An intelligent orchard transport system and method thereof

Abstract

The invention discloses an orchard intelligent transportation system and method thereof, and relates to the technical field of orchard intelligent transportation, including: collecting high-altitude data and ground data of an orchard from high altitude and ground respectively by using a drone and a ground robot; generating a three-dimensional model of the orchard by fusing the high-altitude data and ground data collected by the drone and the ground robot; generating multiple transportation paths based on the three-dimensional model of the orchard by using a path planning algorithm; optimizing transportation paths of different transportation starting points by using a quantum approximate optimization algorithm, and outputting an optimal transportation path; an unmanned automated transportation fleet collaboratively performs transportation tasks based on the optimal transportation path; the invention generates multiple alternative paths based on the three-dimensional model by using a path planning algorithm, and efficiently determines the global optimal path of multi-starting point transportation tasks in combination with a quantum approximate optimization algorithm, thereby significantly improving the efficiency and accuracy of path optimization, and reducing transportation time and energy consumption.

Description

Technical Field

[0001] The present invention relates to the technical field of orchard intelligent transportation, and in particular to an orchard intelligent transportation system and method thereof. Background Art

[0002] Traditional orchard handling mainly relies on manual operation, which is labor-intensive, inefficient and costly. In modern agricultural production, intelligent and unmanned handling technology has gradually become an important direction to improve orchard operation efficiency and reduce operating costs.

[0003] In the existing technology, path planning algorithms are mostly based on classic graph search methods (such as A* or Dijkstra algorithm). Although they can generate feasible paths for a single task, their computational complexity increases rapidly with the size of the orchard. In addition, most existing handling solutions rely only on ground robots or single sensors (such as LiDAR or cameras) for data collection, and cannot fully obtain a high-precision three-dimensional model of the orchard, which leads to incomplete basic data for path planning, further affecting handling efficiency. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an orchard intelligent transportation method to solve the problem of low efficiency in transportation path planning.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides an orchard intelligent transportation method, which comprises:

[0008] Use drones and ground robots to collect aerial and ground data of the orchard from high altitude and ground respectively;

[0009] By integrating the aerial data and ground data collected by drones and ground robots, a three-dimensional model of the orchard is generated;

[0010] Based on the 3D model of the orchard, multiple transport paths are generated using a path planning algorithm;

[0011] Apply quantum approximate optimization algorithm to optimize the transport paths of different transport starting points and output the optimal transport path;

[0012] The unmanned automated transport fleet collaborates to perform transport tasks based on the optimal transport path.

[0013] As a preferred solution of the orchard intelligent transportation method of the present invention, wherein: the aerial data and ground data of the orchard are collected from the high altitude and the ground by the drone and the ground robot respectively, the specific steps are as follows:

[0014] A drone equipped with a LiDAR sensor and an RGB camera collects aerial data of the orchard from high altitude to obtain aerial point cloud data and aerial images of the orchard.

[0015] On the ground, a route collection robot equipped with a LiDAR sensor and an RGB camera is used to drive along the roads in the orchard and collect ground data on the road distribution and obstacles along the way, obtaining point cloud data and images of ground roads and obstacles.

[0016] The collected high-altitude data and ground data are uploaded to the control center through wireless transmission.

[0017] As a preferred solution of the orchard intelligent transportation method of the present invention, wherein: the three-dimensional model of the orchard is generated by integrating the high-altitude data and ground data collected by the drone and the ground robot, and the specific steps are as follows:

[0018] In the control center, the voxel filtering method is used to filter the orchard high-altitude point cloud data and the ground road and obstacle point cloud data collected from the high altitude and the ground;

[0019] Using the ICP algorithm, the aerial point cloud data and the ground point cloud data were initially fused and registered to obtain a preliminary 3D model of the orchard.

[0020] The Mask R-CNN model is used to identify the locations of roads, fruit trees, and obstacles in the aerial images of the orchard and the images of roads and obstacles on the ground.

[0021] Use the Canny edge detection algorithm to extract edge information of roads, fruit trees, and obstacles from the recognition results;

[0022] Use the findContours function in the OpenCV library to extract the contours of roads, fruit trees, and obstacles from the recognition results;

[0023] According to the timestamps and corresponding point cloud data of the orchard aerial images and the ground road and obstacle images, the edge information and contours of the extracted roads, fruit trees and obstacles are projected into the preliminary 3D model of the orchard to generate the final 3D model of the orchard.

