Method and system for intelligently planning aerial photography route for inner cultivated land investigation

Abstract

The invention provides a method and a system for intelligently planning an aerial photography route for inner cultivated land investigation. The method comprises the following steps of: preprocessing obtained multiple pieces of to-be-analyzed unmanned aerial vehicle cruise route data covering multi-layer inner cultivated land investigation data description knowledge; performing simplification processing to obtain target unmanned aerial vehicle cruise route data covering the main description knowledge; and finally, loading the target unmanned aerial vehicle cruise route data to a flight path intelligent planning thread, and processing the target unmanned aerial vehicle cruise route data through the flight path intelligent planning thread to obtain a flight path intelligent planning result corresponding to the unmanned aerial vehicle cruise route data to be analyzed, the flight path intelligent planning thread is obtained based on training of the autonomous training unit and the active debugging unit. On one hand, the flight path intelligent planning thread can be trained based on the autonomous training unit and the active debugging unit, the accuracy of flight path planning is improved, the accuracy of cultivated land inspection is improved, and the reliability of data inspection is guaranteed.

Description

Technical Field

[0001] The present application relates to the technical field of data planning, and more specifically, to a method and system for intelligent planning of aerial photography routes for internal cultivated land surveys. Background Art

[0002] With the continuous development and progress of unmanned aerial vehicles (UAVs), the fields to which UAV technology is applied have become more extensive. When UAV technology is specifically applied to internal cultivated land surveys, it is necessary for humans to plan the flight routes of UAVs. In traditional technologies, the flight path planning is set by relevant technical personnel, which may cause some problems, resulting in inaccurate monitoring or incomplete monitoring. Therefore, there is an urgent need for an intelligent planning method for aerial photography routes to improve the above technical problems. Summary of the Invention

[0003] To improve the technical problems existing in related technologies, the present application provides a method and system for intelligent planning of aerial photography routes for internal cultivated land surveys.

[0004] In a first aspect, a method for intelligent planning of aerial photography routes for internal cultivated land surveys is provided. The method includes: Obtaining a plurality of drone cruise route data to be analyzed that cover multi-level knowledge descriptions of internal cultivated land survey data, and preprocessing the drone cruise route data to be analyzed; Performing simplification processing on the preprocessed drone cruise route data to be analyzed to obtain target drone cruise route data covering main description knowledge; Loading the target drone cruise route data into a flight trajectory intelligent planning thread, and processing the target drone cruise route data through the flight trajectory intelligent planning thread to obtain a flight trajectory intelligent planning result corresponding to the drone cruise route data to be analyzed, where the flight trajectory intelligent planning thread is trained based on an autonomous training unit and an active debugging unit.

[0005] In the present application, the preprocessing of the drone cruise route data to be analyzed includes: Processing the loss value in the drone cruise route data to be analyzed, and performing normalization processing on the processed drone cruise route data to be analyzed.

[0006] In the present application, the processing of the loss value in the drone cruise route data to be analyzed includes: When the description knowledge type of the loss value is quantitative numerical description knowledge, determining a depolarization value, an intermediate value, or a high-heat value according to the drone cruise route data to be analyzed, and compensating the loss value with the depolarization value, the intermediate value, or the high-heat value; When the description knowledge type of the loss value is high-loss value description knowledge, the category with the highest loss degree in the to-be-analyzed UAV cruise route data is used to compensate the loss value.

[0007] In this application, the normalization processing of the processed to-be-analyzed UAV cruise route data includes: Scaling the processed to-be-analyzed UAV cruise route data to a target numerical range according to a specified method.

[0008] In this application, the simplification processing of the preprocessed to-be-analyzed UAV cruise route data to obtain target UAV cruise route data covering the main description knowledge includes: Performing benchmarking processing on the preprocessed to-be-analyzed UAV cruise route data, building a pyramid queue according to the benchmarked to-be-analyzed UAV cruise route data and the corresponding description knowledge, and determining the description knowledge coefficient and description knowledge vector of the pyramid queue; Forming a description knowledge coefficient matrix by arranging the description knowledge coefficients in a set sorting manner, sequentially obtaining target description knowledge vectors corresponding to a preset number of description knowledge coefficients in the description knowledge coefficient matrix, and building a principal component description knowledge matrix according to the target description knowledge vectors; Determining a first simplification matrix according to the benchmarked to-be-analyzed UAV cruise route data and the principal component description knowledge matrix, selecting the description knowledge in the first simplification matrix to obtain a second simplification matrix, and determining the data in the second simplification matrix as the target UAV cruise route data.

