preprocessing module is used for preprocessing the collected surrounding environment data of the seeder; the road test sensing module collects soil data and meteorological data; the cloud side fuses surrounding environment data, positioning data, soil data and meteorological data of the seeder, generates a seeding path plan and a seeding strategy and converts the seeding path plan and the seeding strategy into a seeding control command; the seeder is used for executing seeding behaviors according to the seeding control command. The sensing accuracy of the seeder to the farmland environment is effectively improved, then the seeding precision is improved, precise seeding, automatic navigation and field management can be achieved, and the seeding efficiency and the crop yield are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural sowing, and in particular to an intelligent sowing system and method based on vehicle-road-cloud collaborative perception. Background Art

[0002] With the acceleration of agricultural modernization, the application of crop seeders is becoming more and more widespread. Existing seeders usually rely on sensors on a single vehicle to operate, but due to the complex and changeable farmland environment, this single-point sensing method often has difficulty in accurately obtaining farmland information, resulting in low sowing accuracy, which affects the growth quality and yield of crops. Summary of the invention

[0003] Based on the above problems, the present invention provides an intelligent sowing system and method with vehicle-road-cloud collaborative perception, aiming to solve technical problems in the prior art such as low sowing accuracy affecting the growth quality and yield of crops.

[0004] The present invention provides an intelligent seeding system with vehicle-road-cloud collaborative perception, including a seeding machine, a road test perception module and a cloud, wherein the seeding machine includes a vehicle-side perception module, a preprocessing module, a navigation module and a data transceiver module;

[0005] The vehicle-side sensing module is used to collect the surrounding environment data of the seed drill;

[0006] Navigation module, used to obtain the positioning data of the seed drill;

[0007] The preprocessing module is connected to the vehicle-side sensing module and is used to preprocess the collected data about the surrounding environment of the seed drill;

[0008] The data transceiver module is connected to the preprocessing module and is used to send the preprocessed surrounding environment data and positioning data of the seed drill to the cloud;

[0009] The road test sensing module is used to collect soil data and meteorological data and send the soil data and meteorological data to the cloud;

[0010] The cloud is used to integrate the surrounding environment data, positioning data, soil data and meteorological data of the seed drill to generate seeding path planning and seeding strategy, and convert the seeding strategy and seeding path planning into seeding control commands and send them to the data transceiver module;

[0011] The seeder is used to execute the seeding behavior according to the seeding control command.

[0012] Furthermore, it also includes a status collection module for collecting the sowing status data of the seeder in real time;

[0013] Data transceiver module, connection status collection module, used to feed back seeding status data to the cloud;

[0014] The cloud is used to integrate surrounding environment data, positioning data, soil data, meteorological data, and sowing status data to dynamically update the sowing strategy.

[0015] Furthermore, the vehicle-side perception module includes an image acquisition unit, a laser radar, and a millimeter-wave radar;

[0016] An image acquisition unit, used for acquiring environmental image data around the planter;

[0017] LiDAR, used to collect the first point cloud data around the planter;

[0018] Millimeter wave radar, used to collect the second point cloud data around the planter;

[0019] The preprocessing modules include:

[0020] A first processing unit, configured to identify a first target object based on the environment image data and the first point cloud data;

[0021] A second processing unit, connected to the first processing unit, configured to filter the first target object based on the second point cloud data to obtain a second target object;

[0022] The preprocessed surrounding environment data of the seed drill includes status data of the second target object;

[0023] The cloud is used to dynamically update the seeding path planning based on the positioning data and the status data of the second target object.

[0024] Further, the image acquisition unit acquires two-dimensional environmental image data;

[0025] The first processing unit comprises:

[0026] A region recognition unit is used to perform semantic segmentation and category labeling on two-dimensional environmental image data to obtain semantic recognition results;

[0027] A first detection subunit, configured to identify a two-dimensional target object based on a semantic recognition result, and generate a two-dimensional bounding box for the identified two-dimensional target object;

[0028] The second detection subunit is used to fuse the point cloud data and the semantic recognition result to identify the three-dimensional target object, and generate a three-dimensional bounding box for the identified three-dimensional target object;

[0029] The screening subunit is connected to the first detection subunit and the second detection subunit respectively, and is used to screen out mutually matching two-dimensional bounding boxes and three-dimensional bounding boxes, and use the three-dimensional target object corresponding to the screened three-dimensional bounding box as the first target object.