[0024] As a preferred solution of the orchard intelligent handling method of the present invention, the Mask R-CNN model is used to identify the positions of roads, fruit trees and obstacles in the orchard aerial image and the ground road and obstacle image. The specific steps are as follows:

[0025] Collect RGB images of orchard environments covering different seasons, lighting conditions, and weather conditions and annotate the locations of roads, fruit trees, and obstacles;

[0026] Crop, flip, rotate and scale the annotated RGB image;

[0027] Normalize the pixel values ​​of the processed RGB images and divide them into training set, validation set and test set;

[0028] Input the training set into the Mask R-Mask R-CNN model for training;

[0029] The cross entropy loss is used to measure the difference between the prediction result of the classification branch and the actual label;

[0030] Smooth L1 loss is used to measure the difference between the bounding box predicted by the regression branch and the true bounding box;

[0031] Use binary cross entropy loss to measure the difference between the pixel-level mask generated by the mask branch and the true mask;

[0032] At the end of each training round, the validation set is input and the performance of the model is evaluated by calculating the average precision and intersection-over-union ratio;

[0033] Adjust hyperparameters based on the evaluation results of the validation set;

[0034] When the Mask R-Mask R-CNN model completes training for all RGB images in the training set, the training ends and the performance of the Mask R-Mask R-CNN model is tested using the test set.

[0035] The trained Mask R-Mask R-CNN model is used to identify the locations of roads, fruit trees, and obstacles in aerial images of orchards and images of ground roads and obstacles.

[0036] As a preferred solution of the orchard intelligent transportation method of the present invention, wherein: based on the orchard three-dimensional model, a plurality of transportation paths are generated by using a path planning algorithm, and the specific steps are as follows:

[0037] Based on the distribution of fruit trees, roads and obstacles in the 3D model of the orchard, the orchard is logically partitioned and each partition is used as the starting point for transportation;

[0038] All intersections of roads are used as transport nodes;

[0039] Determine the transportation destination according to the fruit storage and transportation requirements;

[0040] Based on the transport start point, the transport node and the transport end point of the transport path, all transport paths from the transport start point to the transport end point are generated.

[0041] As a preferred solution of the orchard intelligent transportation method of the present invention, wherein: the quantum approximate optimization algorithm is applied to optimize the transportation paths of different transportation starting points and output the optimal transportation path. The specific steps are as follows:

[0042] The path optimization problem is modeled as a graph optimization problem using the quantum approximate optimization algorithm, and the orchard area is represented as a graph. ;

[0043] in, is the graph node set of the transport starting point, transport node and transport end point of all transport paths, is the set of edges connecting two graph nodes;

[0044] Define constraints based on transportation requirements and road conditions;

[0045] For all transport paths with different transport starting points, the length of the path and the transport time are taken as optimization targets, and the objective function is set;

[0046] Based on the objective function, the path optimization problem is converted into a quantum Hamiltonian, which is expressed as:

[0047] ;

[0048] ;

[0049] ;

[0050] in, is the target quantum Hamiltonian value, Indicates the goal, For graph nodes arrive The objective function value of is the diagonal element of the Pauli Z matrix corresponding to the quantum bit, indicating whether the graph node is selected arrive , is the mixed Hamiltonian, Indicates mixed, and They are any two adjacent graph nodes in the graph node set;

[0051] Construct a parameterized quantum circuit based on the target quantum Hamiltonian, and initialize all quantum bits to a uniform superposition state. The expression is as follows:

[0052] ;

[0053] in, is the initial quantum state, the uniform superposition state of all possible paths, is the number of quantum bits, corresponding to the number of optimization targets, Indicates the transport path;

[0054] In each layer of the quantum approximate optimization algorithm, the target Hamiltonian evolution and the mixed Hamiltonian evolution are applied in turn to iteratively update the quantum state. After Q layers of iteration, the quantum state evolves as follows:

[0055] ;

[0056] in, is the quantum state after the Q-layer evolution, To optimize the number of circuit layers, To optimize the index of the number of layers of the circuit, For the layer The control parameters of For the layer The control parameters of

[0057] For quantum states Take measurements and get the result distribution, expressed as:

[0058] ;

[0059] in, For transport path The probability of occurrence;

[0060] The transport path with the highest probability of appearing in the measurement results is the optimal transport path for each transport starting point.

[0061] As a preferred solution of the orchard intelligent transportation method of the present invention, the unmanned automated transportation fleet collaboratively performs the transportation task based on the optimal transportation path, and the specific steps are as follows:

[0062] The task allocation priority is set according to the total length of the optimal transport path for each transport starting point and the position, loading status and remaining power of each unmanned automated transport vehicle;

[0063] The unmanned automated transport vehicle carries out real-time scanning and detection of the road ahead based on the camera and LiDAR sensor;

[0064] When an obstacle is detected, the LiDAR sensor is used to scan the point cloud data of the obstacle and upload it to the control center to update the 3D model of the orchard;

[0065] At the same time, the obstacle location is marked and synchronized to all unmanned automated transport vehicles. After the control center re-optimizes the optimal transport path, they turn around and perform the transport task according to the new optimal transport path.