[0009] In this application, before loading the target UAV cruise route data into the flight trajectory intelligent planning thread, the method further includes: Obtaining a set of cruise trajectory example data, and training the to-be-trained flight trajectory intelligent planning thread according to each group of cruise trajectory example data in the set of cruise trajectory example data to obtain the flight trajectory intelligent planning thread.

[0010] In this application, the training of the to-be-trained flight trajectory intelligent planning thread according to each group of cruise trajectory example data in the set of cruise trajectory example data to obtain the flight trajectory intelligent planning thread includes: Performing preprocessing on the cruise trajectory example data, and performing simplification processing on the preprocessed cruise trajectory example data to obtain target cruise trajectory example data covering the main description knowledge; Divide the cruise trajectory example data in the cruise trajectory example data set into a first cruise trajectory example data subset and a second cruise trajectory example data subset according to the flight label, and build an unannotated data set based on the target cruise trajectory example data in the first cruise trajectory example data subset; Detect the abnormal flight trajectory description coefficient of the target cruise trajectory example data in the second cruise trajectory example data subset, and build an annotated data set according to the target cruise trajectory example data in the second cruise trajectory example data subset and the corresponding directory; Train the to-be-trained flight trajectory intelligent planning thread according to the unannotated data set and the annotated data set to obtain the flight trajectory intelligent planning thread.

[0011] In this application, the detecting the abnormal flight trajectory description coefficient of the target cruise trajectory example data in the second cruise trajectory example data subset includes: Identify the target cruise trajectory example data in the second cruise trajectory example data subset, and determine a random target cruise trajectory example data as the target flight trajectory data; Determine the a-difference value according to the target flight trajectory data and the a-th nearest neighbor data corresponding to the target flight trajectory data, where a is a positive integer; Obtain the real-time difference value between the target flight trajectory data and the a-th nearest neighbor data, and determine the target difference value corresponding to the target flight trajectory data according to the a-difference value and the real-time difference value; Determine the local confidence corresponding to the target flight trajectory data according to the target difference values corresponding to the a nearest neighbor data of the target flight trajectory data, and determine the local influence factors corresponding to the target flight trajectory data based on the local confidence; Judge whether the target flight trajectory data is cultivated land abnormal data according to the local influence factors and the preset specified value, and determine the directory corresponding to the flight label covering the target flight trajectory data according to the judgment result.

[0012] In this application, the training the to-be-trained flight trajectory intelligent planning thread according to the unannotated data set and the annotated data set to obtain the flight trajectory intelligent planning thread includes: Actively select the data with the largest amount of information from the unannotated data set, determine the directory corresponding to the data with the largest amount of information through manual annotation, and update the annotated data set according to the data with the largest amount of information and the directory; Train the to-be-trained intelligent flight trajectory planning thread according to the updated annotation dataset, and optimize the parameters of the to-be-trained intelligent flight trajectory planning thread through the autonomous training unit during the training process; Repeat the above process until the pre-set training requirements are met to obtain the intelligent flight trajectory planning thread.

[0013] In a second aspect, a system for intelligent planning of aerial photography routes for inner cultivated land surveys is provided, including a processor and a memory that communicate with each other. The processor is configured to read and execute a computer program from the memory to implement the above method.