[0030] Further, the second processing unit includes:

[0031] an object recognition subunit, configured to recognize a third target object based on the second point cloud data;

[0032] The object determination subunit is connected to the object identification subunit and is used to screen the first target object according to the identified third target object and determine the screening result as the second target object.

[0033] A vehicle-road-cloud collaborative perception intelligent seeding method, using the aforementioned vehicle-road-cloud collaborative perception intelligent seeding system, comprising:

[0034] Step A1, collecting surrounding environment data and positioning data of the seed drill, preprocessing the collected surrounding environment data of the seed drill, and collecting soil data and meteorological data;

[0035] Step A2, sending the pre-processed surrounding environment data and positioning data of the seed drill to the cloud, and sending the soil data and meteorological data to the cloud;

[0036] Step A3: The cloud integrates the surrounding environment data, positioning data, soil data and meteorological data of the seed drill to generate a seeding strategy and seeding path planning;

[0037] Step A4: The cloud converts the seeding strategy and seeding path planning into a seeding control command and sends it to the seeding machine;

[0038] Step A5: the seeder performs the sowing operation according to the sowing control command.

[0039] Furthermore, it also includes:

[0040] Step A6, collecting the sowing status data of the seeder in real time, and feeding back the sowing status data to the cloud;

[0041] In step A3, the cloud integrates surrounding environment data, positioning data, soil data, meteorological data, and sowing status data to dynamically update the sowing strategy.

[0042] Further, in step A1, the surrounding environment data of the seed drill includes environmental image data, first point cloud data collected by the laser radar, and second point cloud data collected by the millimeter wave radar;

[0043] In step A1, the process of preprocessing the collected surrounding environment data of the seed drill includes:

[0044] Step A11, identifying a first target object based on the environment image data and the first point cloud data;

[0045] Step A12, filtering the first target object based on the second point cloud data to obtain a second target object;

[0046] In step A1, the preprocessed surrounding environment data of the seed drill includes state data of the second target object;

[0047] In step A3, the cloud dynamically updates the seeding path planning according to the positioning data and the status data of the second target object.

[0048] Furthermore, in step A1, two-dimensional environmental image data is collected;

[0049] Step A11 includes:

[0050] Step A111, performing semantic segmentation and category labeling on the two-dimensional environment image data to obtain a semantic recognition result;

[0051] Step A112, identifying a two-dimensional target object based on the semantic recognition result, and generating a two-dimensional bounding box for the identified two-dimensional target object;

[0052] Step A113, fusing the point cloud data and the semantic recognition result to identify the three-dimensional target object, and generating a three-dimensional bounding box for the identified three-dimensional target object;

[0053] Step A114: Filter out the two-dimensional bounding boxes and three-dimensional bounding boxes that match each other, and use the three-dimensional target object corresponding to the filtered three-dimensional bounding box as the first target object.

[0054] Furthermore, step A12 includes:

[0055] Step A121, identifying a third target object based on the second point cloud data;

[0056] Step A122: Screen the first target object according to the identified third target object, and determine the screening result as the second target object.

[0057] The beneficial technical effect of the present invention is that by combining vehicle-side perception information, road-side perception information and farmland data stored in the cloud, the accuracy of the seeder's perception of the farmland environment is effectively improved, thereby improving the sowing accuracy, achieving precise sowing, automatic navigation and field management, and improving sowing efficiency and crop yield. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1-5 A schematic diagram of a module of a vehicle-road-cloud collaborative perception intelligent seeding system according to the present invention;

[0059] Figure 6-10 This is a flowchart of the steps of an intelligent seeding method for vehicle-road-cloud collaborative perception according to the present invention. DETAILED DESCRIPTION

[0060] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0061] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0062] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments, but they are not intended to limit the present invention.