[0066] In a second aspect, the present invention provides an orchard intelligent transport system, including a data acquisition module, a three-dimensional model module, a transport path module, a path optimization module and an execution transport module.

[0067] The data acquisition module is used to collect the high-altitude data and the ground data of the orchard from the high altitude and the ground respectively through the unmanned aerial vehicle and the ground robot;

[0068] The three-dimensional model module is used to generate a three-dimensional model of the orchard by fusing the high-altitude data and ground data collected by the drone and the ground robot;

[0069] The transport path module is used to generate multiple transport paths based on the orchard three-dimensional model using a path planning algorithm;

[0070] The path optimization module is used to apply the quantum approximate optimization algorithm to optimize the transport paths of different transport starting points and output the optimal transport path;

[0071] The execution transport module is used for the unmanned automated transport fleet to collaboratively execute transport tasks based on the optimal transport path.

[0072] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the orchard intelligent transportation method as described in the first aspect of the present invention is implemented.

[0073] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the orchard intelligent transportation method as described in the first aspect of the present invention is implemented.

[0074] The beneficial effects of the present invention are as follows: the present invention uses drones and ground robots to collaboratively collect high-altitude and ground data to generate a high-precision three-dimensional model of an orchard, thereby achieving comprehensive coverage and accurate modeling of the orchard environment, and providing a high-quality data basis for path planning. Based on the three-dimensional model, multiple alternative paths are generated through a path planning algorithm, and combined with a quantum approximate optimization algorithm, the global optimal path of multi-starting point handling tasks is efficiently determined, which significantly improves the efficiency and accuracy of path optimization, reduces handling time and energy consumption, and in the handling execution stage, through the collaborative scheduling of unmanned automated handling fleets, tasks are allocated based on vehicle position, power level and other status, and combined with real-time obstacle detection and dynamic path adjustment functions, the continuity and stability of the handling tasks are guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0076] Figure 1 This is a flow chart of the intelligent orchard transportation method in Example 1.

[0077] Figure 2 This is a module diagram of the orchard intelligent handling system in Example 1. DETAILED DESCRIPTION

[0078] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0079] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0080] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0081] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and provides an orchard intelligent transportation method, comprising the following steps:

[0082] S1. Collect aerial data and ground data of the orchard from high altitude and ground respectively through drones and ground robots.

[0083] A drone equipped with a LiDAR sensor and an RGB camera collects aerial data of the orchard from high altitude to obtain aerial point cloud data and aerial images of the orchard.

[0084] Specifically, the drone takes a high-resolution RGB image every 1 second to record the current geographic location and attitude information. The resolution of the orchard aerial image is set to 1920x1080 pixels to ensure that the details of the road and fruit trees can be clearly captured;

[0085] The drone's LiDAR sensor generates point cloud data at a rate of 100,000 points per second, recording the terrain and three-dimensional coordinates of objects in the orchard. Every time it takes a photo, the drone will synchronously record the LiDAR data to ensure the temporal consistency of the orchard's aerial images and the orchard's aerial point cloud data.

[0086] In order to improve the accuracy and reliability of high-altitude data, drones will conduct multiple scans in each key area (such as road intersections and areas with dense fruit trees) to ensure the integrity and consistency of high-altitude data;

[0087] On the ground, a route collection robot equipped with a LiDAR sensor and an RGB camera is used to drive along the roads in the orchard and collect ground data on the road distribution and obstacles along the way, obtaining point cloud data and images of ground roads and obstacles.

[0088] Specifically, the ground road and obstacle point cloud data and ground road and obstacle images contain the height of the orchard terrain, the road width, the obstacle conditions on the road, etc.;

[0089] The collected high-altitude data and ground data are uploaded to the control center through wireless transmission.

[0090] S2. Generate a three-dimensional model of the orchard by integrating the aerial data and ground data collected by drones and ground robots.

[0091] In the control center, the voxel filtering method is used to filter the orchard high-altitude point cloud data and the ground road and obstacle point cloud data collected from the high altitude and the ground to remove outliers and noise points and improve the point cloud quality;

[0092] The ICP (Iterative Closest Point) algorithm is used to initially fuse and register the aerial point cloud data with the ground point cloud data to obtain a preliminary 3D model of the orchard.

[0093] Specifically, the KD Tree spatial index structure is used to find the closest point in the ground point cloud from each point in the high-altitude point cloud to form a pair of corresponding points. Then, a two-way matching strategy is adopted, that is, starting from each point in the ground point cloud, the closest point in the high-altitude point cloud is found to ensure the symmetry and consistency of the matching results.