[0014] A method and system for intelligent planning of aerial photography routes for inner cultivated land surveys provided by an embodiment of the present application preprocess the obtained multiple pieces of to-be-analyzed UAV cruise route data covering multi-level inner cultivated land survey data description knowledge; then perform simplification processing to obtain target UAV cruise route data covering the main description knowledge; finally, load the target UAV cruise route data into the intelligent flight trajectory planning thread, and process the target UAV cruise route data through the intelligent flight trajectory planning thread to obtain an intelligent flight trajectory planning result corresponding to the to-be-analyzed UAV cruise route data, where the intelligent flight trajectory planning thread is trained based on an autonomous training unit and an active debugging unit. On the one hand, the present application can train the intelligent flight trajectory planning thread based on the autonomous training unit and the active debugging unit, improve the accuracy of flight trajectory planning, improve the accuracy of cultivated land inspection, and ensure the reliability of data inspection. Description of the Drawings

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings required in the embodiments. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0016] Figure 1 It is a flowchart of a method for intelligent planning of aerial photography routes for inner cultivated land surveys provided by an embodiment of the present application. Detailed Embodiments

[0017] In order to better understand the above technical solutions, the following will provide a detailed description of the technical solutions of the present application through the drawings and specific embodiments. It should be understood that the embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical solutions of the present application, rather than limitations on the technical solutions of the present application. Without conflict, the technical features in the embodiments of the present application and the embodiments can be combined with each other.

[0018] Please refer toFigure 1 shows a method for intelligent planning of aerial photography routes for internal cultivated land surveys, and the method may include the technical solutions described in the following steps S210 to step S230.

[0019] Step S210: Obtain a plurality of drone cruise route data to be analyzed that cover multi-level knowledge descriptions of internal cultivated land survey data, and preprocess the drone cruise route data to be analyzed; Exemplarily, the description knowledge is understood as features in this application.

[0020] Among them, the drone cruise route data to be analyzed can be understood as a flight trajectory set through internal cultivated land survey data.

[0021] The specific processing steps of preprocessing in this application include: analyzing and screening the drone cruise route data to be analyzed, simply processing it, and deleting some incorrect or missing data.

[0022] Step S220: Simplify the preprocessed drone cruise route data to be analyzed to obtain target drone cruise route data that covers the main description knowledge; Exemplarily, the main description knowledge is understood as important features.

[0023] Among them, the simplification process can be understood as processing the data in a way such as dimensionality reduction.

[0024] Step S230: Load the target drone cruise route data into the flight trajectory intelligent planning thread, and process the target drone cruise route data through the flight trajectory intelligent planning thread to obtain a flight trajectory intelligent planning result corresponding to the drone cruise route data to be analyzed, where the flight trajectory intelligent planning thread is trained based on an autonomous training unit and an active debugging unit.

[0025] Exemplarily, the flight trajectory intelligent planning thread is a type of artificial intelligence thread, which includes processing methods such as a cnn processing thread.

[0026] In the method for intelligent planning of aerial photography routes for internal cultivated land surveys provided in the embodiments of the present application, preprocessing is performed on the obtained multiple pieces of drone cruise route data to be analyzed that cover multi-level internal cultivated land survey data description knowledge; then, simplification processing is carried out to obtain target drone cruise route data covering the main description knowledge; finally, the target drone cruise route data is loaded into the flight trajectory intelligent planning thread, and the flight trajectory intelligent planning thread processes the target drone cruise route data to obtain a flight trajectory intelligent planning result corresponding to the drone cruise route data to be analyzed, where the flight trajectory intelligent planning thread is trained based on an autonomous training unit and an active debugging unit. On the one hand, the present application can train the flight trajectory intelligent planning thread based on the autonomous training unit and the active debugging unit, improve the accuracy of flight trajectory planning, improve the accuracy of cultivated land inspection, and ensure the reliability of data inspection. The following will detail the specific implementation methods of each method step in the method for intelligent planning of aerial photography routes for internal cultivated land surveys in the embodiments of the present application.

[0027] In step S210, multiple pieces of drone cruise route data to be analyzed that cover multi-level internal cultivated land survey data description knowledge are obtained, and preprocessing is performed on the drone cruise route data to be analyzed.

[0028] In an embodiment of the present application, in order to improve the accuracy of anomaly detection for high-dimensional drone cruise route data, first, multiple pieces of drone cruise route data to be analyzed that cover multi-level internal cultivated land survey data description knowledge can be obtained and preprocessed to provide a basis for subsequent detection of cultivated land anomaly data.