[0063] See also Figure 1 The present invention provides an intelligent seeding system with vehicle-road-cloud collaborative perception, comprising a seeding machine (1), a road test perception module (2) and a cloud (3), wherein the seeding machine (1) comprises a vehicle-side perception module (11), a preprocessing module (12), a navigation module (13) and a data transceiver module (14);

[0064] A vehicle-side sensing module (11) is used to collect data about the surrounding environment of the seed drill;

[0065] A navigation module (12), used for obtaining positioning data of the seed drill;

[0066] A preprocessing module (13), connected to the vehicle-side sensing module (11), is used to preprocess the collected surrounding environment data of the seed drill;

[0067] A data transceiver module (14) is connected to the preprocessing module (13) and the navigation module (12) respectively, and is used to send the preprocessed surrounding environment data and positioning data of the seed drill to the cloud;

[0068] A road test sensing module (2), used for collecting soil data and meteorological data, and sending the soil data and meteorological data to the cloud;

[0069] The cloud (3) is used to integrate the surrounding environment data, positioning data, soil data and meteorological data of the seed drill to generate a seeding path plan and a seeding strategy, and convert the seeding strategy and seeding path plan into a seeding control command and send it to the data transceiver module (14);

[0070] The seed drill (1) is used for executing seeding behavior according to a seeding control command.

[0071] Specifically, the hardware platform in the seed drill mainly includes a seeding actuator (15), a navigation module (12), a vehicle-side sensing module (11), a preprocessing module (13), a control module (16), and a data transceiver module (14).

[0072] The data transceiver module (14) mainly sends pre-processed surrounding environment data and positioning data to the cloud (3), and receives the sowing control command generated by the cloud. The control module (16) is responsible for issuing the sowing control command to control the sowing actuator (15) to sow. The control module (16) is the brain of the entire seeder system and is responsible for issuing control instructions.

[0073] Furthermore, the sowing strategy includes sowing depth, sowing speed and sowing amount. That is, the sowing control command includes sowing depth, sowing speed and sowing amount. The sowing execution mechanism adjusts the sowing behavior according to the sowing control instruction, executes the sowing operation, and sows the seeds into the field according to the predetermined depth and spacing.

[0074] The vehicle-side perception module (11) uses vehicle-mounted sensors to obtain environmental data around the planter, including but not limited to image acquisition units, laser radars, millimeter-wave radars, etc. The road test perception module (2) obtains soil data and meteorological data through roadside sensors deployed in the field, such as soil sensors and meteorological stations. The soil sensor collects soil data, including key data such as soil moisture, soil temperature, soil pH value, etc. This key data is crucial for evaluating the crop growth environment. The soil sensor sends the measurement results to the cloud via wireless signals, providing real-time data for precision agriculture.

[0075] The weather station collects key meteorological information, including temperature, humidity, wind speed, etc., which are essential for predicting crop growth conditions and adjusting sowing strategies. The meteorological data collected by the weather station is also transmitted to the cloud via wireless network and analyzed together with other sensor data.

[0076] Specifically, wireless communication technologies such as 4G or 5G are used to transmit meteorological data collected by soil moisture sensors and weather stations to the cloud processing center in real time, ensuring the stability and security of data transmission and ensuring that the data is not tampered with or lost during transmission.

[0077] In addition to receiving positioning data, the navigation module also collects real-time path data to ensure that the planter travels along the predetermined path.

[0078] The navigation module may integrate GPS and other sensors to achieve high-precision positioning and path tracking.

[0079] The cloud will fuse data from different sensors to form a comprehensive view of the field environment. The data fusion takes into account the time synchronization and spatial correlation of the data to ensure the accuracy and reliability of the analysis results. The cloud will first pre-process the received data, such as cleaning, verification and formatting, in preparation for subsequent analysis.

[0080] Furthermore, the cloud integrates current and historical surrounding environment data, positioning data, soil data, and meteorological data to form seeding strategies and seeding path planning.

[0081] The present invention realizes precise sowing and significantly improves sowing accuracy through vehicle-road-cloud collaborative technology. This technology uses on-board sensors and roadside sensors to collect field data, and conducts in-depth analysis and processing through a cloud data processing center. In this way, the seeder can adjust the sowing behavior according to real-time data to ensure that seeds are sown under optimal conditions, thereby improving the growth potential and yield of crops. At the same time, automated navigation and sowing control reduce human intervention and effectively improve sowing efficiency. The navigation system of the seeder can automatically drive according to the generated sowing path planning, while the sowing control module accurately controls the sowing depth and spacing according to the sowing control commands generated by the cloud, reducing human errors and improving work efficiency. In addition, the system can monitor field environmental data in real time and dynamically adjust the sowing strategy, optimizing field management. By continuously monitoring key parameters such as soil moisture, temperature, and pH value, the system can respond to environmental changes in a timely manner, adjust the sowing strategy, and ensure that crops grow under optimal growth conditions. These innovations not only improve crop growth conditions, but also enhance crop yields, bringing significant benefits to modern agricultural production. By reducing resource waste and improving crop growth quality and yield, the present invention helps to achieve sustainable agricultural development while providing farmers with an efficient and economical sowing solution. The implementation of this invention not only improves the efficiency and quality of agricultural production, but also provides a new direction for the further development of agricultural science and technology, and promotes the progress of precision agriculture and intelligent agricultural machinery technology. Through vehicle-road-cloud collaborative technology, this invention provides strong support for the intelligent, automated and data-driven transformation of modern agriculture.