[0094] After finding the point pairs corresponding to the orchard aerial point cloud and the ground road and obstacle point cloud, the optimal rotation matrix and translation vector for transforming the orchard aerial point cloud to the ground road and obstacle point cloud is calculated by minimizing the sum of squared Euclidean distances between the point pairs. The expression is as follows:

[0095] ;

[0096] in, is the optimal rotation matrix, is the translation vector, is the number of corresponding point pairs, is the index of the number of corresponding point pairs, For the Ground road and obstacle point clouds, For the A high-altitude point cloud of an orchard;

[0097] According to the point pairs corresponding to the orchard aerial point cloud and the ground road and obstacle point cloud, the covariance matrix is ​​constructed. The expression is as follows:

[0098] ;

[0099] in, is the covariance matrix, is the centroid of the ground road and obstacle point cloud, is the centroid of the high-altitude point cloud of the garden;

[0100] By changing the covariance matrix Perform singular value decomposition and calculate the rotation matrix and translation vector. The expression is as follows:

[0101] ;

[0102] ;

[0103] ;

[0104] in, and Each is a 3×3×3 orthogonal matrix, is a 3×3×3 diagonal matrix;

[0105] Apply the calculated rotation matrix and translation vector to the high-altitude point cloud to update its posture and achieve the best registration effect;

[0106] The Mask R-CNN model is used to identify the locations of roads, fruit trees, and obstacles in the aerial images of the orchard and the images of roads and obstacles on the ground.

[0107] Collect RGB images of orchard environments covering different seasons, lighting conditions, and weather conditions and annotate the locations of roads, fruit trees, and obstacles;

[0108] Use professional annotation tools (such as LabelMe, VIA, CVAT) to annotate RGB images at the pixel level. The annotation content includes important types of objects such as roads, fruit trees, obstacles, etc. Each object should be annotated with its bounding box and pixel-level mask.

[0109] The annotated RGB images are cropped, flipped, rotated, and scaled to increase the generalization ability of the model, simulate different perspectives, and increase the robustness of the model to different scales and angles;

[0110] Normalize the pixel values ​​of the processed RGB images and divide them into training set, validation set and test set;

[0111] Input the training set into the Mask R-Mask R-CNN model for training;

[0112] The cross entropy loss is used to measure the difference between the prediction result of the classification branch and the true label. The expression is as follows:

[0113] ;

[0114] in, is the cross entropy loss value, is the number of training samples, is the index of the number of training samples, For the The true annotations of training samples, For the The predicted annotations of training samples;

[0115] The smooth L1 loss is used to measure the difference between the bounding box predicted by the regression branch and the true bounding box. The expression is as follows:

[0116] ;

[0117] in, To smooth the L1 loss value, For the The coordinates of the true bounding box of the training samples are usually expressed as Represent the center point coordinates, width and height of the bounding box respectively. For the The coordinates of the predicted bounding boxes for each training sample, represents the smoothed L1 loss function;

[0118] The binary cross entropy loss is used to measure the difference between the pixel-level mask generated by the mask branch and the true mask, which is expressed as follows:

[0119] ;

[0120] in, is the binary cross entropy loss value, is the height of the mask, that is, the vertical resolution of the mask image, is the index of the mask image row, ranging from 1 to , each row represents a horizontal scan line in the mask image, is the width of the mask, that is, the horizontal resolution of the mask image, The index of the mask image column, ranging from 1 to , each column represents a vertical scan line in the mask image, For the Line The true mask value of the column is 0 or 1, indicating whether the pixel belongs to the target object. For the Line The predicted mask value of the column indicates the model's confidence that the pixel belongs to the target object, and its value range is [0,1];

[0121] At the end of each training round, the validation set is input and the performance of the model is evaluated by calculating the average precision and intersection-over-union ratio;

[0122] Specifically, based on the recognition accuracy of the Mask R-CNN model for the locations of roads, fruit trees, and obstacles in the RGB image, an accuracy curve is drawn, and the area under the entire curve is calculated using the interpolation method to obtain the average accuracy;

[0123] Calculate the ratio of the intersection area and the union area of ​​the predicted bounding box and the true bounding box to get the intersection-union ratio, which ranges from [0,1]. The closer the intersection-union ratio is to 1, the higher the overlap between the predicted bounding box and the true bounding box.

[0124] Adjust hyperparameters (such as learning rate, regularization coefficient, etc.) based on the evaluation results of the validation set;

[0125] Specifically, it is common to start with a higher initial learning rate (such as 0.001) and gradually reduce it during training. For example, every fixed number of training rounds, the learning rate is multiplied by a factor less than 1 (such as 0.1);

[0126] By applying L2 regularization to the model's weights, excessive weights are penalized, thereby encouraging the model to learn simpler features. The value of the regularization coefficient is usually small (such as 0.0001). During the Mask R-CNN model training process, the regularization coefficient can be dynamically adjusted according to the performance of the Mask R-Mask R-CNN model on the training set and the validation set. For example, when the loss of the training set continues to decrease, but the loss of the validation set begins to increase, it means that the Mask R-Mask R-CNN model begins to overfit, and the regularization strength is increased. When the losses of both the training set and the validation set are high and there is no obvious downward trend, it means that the Mask R-MaskR-CNN model is underfitting, and the regularization strength is reduced.