[0029] In an embodiment of the present application, preprocessing the drone cruise route data to be analyzed mainly focuses on processing the loss value. For different types of description knowledge, the processing methods for the loss value are different. When the description knowledge type of the loss value is quantitative numerical description knowledge, the depolarization value, intermediate value, or high-heat value is determined according to the drone cruise route data to be analyzed, and the loss value is compensated with the depolarization value, intermediate value, or high-heat value; when the description knowledge type of the loss value is high-loss value description knowledge, the category with the highest loss degree in the drone cruise route data to be analyzed is used to compensate the loss value. For example, when a category is missing, the occurrence losses of each description knowledge category in other groups of drone cruise route data to be analyzed can be counted, and the loss value is compensated according to the description knowledge category with the highest occurrence loss. It should be noted that the categories in the embodiments of the present application are information in different dimensions of the flight label.

[0030] Furthermore, the to-be-analyzed UAV cruise route data after loss value compensation can be normalized, and the processed to-be-analyzed UAV cruise route data can be scaled into a target numerical range according to a specified method. The target numerical range is [0, 1] or [-1, 1], which can avoid the influence of individual data on the subsequent detection of cultivated land anomaly data.

[0031] In an embodiment of the present application, after compensating for the loss value, the to-be-analyzed UAV cruise route data can also be processed with an abnormal flight trajectory description coefficient to initially identify the abnormal flight trajectory description coefficient in the to-be-analyzed UAV cruise route data, and replace it with the depolarized value of the upper and lower bounds in the normal value range or delete the abnormal flight trajectory description coefficient. By processing the abnormal flight trajectory description coefficient, the data processing volume for subsequent anomaly analysis can be reduced, the data processing efficiency is improved, and the cost is reduced. After completing the processing of the abnormal flight trajectory description coefficient, the data can be normalized and scaled into the target numerical range.

[0032] In step S220, the preprocessed to-be-analyzed UAV cruise route data is simplified to obtain target UAV cruise route data covering the main description knowledge.

[0033] In an embodiment of the present application, since the to-be-analyzed UAV cruise route data is high-dimensional data, in order to improve the data processing efficiency, anomaly detection can be performed on the principal component data in the to-be-analyzed UAV cruise route data, and this principal component is the relatively important description knowledge in the to-be-analyzed UAV cruise route data.

[0034] The process of the simplification processing in the present application at least covers steps S301 - S303, specifically: In step S301, the preprocessed to-be-analyzed UAV cruise route data is benchmarked, a pyramid queue is built based on the benchmarked to-be-analyzed UAV cruise route data and the corresponding description knowledge, and the description knowledge coefficient and description knowledge vector of the pyramid queue are determined.

[0035] In an embodiment of the present application, all the to-be-analyzed UAV cruise route data is built into an original matrix according to the corresponding description knowledge. The size of the original matrix is n×p, where n is the data volume and p is the number of description knowledge. Then, benchmarking processing is performed based on the original matrix to make the mean of the data zero, and the benchmarking processing is achieved by subtracting each data from the mean of the corresponding description knowledge.

[0036] In step S302, the description knowledge coefficients are formed into a description knowledge coefficient matrix according to a set sorting method, and the target description knowledge vectors corresponding to the preset number of description knowledge coefficients in the description knowledge coefficient matrix are obtained in sequence. A main description knowledge matrix is constructed based on the target description knowledge vectors.

[0037] In an embodiment of the present application, after obtaining all the description knowledge coefficients of the pyramid queue C, they can be arranged in descending order to form a description knowledge coefficient matrix, and a largest description knowledge coefficients are extracted from the description knowledge coefficient matrix in sequence. The description knowledge vectors corresponding to the a largest description knowledge coefficients are determined as the target description knowledge vectors, and a principal component description knowledge matrix is constructed based on the target description knowledge vectors. The size of the principal component description knowledge matrix is p×a. That is to say, the description knowledge vectors corresponding to the a largest description knowledge coefficients are the description knowledge vectors corresponding to the required principal components. Among them, a is a non-zero hyperparameter and is much smaller than the total number of description knowledge corresponding to the drone cruise route data to be analyzed, and can be adjusted according to actual needs.

[0038] In step S303, a first simplified matrix is determined based on the benchmarked drone cruise route data to be analyzed and the principal component description knowledge matrix. The description knowledge in the first simplified matrix is selected to obtain a second simplified matrix, and the data in the second simplified matrix is determined as the target drone cruise route data.