[0082] Furthermore, a status acquisition module (17) is used to collect the sowing status data of the sowing machine in real time;

[0083] A data transceiver module (14) and a connection status acquisition module (17) are used to feed back the sowing status data to the cloud;

[0084] The cloud (3) is used to integrate surrounding environment data, positioning data, soil data, meteorological data and sowing status data to dynamically update the sowing strategy.

[0085] The cloud also combines the real-time sowing status data of the seeder to dynamically adjust the sowing strategy.

[0086] Furthermore, the cloud integrates current and historical surrounding environment data, positioning data, soil data, meteorological data, and sowing status data to update the sowing strategy.

[0087] The cloud integrates data, considers crop distribution, soil conditions and expected operation results, and uses algorithms to plan the most efficient sowing path, generating the optimal operation path for agricultural machinery. Based on the path planning and field environment, a specific sowing strategy is generated, and parameters such as sowing depth, sowing speed and sowing amount are determined to generate an executable sowing strategy. In other words, as an implementation method, the cloud first generates a sowing path plan, and then combines the sowing path plan to form a sowing strategy.

[0088] The real-time sowing status data of the seeder includes real-time sowing depth, real-time sowing speed, real-time sowing position, etc. Optimize the sowing strategy based on real-time feedback. The cloud dynamically adjusts the sowing control instructions based on the real-time feedback from the seeder to ensure the accuracy and adaptability of the sowing operation.

[0089] See also Figure 2-3 , further, the vehicle-side perception module (11) includes an image acquisition unit (111), a laser radar (112) and a millimeter wave radar (113);

[0090] An image acquisition unit (111) is used to acquire environmental image data around the seed drill;

[0091] A laser radar (112), used for collecting first point cloud data around the planter;

[0092] The millimeter wave radar (113) is used to collect second point cloud data around the seed drill; the preprocessing module (12) includes:

[0093] A first processing unit (121) is used to identify a first target object based on the environment image data and the first point cloud data;

[0094] A second processing unit (122), connected to the first processing unit (121), configured to filter the first target object based on the second point cloud data to obtain a second target object;

[0095] The preprocessed surrounding environment data of the seed drill includes status data of the second target object;

[0096] The cloud (3) is used to dynamically update the seeding path planning according to the positioning data and the status data of the second target object.

[0097] The first target object is an obstacle identified by combining the environment image data and the first point cloud data.

[0098] The second target object is a dynamic or static obstacle that affects the sowing path of the seed drill, which is screened out from the first target object. When working in the field, there may not be only one seed drill, but multiple seed drills may be sowing at the same time. The seed drill may be stopped or moving, and of course it may also be other moving objects. The present invention identifies whether a moving object causes an effective obstacle to the seed drill. If the moving object does not cause an effective obstacle, for example, the moving object may be able to leave when the seed drill arrives, there is no need to avoid the obstacle. This can reduce unnecessary obstacle avoidance of the seed drill and improve sowing efficiency.

[0099] As a way, the cloud integrates current and historical surrounding environment data, positioning data, soil data, meteorological data, and sowing status data, and combines the sown and unsown areas of the farmland to form sowing path planning and sowing strategies in real time.