[0127] When the Mask R-Mask R-CNN model completes the training of all RGB images in the training set, the training of the Mask R-Mask R-CNN model ends, and the performance of the Mask R-Mask R-CNN model is tested using the test set. When the performance of the validation set does not improve within 50 consecutive training rounds, the training is stopped in advance to prevent overfitting.

[0128] Use the trained Mask R-Mask R-CNN model to identify the locations of roads, fruit trees, and obstacles in aerial images of the orchard and images of ground roads and obstacles.

[0129] Use the Canny edge detection algorithm to extract edge information of roads, fruit trees, and obstacles from the recognition results;

[0130] Use the findContours function in the OpenCV library to extract the contours of roads, fruit trees, and obstacles from the recognition results;

[0131] According to the timestamps and corresponding point cloud data of the orchard aerial images and ground road and obstacle images, the edge information and contours of the extracted roads, fruit trees and obstacles are projected into the preliminary 3D model of the orchard to generate the final 3D model of the orchard;

[0132] Specifically, according to the timestamps of the aerial images of the orchard and the images of the ground roads and obstacles, the point cloud data collected at the same time are found, and the feature points in the two-dimensional image are converted into three-dimensional coordinates using the camera intrinsic parameter matrix and IMU data to ensure that the spatial position of the feature points is consistent with the geometric structure of the point cloud data. The ICP algorithm is used to align the edge information and contour three-dimensional coordinates of the roads, fruit trees, and obstacles with the preliminary three-dimensional model to ensure their precise spatial matching.

[0133] S3. Based on the three-dimensional model of the orchard, a path planning algorithm is used to generate multiple transportation paths.

[0134] Based on the distribution of fruit trees, roads and obstacles in the 3D model of the orchard, the orchard is logically partitioned and each partition is used as the starting point for transportation;

[0135] Specifically, the spatial coordinates of each fruit tree are extracted from the three-dimensional model of the orchard to establish a point set of the fruit tree;

[0136] A simple grid division method is used to divide the orchard into K sub-areas by taking the location distribution of roads in the orchard as the sub-area boundaries;

[0137] All intersections of roads are used as transport nodes;

[0138] Determine the transportation destination according to the fruit storage and transportation requirements;

[0139] Based on the transport path start point, transport node and end point, all transport paths from the transport start point to the end point are generated.

[0140] S4. Apply the quantum approximate optimization algorithm to optimize the transport paths of different transport starting points and output the optimal transport path.

[0141] The path optimization problem is modeled as a graph optimization problem using the quantum approximate optimization algorithm, and the orchard area is represented as a graph. ;

[0142] in, is the graph node set of the transport starting point, transport node and transport end point of all transport paths, is the set of edges connecting two graph nodes;

[0143] Define constraints based on transportation requirements and road conditions, including:

[0144] Each transport route must start from the transport starting point and end at the transport end point;

[0145] Each origin and destination must be connected by only one path;

[0146] The path must not cross obstacles;

[0147] For all transport paths with different transport starting points, the length of the path and the transport time are taken as optimization targets, and the objective function is set. The expression is:

[0148] ;

[0149] in, For graph nodes arrive Length, For graph nodes arrive The transportation time, Representing a graph node arrive Is there an obstacle between them? If there is an obstacle and the road is completely impassable, , when there are obstacles but the road allows the unmanned automated transport vehicle to pass, , when the road is clear, , and They are any two adjacent graph nodes in the graph node set;

[0150] Based on the objective function, the path optimization problem is converted into a quantum Hamiltonian, which is expressed as:

[0151] ;

[0152] ;

[0153] ;

[0154] in, is the target quantum Hamiltonian value, For graph nodes arrive The objective function value of is the diagonal element of the Pauli Z matrix corresponding to the quantum bit, indicating whether the graph node is selected arrive , if selected, then If not selected, , is the mixed Hamiltonian, indicating the target state flip, and They are any two adjacent graph nodes in the graph node set;

[0155] Construct a parameterized quantum circuit based on the target quantum Hamiltonian, and initialize all quantum bits to a uniform superposition state. The expression is as follows:

[0156] ;

[0157] in, is the initial quantum state, the uniform superposition state of all possible paths, is the number of quantum bits, corresponding to the number of optimization targets, Indicates the transport path;

[0158] In each layer of the quantum approximate optimization algorithm, the target Hamiltonian evolution and the mixed Hamiltonian evolution are applied in turn to iteratively update the quantum state. After Q layers of iteration, the quantum state evolves as follows:

[0159] ;

[0160] in, For passing The quantum state after layer evolution, To optimize the number of circuit layers, To optimize the index of the number of layers of the circuit, For the layer The control parameters of For the layer The control parameters of

[0161] For quantum states Take a measurement and get a result distribution, expressed as:

[0162] ;

[0163] in, For transport path The probability of occurrence;

[0164] The transport path with the highest probability in the measurement results is the optimal transport path for each starting point.