[0039] In an embodiment of the present application, although the first simplified matrix is determined based on the principal component description knowledge matrix, in order to further reduce the data processing volume, it is also necessary to select the description knowledge in the first simplified matrix to obtain a second simplified matrix.

[0040] In step S230, the target drone cruise route data is loaded into the flight trajectory intelligent planning thread, and the target drone cruise route data is processed by the flight trajectory intelligent planning thread to obtain a flight trajectory intelligent planning result corresponding to the drone cruise route data to be analyzed. Among them, the flight trajectory intelligent planning thread is trained based on an autonomous training unit and an active debugging unit.

[0041] In one embodiment of the present application, the drone cruise route data covered in the second simplified matrix is the target drone cruise route data. By detecting the target drone cruise route data through the trained flight trajectory intelligent planning thread, the cultivated land anomaly data therein can be obtained, and the flight trajectory intelligent planning results corresponding to different flight labels can be obtained. When at least one data in the drone cruise route data is detected as cultivated land anomaly data, the flight label corresponding to the cultivated land anomaly data is the abnormal flight label. According to the type of the cultivated land anomaly data, the flight trajectory intelligent planning result corresponding to the abnormal flight label can be determined, such as account theft, brush orders, false comments, change of shopping habits, and so on. The flight trajectory intelligent planning thread in the present application can be any classification model, such as decision tree, random forest, etc. The embodiments of the present application do not make specific limitations thereto.

[0042] In one embodiment of the present application, before using the flight trajectory intelligent planning thread to perform anomaly detection on the target drone cruise route data, it is also necessary to train the flight trajectory intelligent planning thread to be trained to obtain a converged and stable flight trajectory intelligent planning thread to ensure the accuracy of the anomaly detection result. In the embodiments of the present application, an active debugging unit and an autonomous training unit can be used for model training. Next, the model training process will be described in detail.

[0043] In one embodiment of the present application, first, a set of cruise trajectory example data is obtained, and then each group of cruise trajectory example data covered in the set of cruise trajectory example data is used to train the flight trajectory intelligent planning thread to be trained to obtain the flight trajectory intelligent planning thread.

[0044] In one embodiment of the present application, the cruise trajectory example data set includes an annotated data set and an unannotated data set. The annotated data set includes cruise trajectory example data and an abnormal directory corresponding to the cruise trajectory example data, and the unannotated data set only includes cruise trajectory example data. For the process of obtaining the cruise trajectory example data set, in step S401, the drone cruise route data corresponding to multiple flight tags is obtained from an e-commerce platform and determined as cruise trajectory example data; in step S402, all the cruise trajectory example data is preprocessed, and the loss value in the cruise trajectory example data is compensated. The loss value includes the absence of quantitative numerical description knowledge or high-loss value description knowledge. When the loss value is quantitative numerical description knowledge, the depolarization value, intermediate value or high-heat value of the cruise trajectory example data is used for filling. When the loss value is high-loss value description knowledge, the category with the highest loss degree in other cruise trajectory example data is used for filling. After completing the loss value compensation, the data is normalized and scaled to the target flight trajectory data range; then, all the scaled cruise trajectory example data is simplified to obtain target cruise trajectory example data covering the main description knowledge; in step S403, the target cruise trajectory example data is divided into a first cruise trajectory example data subset and a second cruise trajectory example data subset according to the corresponding flight tags, and an unannotated data set is built according to the target cruise trajectory example data in the first cruise trajectory example data subset; in step S404, the abnormal flight trajectory description coefficient of the target cruise trajectory example data in the second cruise trajectory example data subset is detected, and an annotated data set is built according to the target cruise trajectory example data in the second cruise trajectory example data subset and the corresponding directory; in step S405, a cruise trajectory example data set is built according to the unannotated data set and the annotated data set.

[0045] Among them, in step S404, when detecting the abnormal flight trajectory description coefficient of the target cruise trajectory example data, it can be implemented by means of manual annotation or detector detection, and the embodiments of the present application do not make specific limitations on this.