[0100] See also Figure 4 , further, the image acquisition unit (111) acquires two-dimensional environmental image data;

[0101] The first processing unit (121) comprises:

[0102] The region recognition subunit (1211) is used to perform semantic segmentation and category labeling on the two-dimensional environment image data to obtain a semantic recognition result;

[0103] The first detection subunit (1212) is connected to the region recognition subunit (1211), and is used to recognize the two-dimensional target object based on the semantic recognition result, and generate a two-dimensional bounding box for the recognized two-dimensional target object;

[0104] The second detection subunit (1213) is connected to the region recognition subunit (1211) and is used to fuse the point cloud data and the semantic recognition result to identify the three-dimensional target object, and generate a three-dimensional bounding box for the identified three-dimensional target object;

[0105] The screening subunit (1214) is connected to the first detection subunit (1212) and the second detection subunit (1213) respectively, and is used to screen out two-dimensional bounding boxes and three-dimensional bounding boxes that match each other, and use the three-dimensional target object corresponding to the screened three-dimensional bounding box as the first target object.

[0106] The region recognition subunit performs semantic segmentation on the 2D image to segment different regions and objects in the 2D image. Through deep learning algorithms, the image is segmented into different semantic regions, such as roads, vehicles, pedestrians, etc. Then, based on the semantic segmentation, the specific category of each segmented region is further identified and labeled, providing a basis for 3D target object detection.

[0107] The first detection subunit (1212) detects obstacles in the 2D image of the semantic recognition result, detects obstacles in the image, and generates a 2D bounding box to provide the planar position information of the obstacles.

[0108] The second detection subunit (1213) detects the 3D target in the two-dimensional image using the two-dimensional image segmented and annotated by the semantic recognition result and the first point cloud data of the laser radar, generates a 3D bounding box, and provides spatial position information of the obstacle.

[0109] The screening subunit (1214) matches the two-dimensional bounding box with the three-dimensional bounding box by calculating the shortest distance, and then calculates the intersection-and-union ratio of the matched two-dimensional bounding box and the three-dimensional bounding box. If the intersection-and-union ratio exceeds the threshold, it is retained; if the intersection-and-union ratio does not exceed the threshold, it is deleted, and the three-dimensional target object corresponding to the three-dimensional bounding box whose intersection-and-union ratio exceeds the threshold is used as the first target object. By evaluating the consistency of the detection results of the two-dimensional target object and the three-dimensional target object, a basis is provided for obstacle filtering. The 3D bounding box whose intersection-and-union ratio does not exceed the threshold is removed to retain accurate obstacle information. The 3D bounding box that meets the conditions is retained to retain accurate 3D target object information for subsequent obstacle processing and seeding path planning.

[0110] LiDAR can obtain accurate distance and shape information of the vehicle's surroundings, generate the first point cloud data, provide accurate spatial information, project the point cloud data generated by LiDAR into a 2D image, enhance the depth information of the image, and assist in 3D target object detection.

[0111] See also Figure 5 , further, the second processing unit (122) comprises:

[0112] An object recognition subunit (1221), configured to recognize a third target object based on the second point cloud data;

[0113] The object determination subunit (1222) is connected to the object identification subunit (1221) and is used to filter the first target object according to the identified third target object and determine the screening result as the second target object.

[0114] The millimeter-wave radar provides speed and direction information of obstacles around the vehicle, detects dynamic obstacles around the vehicle, and provides speed and direction information. Combined with the identified first target object and the second point cloud data generated by the millimeter-wave radar, obstacles that affect the vehicle's travel are identified, providing safety protection for seeding path planning.

[0115] See also Figure 6 The present invention also provides a vehicle-road-cloud collaborative perception intelligent seeding method, using the aforementioned vehicle-road-cloud collaborative perception intelligent seeding system, including:

[0116] Step A1, collecting surrounding environment data and positioning data of the seed drill, preprocessing the collected surrounding environment data of the seed drill, and collecting soil data and meteorological data;

[0117] Step A2, sending the pre-processed surrounding environment data and positioning data of the seed drill to the cloud, and sending the soil data and meteorological data to the cloud;

[0118] Step A3: The cloud integrates the surrounding environment data, positioning data, soil data and meteorological data of the seed drill to generate a seeding strategy and seeding path planning;

[0119] Step A4: The cloud converts the seeding strategy and seeding path planning into a seeding control command and sends it to the seeding machine;

[0120] Step A5: the seeder performs the sowing operation according to the sowing control command.

[0121] Furthermore, the sowing strategy includes sowing depth, sowing speed and sowing amount. That is, the sowing control command includes sowing depth, sowing speed and sowing amount. The sowing execution mechanism adjusts the sowing behavior according to the sowing control instruction, executes the sowing operation, and sows the seeds into the field according to the predetermined depth and spacing.