[0165] S5. The unmanned automated transport fleet collaborates to perform transport tasks based on the optimal transport path.

[0166] The task allocation priority is set according to the total length of the optimal path for each transport starting point and the location, loading status and remaining power of each unmanned automated transport vehicle;

[0167] Specifically, vehicles with sufficient power are given priority, and vehicles that are empty and closest to the transport starting point are given priority;

[0168] The unmanned automated transport vehicle uses cameras and LiDAR sensors to scan and detect the road ahead in real time;

[0169] When an obstacle is detected, the LiDAR sensor is used to scan the point cloud data of the obstacle and upload it to the control center to update the 3D model of the orchard;

[0170] At the same time, the location is marked and synchronized to all unmanned automated transport vehicles. After the control center re-optimizes the optimal transport path, they turn around and perform the transport task according to the new optimal transport path.

[0171] The present embodiment also provides an orchard intelligent transportation system, including: a data acquisition module, a three-dimensional model module, a transportation path module, a path optimization module and a transportation execution module, the data acquisition module is used to collect high-altitude data and ground data of the orchard from high altitude and ground respectively through unmanned aerial vehicles and ground robots; the three-dimensional model module is used to generate a three-dimensional model of the orchard by integrating the high-altitude data and ground data collected by the unmanned aerial vehicle and the ground robot; the transportation path module is used to generate multiple transportation paths based on the orchard three-dimensional model using a path planning algorithm; the path optimization module is used to apply a quantum approximate optimization algorithm to optimize the transportation paths with different transportation starting points and output the optimal transportation path; the transportation execution module is used for the unmanned automated transportation fleet to collaboratively execute the transportation task based on the optimal transportation path.

[0172] This embodiment also provides a computer device, which is suitable for the case of an intelligent handling method for an orchard, and includes: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the intelligent handling method for an orchard as proposed in the above embodiment.

[0173] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0174] The present embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for realizing intelligent orchard transportation as proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0175] In summary, the present invention achieves comprehensive coverage and accurate modeling of the orchard environment by: UAVs and ground robots collaborate to collect high-altitude and ground data, generate a high-precision three-dimensional model of the orchard, and provide a high-quality data foundation for path planning. Based on the three-dimensional model, a plurality of alternative paths are generated by a path planning algorithm, and combined with a quantum approximate optimization algorithm, the global optimal path of multi-starting point transportation tasks is efficiently determined, which significantly improves the efficiency and accuracy of path optimization, reduces transportation time and energy consumption, and in the transportation execution stage, through the collaborative scheduling of unmanned automated transportation fleets, tasks are allocated based on vehicle position, power level and other status, and combined with real-time obstacle detection and dynamic path adjustment functions, to ensure the continuity and stability of transportation tasks.

[0176] Example 2, referring to Table 1, is the second example of the present invention. To further verify the technical solution of the present invention, experimental simulation data of an orchard intelligent handling method are provided.

[0177] In order to verify the superiority of the "an intelligent transportation method for orchards" of the present invention, an experiment was designed for its core steps and compared with the existing technologies (the traditional path planning method based on the A* algorithm and the data collection method of a single ground robot).

[0178] The orchard area is 50 mu, with densely distributed fruit trees and terrain containing roads of varying slopes and area obstacles (such as fallen branches and mechanical equipment).

[0179] In the inventive method, a drone equipped with LiDAR and RGB camera (sampling frequency 100,000 points / second) and a ground robot (LiDAR sampling frequency 200,000 points / second) are used.

[0180] The control group used a traditional single ground robot to collect data, and the path planning was based on the A* algorithm.

[0181] The method of the present invention collects aerial and ground data of the orchard through drones and ground robots respectively. The drone focuses on covering the overall terrain and fruit tree distribution of the orchard, and the ground robot is responsible for collecting detailed information on roads and obstacle areas. The data is integrated to the control center through wireless transmission.

[0182] The present invention uses voxel filtering and ICP algorithm to generate a high-precision three-dimensional model of the orchard. The comparison group did not generate a three-dimensional model and only planned through a two-dimensional path map.

[0183] The method of the present invention combines the three-dimensional model and the quantum approximate optimization algorithm (QAOA) to generate the optimal path. The control group uses the A* algorithm for path planning.