[0046] The process of the detector detecting the abnormal flight trajectory description coefficient of the target cruise trajectory example data is given. This process includes at least steps S501-S505, which are specifically as follows: In step S501, the target cruise trajectory example data in the second cruise trajectory example data subset is identified, and a random one of the target cruise trajectory example data is determined as the target flight trajectory data.

[0047] In an embodiment of the present application, the second cruise trajectory example data subset includes a plurality of data corresponding to different main description knowledges. As long as there is abnormal data in each flight tag, the corresponding flight tag is also abnormal. Therefore, it is necessary to perform anomaly detection on each data in the second cruise trajectory example data subset. Correspondingly, any target cruise trajectory example data in the second cruise trajectory example data subset can be determined as target flight trajectory data, and anomaly flight trajectory description coefficient detection can be performed on it.

[0048] In step S502, an a-difference value is determined according to the target flight trajectory data and the a-th nearest neighbor data corresponding to the target flight trajectory data, where a is a positive integer.

[0049] In an embodiment of the present application, for the target flight trajectory data xi, the difference value between it and other sample data in the target cruise trajectory example data can be calculated. According to the calculated difference value, the a-th nearest neighbor data corresponding to the target flight trajectory data is determined, and the nearest neighbor difference value a-dist(xi) between the target flight trajectory data and the a-th nearest neighbor data is obtained.

[0050] In step S503, a real-time difference value between the target flight trajectory data and the a-th nearest neighbor data is obtained, and a target difference value corresponding to the target flight trajectory data is determined according to the a-difference value and the real-time difference value.

[0051] In an embodiment of the present application, a target difference value corresponding to the target flight trajectory data can be obtained. The target difference value represents at least how far it takes to go from the target flight trajectory data xi to the a-th nearest neighbor data xj.

[0052] In step S504, a local confidence corresponding to the target flight trajectory data is determined according to the target difference value corresponding to the a nearest neighbor data of the target flight trajectory data, and a local influence factor corresponding to the target flight trajectory data is determined based on the local confidence.

[0053] In an embodiment of the present application, based on step S502 and step S503, the target difference value between the target flight trajectory data xi and its a nearest neighbor data can be obtained. Based on the target difference value between the target flight trajectory data xi and its a nearest neighbor data, the local confidence of the target flight trajectory data xi can be calculated.

[0054] In step S505, it is determined whether the target flight trajectory data is cultivated land abnormal data according to the local influence factor and a preset specified value, and a directory corresponding to the target flight trajectory data is determined according to the judgment result.

[0055] In an embodiment of the present application, the local influence factor represents the relative density of the target flight trajectory data xi and other points within its neighborhood. After obtaining the local influence factor of the target flight trajectory data xi, the local influence factor can be compared with a pre-set specified value, and it can be determined whether the target flight trajectory data is cultivated land anomaly data according to the comparison result. Furthermore, the directory of the flight label covering the target flight trajectory data can be determined according to the anomaly detection result of the target flight trajectory data.

[0056] In an embodiment of the present application, when training the to-be-trained intelligent flight trajectory planning thread with each group of cruise trajectory example data covered in the cruise trajectory example data set, the active debugging unit and the self-training unit can be used for training. The self-training unit iteratively executes along with the iteration of active learning. That is to say, in each round of training process, the model parameters are determined based on the self-training unit for model optimization.

[0057] After obtaining the annotated data set and the unannotated data set, one or more data with the largest amount of information can be actively selected from the unannotated data set. Then, the data with the largest amount of information is manually annotated to determine the corresponding anomaly directory. Then, the data with the largest amount of information and the corresponding anomaly directory are added to the annotated data set, and the to-be-trained intelligent flight trajectory planning thread is trained with the annotated data set. During the training process, the self-training unit is used to determine the model parameters to optimize the model until the pre-set number of training times is completed or the model performance reaches the specified value, and then the training is stopped, and the final intelligent flight trajectory planning thread can be obtained.

[0058] In an embodiment of the present application, the data with the largest amount of information and the corresponding anomaly directory can be added to the annotated data set, and then the to-be-trained intelligent flight trajectory planning thread is trained according to the annotated data set. During the training process, the model parameters can be determined based on the self-training unit. Next, the method for the self-training unit to determine the model parameters will be described in detail.