[0122] In step A1, the surrounding environment data includes but is not limited to data collected by the image acquisition unit, laser radar, millimeter wave radar, etc. Soil data includes but is not limited to key data such as soil moisture, soil temperature, soil pH value, etc. This key data is crucial for evaluating the crop growth environment. The weather station collects key meteorological data including temperature, humidity, wind speed, etc.

[0123] See also Figure 7 , further including:

[0124] Step A6, collecting the sowing status data of the seeder in real time, and feeding back the sowing status data to the cloud;

[0125] In step A3, the cloud integrates surrounding environment data, positioning data, soil data, meteorological data, and sowing status data to dynamically update the sowing strategy.

[0126] The real-time sowing status data of the seeder includes real-time sowing depth, real-time sowing speed, real-time sowing position, etc. Optimize the sowing strategy based on real-time feedback. The cloud dynamically adjusts the sowing control instructions based on the real-time feedback from the seeder to ensure the accuracy and adaptability of the sowing operation.

[0127] See also Figure 8,Further, in step A1, the surrounding environment data of the seed drill includes environmental image data, first point cloud data collected by the laser radar, and second point cloud data collected by the millimeter wave radar;

[0128] In step A1, the process of preprocessing the collected surrounding environment data of the seed drill includes:

[0129] Step A11, identifying a first target object based on the environment image data and the first point cloud data;

[0130] Step A12, filtering the first target object based on the second point cloud data to obtain a second target object;

[0131] In step A1, the preprocessed surrounding environment data of the seed drill includes state data of the second target object;

[0132] In step A3, the cloud dynamically updates the seeding path planning according to the positioning data and the status data of the second target object.

[0133] The first target object is an obstacle identified by combining the environment image data and the first point cloud data.

[0134] The second target object is a dynamic or static obstacle that affects the sowing path of the seed drill, which is screened out from the first target object. When working in the field, there may not be only one seed drill, but multiple seed drills may be sowing at the same time. The seed drill may be stopped or moving, and of course it may also be other moving objects. The present invention identifies whether a moving object causes an effective obstacle to the seed drill. If the moving object does not cause an effective obstacle, for example, the moving object may be able to leave when the seed drill arrives, there is no need to avoid the obstacle. This can reduce unnecessary obstacle avoidance of the seed drill and improve sowing efficiency.

[0135] See also Fig. 9 ,Further, in step A1, two-dimensional environmental image data is collected;

[0136] Step A11 includes:

[0137] Step A111, performing semantic segmentation and category labeling on the two-dimensional environment image data to obtain a semantic recognition result;

[0138] Step A112, identifying a two-dimensional target object based on the semantic recognition result, and generating a two-dimensional bounding box for the identified two-dimensional target object;

[0139] Step A113, fusing the point cloud data and the semantic recognition result to identify the three-dimensional target object, and generating a three-dimensional bounding box for the identified three-dimensional target object;

[0140] Step A114: Filter out the two-dimensional bounding boxes and three-dimensional bounding boxes that match each other, and use the three-dimensional target object corresponding to the filtered three-dimensional bounding box as the first target object.

[0141] Semantic segmentation is performed on 2D images to segment different regions and objects in the 2D images. Through deep learning algorithms, images are segmented into different semantic regions, such as roads, vehicles, pedestrians, etc. Then, based on semantic segmentation, the specific categories of each segmented region are further identified and labeled, providing a basis for 3D target object detection.

[0142] Detect obstacles in the 2D image of the semantic recognition result. Detect obstacles in the image and generate a 2D bounding box to provide the planar position information of the obstacle.

[0143] The 2D image segmented and annotated by the semantic recognition result and the first point cloud data of the LiDAR are used to detect the 3D target in the 2D image, generate a 3D bounding box, and provide the spatial position information of the obstacle.