[0184] The method of the present invention uses an unmanned automated fleet to complete the handling task based on the optimal path, and the fleet cooperates to adopt a dynamic obstacle avoidance mechanism. The comparison group uses a single transport vehicle, which lacks the ability to adjust the dynamic path.

[0185] The details are shown in Table 1 below:

[0186] Table 1 Experimental data comparison table

[0187]

[0188] By comparing the test data, we can clearly see the significant advantages of the method of the present invention:

[0189] The method of the present invention utilizes the quantum approximate optimization algorithm (QAOA) to significantly shorten the path planning time. It only takes 12 seconds to complete the path planning on complex terrain. The A algorithm of the comparison method takes 45 seconds, and the efficiency is improved by about 275%.

[0190] The present invention reduces the transportation time from 45 minutes of the comparative method to 28 minutes based on the collaborative operation of the optimal path and the automated fleet. The main reason is that the unmanned automated transportation fleet utilizes the optimal path for division of labor and collaboration, and combines the dynamic obstacle avoidance function to reduce the number of path detours (the present invention only detoured twice, while the comparative method detoured eight times).

[0191] Due to the optimized path and fleet collaboration, the total energy consumption of the method of the present invention is only 35 units, while the comparison method consumes 60 units of energy, with significant energy-saving effects. At the same time, the completion rate of the handling task is increased to 98%, while the comparison method only completes 80% due to path duplication and task interruption.

[0192] The method of the present invention achieves a data coverage rate of 95% through multi-source data fusion of drones and ground robots, which is significantly higher than the 65% of the comparative method. At the same time, the generated three-dimensional model not only improves the accuracy of path planning, but also provides real-time support for dynamic obstacle avoidance. The comparative method only relies on two-dimensional data and lacks a comprehensive understanding of the overall terrain of the orchard.

[0193] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An intelligent handling method for an orchard, characterized in that: include, Use drones and ground robots to collect aerial and ground data of the orchard from high altitude and ground respectively; By integrating the aerial data and ground data collected by drones and ground robots, a three-dimensional model of the orchard is generated; Based on the 3D model of the orchard, multiple transport paths are generated using a path planning algorithm; Apply quantum approximate optimization algorithm to optimize the transport paths of different transport starting points and output the optimal transport path; The unmanned automated transport fleet collaborates to perform transport tasks based on the optimal transport path; The quantum approximate optimization algorithm is applied to optimize the transport paths of different transport starting points and output the optimal transport path. The specific steps are as follows: The path optimization problem is modeled as a graph optimization problem using the quantum approximate optimization algorithm, and the orchard area is represented as a graph. ; in, is the graph node set of the transport starting point, transport node and transport end point of all transport paths, is the set of edges connecting two graph nodes; Define constraints based on transportation requirements and road conditions; For all transport paths with different transport starting points, the length of the transport path and the transport time are taken as optimization targets, and the objective function is set; Convert the path optimization problem into a quantum Hamiltonian based on the objective function , the expression is: ； ； ； in, is the target quantum Hamiltonian, Indicates the goal, For graph nodes arrive The objective function value of is the Pauli The diagonal elements of the matrix, is the mixed Hamiltonian, Indicates mixed, and They are any two adjacent graph nodes in the graph node set; Construct a parameterized quantum circuit based on the target quantum Hamiltonian, and initialize all quantum bits to a uniform superposition state. The expression is as follows: ； in, is the initial quantum state, is the number of quantum bits, Indicates the transport path; In each layer of the quantum approximate optimization algorithm, the target Hamiltonian evolution and the mixed Hamiltonian evolution are applied in turn to iteratively update the quantum state. Layer iteration, the quantum state evolves to: ； in, is the quantum state after the Q-layer evolution, To optimize the number of circuit layers, To optimize the index of the number of layers of the circuit, For the layer The control parameters of For the layer The control parameters of For quantum states Take measurements and get the result distribution, expressed as: ； in, For transport path The probability of occurrence; The transport path with the highest probability of appearing in the measurement results is the optimal transport path for each transport starting point.

2. The orchard intelligent transportation method according to claim 1, characterized in that: The specific steps of collecting the aerial data and ground data of the orchard from the high altitude and the ground by using the drone and the ground robot are as follows: A drone equipped with a LiDAR sensor and an RGB camera collects aerial data of the orchard from high altitude to obtain aerial point cloud data and aerial images of the orchard. On the ground, a route collection robot equipped with a LiDAR sensor and an RGB camera is used to drive along the roads in the orchard and collect ground data on the road distribution and obstacles along the way, obtaining point cloud data and images of ground roads and obstacles. The collected high-altitude data and ground data are uploaded to the control center through wireless transmission.