[0059] After determining the model parameters through the self-training unit, the model parameters can be assigned to the to-be-trained intelligent flight trajectory planning thread, and then the to-be-trained intelligent flight trajectory planning thread is trained again according to the active debugging unit until the pre-set number of training times is reached or the model performance reaches the specified value.

[0060] In an embodiment of the present application, after performing anomaly detection on the to-be-analyzed UAV cruise route data through the intelligent flight trajectory planning thread, the detection result can be output and displayed on the display interface of the terminal device.

[0061] The method for intelligent planning of aerial photography routes for internal cultivated land surveys provided by the embodiments of the present application preprocesses the obtained multiple pieces of drone cruise route data to be analyzed that cover multi-level internal cultivated land survey data description knowledge; then performs simplification processing to obtain target drone cruise route data covering the main description knowledge; finally, loads the target drone cruise route data into the flight trajectory intelligent planning thread, and processes the target drone cruise route data through the flight trajectory intelligent planning thread to obtain a flight trajectory intelligent planning result corresponding to the drone cruise route data to be analyzed, wherein the flight trajectory intelligent planning thread is trained based on an independent training unit and an active debugging unit. On the one hand, the present application can train the flight trajectory intelligent planning thread based on the independent training unit and the active debugging unit, improve the accuracy of flight trajectory planning, improve the accuracy of cultivated land inspection, and ensure the reliability of data inspection.

[0062] On this basis, a device for intelligent planning of aerial photography routes for internal cultivated land surveys is provided. The device includes: A data preprocessing module, configured to obtain multiple pieces of drone cruise route data to be analyzed that cover multi-level internal cultivated land survey data description knowledge, and preprocess the drone cruise route data to be analyzed; A data simplification module, configured to perform simplification processing on the preprocessed drone cruise route data to be analyzed to obtain target drone cruise route data covering the main description knowledge; A result planning module, configured to load the target drone cruise route data into the flight trajectory intelligent planning thread, and process the target drone cruise route data through the flight trajectory intelligent planning thread to obtain a flight trajectory intelligent planning result corresponding to the drone cruise route data to be analyzed, wherein the flight trajectory intelligent planning thread is trained based on an independent training unit and an active debugging unit.

[0063] On this basis, a system for intelligent planning of aerial photography routes for internal cultivated land surveys is shown, including a processor and a memory that communicate with each other. The processor is configured to read and execute a computer program from the memory to implement the above method.

[0064] On this basis, a computer-readable storage medium is further provided, and the computer program stored thereon implements the above method when running.

[0065] In summary, based on the above solution, preprocess the obtained multiple pieces of drone cruise route data to be analyzed that cover the description knowledge of cultivated land survey data at multiple levels; then perform simplification processing to obtain the target drone cruise route data covering the main description knowledge; finally, load the target drone cruise route data into the flight trajectory intelligent planning thread, and process the target drone cruise route data through the flight trajectory intelligent planning thread to obtain the flight trajectory intelligent planning result corresponding to the drone cruise route data to be analyzed, where the flight trajectory intelligent planning thread is trained based on the autonomous training unit and the active debugging unit. On the one hand, this application can train the flight trajectory intelligent planning thread based on the autonomous training unit and the active debugging unit, improve the accuracy of flight trajectory planning, improve the accuracy of cultivated land inspection, and ensure the reliability of data inspection.

[0066] It should be understood that the above-described systems and their modules can be implemented in various ways. For example, in some embodiments, the systems and their modules can be implemented through hardware, software, or a combination of software and hardware. Among them, the hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. Those skilled in the art can understand that the above methods and systems can be implemented using computer-executable instructions and / or included in processor control code, for example, such code is provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The systems and their modules of this application can be implemented not only by hardware circuits of programmable hardware devices such as very large scale integrated circuits or gate arrays, semiconductors such as logic chips and transistors, or field programmable gate arrays and programmable logic devices, but also by software executed by various types of processors, or by a combination of the above hardware circuits and software (for example, firmware).

[0067] It should be noted that the beneficial effects that may be produced by different embodiments are different. In different embodiments, the beneficial effects that may be produced can be any one or several combinations of the above, or any other beneficial effects that may be obtained.