[0144] Specifically, in step A114, the two-dimensional bounding box and the three-dimensional bounding box are matched by calculating the shortest distance, and then the intersection and union ratio of the matched two-dimensional bounding box and the three-dimensional bounding box is calculated. If the intersection and union ratio exceeds the threshold, it is retained. If the intersection and union ratio does not exceed the threshold, it is deleted, and the three-dimensional target object corresponding to the three-dimensional bounding box whose intersection and union ratio exceeds the threshold is used as the first target object. By evaluating the consistency of the detection results of the two-dimensional target object and the three-dimensional target object, a basis is provided for obstacle filtering. The 3D bounding box whose intersection and union ratio does not exceed the threshold is removed to retain accurate obstacle information. The 3D bounding box that meets the conditions is retained to retain accurate 3D target object information for subsequent obstacle processing and seeding path planning.

[0145] See also Fig.10 , further, step A12 includes:

[0146] Step A121, identifying a third target object based on the second point cloud data;

[0147] Step A122: Screen the first target object according to the identified third target object, and determine the screening result as the second target object.

[0148] The millimeter-wave radar provides speed and direction information of obstacles around the vehicle, detects dynamic obstacles around the vehicle, and provides speed and direction information. Combined with the identified first target object and the second point cloud data generated by the millimeter-wave radar, obstacles that affect the vehicle's travel are identified, providing safety protection for seeding path planning.

[0149] It can be seen that the present invention combines vehicle-road-cloud collaborative perception to control the seed drill to sow, which can:

[0150] Improve seeding accuracy: Through advanced vehicle-side and roadside sensing systems, precise control of seeding depth and spacing can be achieved.

[0151] Enhanced seeding efficiency: Utilize cloud data processing and collaborative control algorithms to automate the seeding process and reduce manual intervention.

[0152] Optimize field management: monitor the field environment in real time, dynamically adjust sowing strategies, and improve crop growth conditions.

[0153] Increase crop yield: Reasonable sowing density and depth improve the uniformity of crop growth and yield.

[0154] Improve system reliability: Through vehicle-road-cloud collaborative technology, the adaptability and stability of the planter under different field conditions can be improved.

[0155] The above are only preferred embodiments of the present invention, and are not intended to limit the implementation methods and protection scope of the present invention. Those skilled in the art should be aware that all solutions obtained by equivalent substitutions and obvious changes made using the description and illustrations of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent seeding system based on vehicle-road-cloud collaborative perception, characterized in that: It includes a seeder, a road test perception module and a cloud, wherein the seeder includes a vehicle-side perception module, a preprocessing module, a navigation module and a data transceiver module; The vehicle-side sensing module is used to collect the surrounding environment data of the seed drill; The navigation module is used to obtain the positioning data of the seed drill; The preprocessing module is connected to the vehicle-side sensing module and is used to preprocess the collected surrounding environment data of the seed drill; The data transceiver module is connected to the preprocessing module and the navigation module respectively, and is used to send the preprocessed surrounding environment data of the seed drill and the positioning data to the cloud; A drive test sensing module, used for collecting soil data and meteorological data, and sending the soil data and the meteorological data to the cloud; The cloud is used to integrate the surrounding environment data of the seed drill, the positioning data, the soil data and the meteorological data, generate a seeding path plan and a seeding strategy, and convert the seeding strategy and the seeding path plan into a seeding control command and send it to the data transceiver module; The seed drill is used to perform seeding behavior according to the seeding control command.

2. The intelligent seeding system based on vehicle-road-cloud collaborative perception as claimed in claim 1, characterized in that: A status collection module is used to collect the sowing status data of the seeder in real time; The data transceiver module is connected to the status acquisition module and is used to feed back the sowing status data to the cloud; The cloud is used to integrate the surrounding environment data, the positioning data, the soil data, the meteorological data and the sowing status data to dynamically update the sowing strategy.

3. The intelligent seeding system based on vehicle-road-cloud collaborative perception as claimed in claim 2, characterized in that: The vehicle-side perception module includes an image acquisition unit, a laser radar and a millimeter-wave radar; An image acquisition unit, used for acquiring environmental image data around the seed drill; A laser radar, used for collecting first point cloud data around the planter; A millimeter wave radar, used for collecting second point cloud data around the planter; 、 The preprocessing module comprises: A first processing unit, configured to identify a first target object based on the environment image data and the first point cloud data; a second processing unit, connected to the first processing unit, and configured to filter the first target object based on the second point cloud data to obtain a second target object; The preprocessed surrounding environment data of the seed drill includes the status data of the second target object; The cloud is used to dynamically update the seeding path planning according to the positioning data and the status data of the second target object.