3. The intelligent orchard transport method according to claim 2, characterized in that: The three-dimensional model of the orchard is generated by integrating the high-altitude data and ground data collected by the drone and the ground robot. The specific steps are as follows: In the control center, the voxel filtering method is used to filter the orchard high-altitude point cloud data and the ground road and obstacle point cloud data collected from the high altitude and the ground; Using the ICP algorithm, the aerial point cloud data and the ground point cloud data were initially fused and registered to obtain a preliminary 3D model of the orchard. The Mask R-CNN model is used to identify the locations of roads, fruit trees, and obstacles in the aerial images of the orchard and the images of roads and obstacles on the ground. Use the Canny edge detection algorithm to extract edge information of roads, fruit trees, and obstacles from the recognition results; Use the findContours function in the OpenCV library to extract the contours of roads, fruit trees, and obstacles from the recognition results; According to the timestamps and corresponding point cloud data of the orchard aerial images and the ground road and obstacle images, the edge information and contours of the extracted roads, fruit trees and obstacles are projected into the preliminary 3D model of the orchard to generate the final 3D model of the orchard.

4. The intelligent orchard transport method according to claim 3, characterized in that: The Mask R-CNN model is used to identify the locations of roads, fruit trees and obstacles in the orchard aerial images and ground road and obstacle images. The specific steps are as follows: Collect RGB images of orchard environments covering different seasons, lighting conditions, and weather conditions and annotate the locations of roads, fruit trees, and obstacles; Crop, flip, rotate and scale the annotated RGB image; Normalize the pixel values ​​of the processed RGB images and divide them into training set, validation set and test set; Input the training set into the Mask R-Mask R-CNN model for training; The cross entropy loss is used to measure the difference between the prediction result of the classification branch and the actual label; Smooth L1 loss is used to measure the difference between the bounding box predicted by the regression branch and the true bounding box; Use binary cross entropy loss to measure the difference between the pixel-level mask generated by the mask branch and the true mask; At the end of each training round, the validation set is input and the performance of the model is evaluated by calculating the average precision and intersection-over-union ratio; Adjust hyperparameters based on the evaluation results of the validation set; When the Mask R-Mask R-CNN model completes the training of all RGB images in the training set, the training of the Mask R-Mask R-CNN model ends, and the performance of the Mask R-Mask R-CNN model is tested using the test set. The trained Mask R-Mask R-CNN model is used to identify the locations of roads, fruit trees, and obstacles in aerial images of orchards and images of ground roads and obstacles.

5. The orchard intelligent transportation method according to claim 4, characterized in that: Based on the orchard three-dimensional model, a path planning algorithm is used to generate multiple transport paths. The specific steps are as follows: Based on the distribution of fruit trees, roads and obstacles in the 3D model of the orchard, the orchard is logically partitioned and each partition is used as the starting point for transportation; All intersections of roads are used as transport nodes; Determine the transportation destination according to the fruit storage and transportation requirements; Based on the transport start point, the transport node and the transport end point of the transport path, all transport paths from the transport start point to the transport end point are generated.

6. The orchard intelligent transportation method according to claim 1, characterized in that: The unmanned automated transport fleet collaboratively performs the transport task based on the optimal transport path. The specific steps are as follows: The task allocation priority is set according to the total length of the optimal transport path for each transport starting point and the position, loading status and remaining power of each unmanned automated transport vehicle; The unmanned automated transport vehicle uses cameras and LiDAR sensors to scan and detect the road ahead in real time; When an obstacle is detected, the LiDAR sensor is used to scan the point cloud data of the obstacle and upload it to the control center to update the 3D model of the orchard; At the same time, the obstacle location is marked and synchronized to all unmanned automated transport vehicles. After the control center re-optimizes the optimal transport path, they turn around and perform the transport task according to the new optimal transport path.

7. An orchard intelligent handling system, based on the orchard intelligent handling method according to any one of claims 1 to 6, characterized in that: It includes data acquisition module, 3D model module, transport path module, path optimization module and transport execution module. The data acquisition module is used to collect the high-altitude data and the ground data of the orchard from the high altitude and the ground respectively through the unmanned aerial vehicle and the ground robot; The three-dimensional model module is used to generate a three-dimensional model of the orchard by fusing the high-altitude data and ground data collected by the drone and the ground robot; The transport path module is used to generate multiple transport paths based on the orchard three-dimensional model using a path planning algorithm; The path optimization module is used to apply the quantum approximate optimization algorithm to optimize the transport paths of different transport starting points and output the optimal transport path; The execution transport module is used for the unmanned automated transport fleet to collaboratively execute transport tasks based on the optimal transport path.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the orchard intelligent transportation method described in any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the orchard intelligent transportation method described in any one of claims 1 to 6 are implemented.

Images