4. The intelligent seeding system based on vehicle-road-cloud collaborative perception as claimed in claim 3, characterized in that: The image acquisition unit acquires two-dimensional image data of the environment; The first processing unit comprises: A region recognition subunit is used to perform semantic segmentation and category labeling on the two-dimensional environment image data to obtain a semantic recognition result; a first detection subunit, connected to the region recognition subunit, configured to recognize a two-dimensional target object based on the semantic recognition result, and generate a two-dimensional bounding box for the recognized two-dimensional target object; A second detection subunit, connected to the region recognition subunit, is used to fuse the point cloud data and the semantic recognition result to identify a three-dimensional target object, and generate a three-dimensional bounding box for the identified three-dimensional target object; The screening subunit is connected to the first detection subunit and the second detection subunit, respectively, and is used to screen out the two-dimensional bounding box and the three-dimensional bounding box that match each other, and use the three-dimensional target object corresponding to the screened three-dimensional bounding box as the first target object.

5. The intelligent seeding system based on vehicle-road-cloud collaborative perception as claimed in claim 3, characterized in that: The second processing unit comprises: an object recognition subunit, configured to recognize a third target object based on the second point cloud data; The object determination subunit is connected to the object identification subunit and is used to screen the first target object according to the identified third target object and determine the screening result as the second target object.

6. An intelligent seeding method based on vehicle-road-cloud collaborative perception, characterized in that: An intelligent seeding system using vehicle-road-cloud collaborative perception as described in any one of claims 1 to 5, comprising: Step A1, collecting the surrounding environment data and positioning data of the seed drill, preprocessing the collected surrounding environment data of the seed drill, and collecting soil data and meteorological data; Step A2, sending the preprocessed surrounding environment data and the positioning data of the seed drill to the cloud, and sending the soil data and the meteorological data to the cloud; Step A3, the cloud integrates the surrounding environment data of the seed drill, the positioning data, the soil data and the meteorological data to generate a seeding strategy and a seeding path planning; Step A4, the cloud converts the seeding strategy and the seeding path planning into a seeding control command and sends it to the seeding machine; Step A5: the seeder performs the sowing operation according to the sowing control command.

7. The intelligent seeding method of vehicle-road-cloud collaborative perception as claimed in claim 6, characterized in that: Also includes: Step A6, collecting the sowing status data of the seeder in real time, and feeding back the sowing status data to the cloud; In the step A3, the cloud merges the surrounding environment data, the positioning data, the soil data, the meteorological data and the sowing status data to dynamically update the sowing strategy.

8. The intelligent seeding method of vehicle-road-cloud collaborative perception as claimed in claim 6, characterized in that: In the step A1, the surrounding environment data of the seed drill includes environmental image data, first point cloud data collected by a laser radar, and second point cloud data collected by a millimeter wave radar; In step A1, the process of preprocessing the collected surrounding environment data of the seed drill includes: Step A11, identifying a first target object based on the environment image data and the first point cloud data; Step A12, filtering the first target object based on the second point cloud data to obtain a second target object; In the step A1, the preprocessed surrounding environment data of the seed drill includes the status data of the second target object; In the step A3, the cloud dynamically updates the seeding path planning according to the positioning data and the status data of the second target object.

9. The intelligent seeding method of vehicle-road-cloud collaborative perception as claimed in claim 8, characterized in that: In the step A1, two-dimensional image data of the environment is collected; The step A11 comprises: Step A111, performing semantic segmentation and category labeling on the two-dimensional environment image data to obtain a semantic recognition result; Step A112, identifying a two-dimensional target object based on the semantic recognition result, and generating a two-dimensional bounding box for the identified two-dimensional target object; Step A113, fusing the point cloud data and the semantic recognition result to identify a three-dimensional target object, and generating a three-dimensional bounding box for the identified three-dimensional target object; Step A114: Filter out the two-dimensional bounding box and the three-dimensional bounding box that match each other, and use the three-dimensional target object corresponding to the filtered three-dimensional bounding box as the first target object.

10. The intelligent seeding method of vehicle-road-cloud collaborative perception as claimed in claim 8, characterized in that: The step A12 comprises: Step A121, identifying a third target object based on the second point cloud data; Step A122: Screen the first target object according to the identified third target object, and determine the screening result as the second target object.

Images