Claims

1. A method for intelligent planning of aerial photography routes for inland cultivated land survey, characterized in that: include: Obtaining a plurality of unmanned aerial vehicle cruise route data to be analyzed that covers multi-level internal cultivated land survey data description knowledge, and preprocessing the unmanned aerial vehicle cruise route data to be analyzed; Simplifying the preprocessed UAV cruise route data to be analyzed to obtain target UAV cruise route data covering the main description knowledge; The target UAV cruise route data is loaded into the flight trajectory intelligent planning thread, and the target UAV cruise route data is processed by the flight trajectory intelligent planning thread to obtain a flight trajectory intelligent planning result corresponding to the UAV cruise route data to be analyzed, wherein the flight trajectory intelligent planning thread is obtained based on the training of the autonomous training unit and the active debugging unit.

2. The method according to claim 1, characterized in that The preprocessing of the drone cruise route data to be analyzed includes: The loss values ​​in the unmanned aerial vehicle cruise route data to be analyzed are processed, and the processed unmanned aerial vehicle cruise route data to be analyzed are normalized.

3. The method according to claim 2, characterized in that The processing of the loss value in the cruise route data of the unmanned aerial vehicle to be analyzed includes: When the description knowledge type of the loss value is quantitative numerical description knowledge, a depolarization value, an intermediate value or a high heat value is determined according to the cruise route data of the unmanned aerial vehicle to be analyzed, and the depolarization value, the intermediate value or the high heat value is used to compensate the loss value; When the description knowledge type of the loss value is high loss value description knowledge, the loss value is compensated by using the category with the highest loss degree in the cruise route data of the unmanned aerial vehicle to be analyzed.

4. The method according to claim 2, characterized in that: The step of normalizing the processed drone cruise route data to be analyzed includes: The processed drone cruise route data to be analyzed is scaled to a target value range according to a specified method.

5. The method according to claim 1, characterized in that The pre-processed UAV cruise route data to be analyzed is simplified to obtain the target UAV cruise route data covering the main description knowledge, including: Benchmarking the preprocessed UAV cruise route data to be analyzed, building a pyramid queue according to the benchmarked UAV cruise route data to be analyzed and the corresponding description knowledge, and determining the description knowledge coefficient and description knowledge vector of the pyramid queue; The description knowledge coefficients are arranged into a description knowledge coefficient matrix according to a set sorting method, target description knowledge vectors corresponding to a predetermined number of description knowledge coefficients in the description knowledge coefficient matrix are obtained in sequence, and a principal component description knowledge matrix is ​​constructed according to the target description knowledge vectors; A first simplified matrix is ​​determined based on the UAV cruise route data to be analyzed after benchmarking and the principal component description knowledge matrix, the descriptive knowledge in the first simplified matrix is ​​selected to obtain a second simplified matrix, and the data in the second simplified matrix is ​​determined as the target UAV cruise route data.

6. The method according to claim 1, characterized in that Before loading the target UAV cruise route data into the flight trajectory intelligent planning thread, the method further includes: A cruise trajectory example data set is obtained, and a flight trajectory intelligent planning thread to be trained is trained according to each group of cruise trajectory example data in the cruise trajectory example data set to obtain the flight trajectory intelligent planning thread.

7. The method according to claim 6, characterized in that The training of the to-be-trained flight trajectory intelligent planning thread according to each group of cruise trajectory example data in the cruise trajectory example data set to obtain the flight trajectory intelligent planning thread includes: Preprocessing the cruise trajectory example data, and simplifying the preprocessed cruise trajectory example data to obtain target cruise trajectory example data covering main description knowledge; Dividing the cruise trajectory example data in the cruise trajectory example data set into a first cruise trajectory example data subset and a second cruise trajectory example data subset according to flight tags, and building an unannotated data set according to the target cruise trajectory example data in the first cruise trajectory example data subset; Perform abnormal flight trajectory description coefficient detection on the target cruise trajectory example data in the second cruise trajectory example data subset, and build an annotation data set according to the target cruise trajectory example data in the second cruise trajectory example data subset and the corresponding directory; The flight trajectory intelligent planning thread to be trained is trained according to the unannotated data set and the annotated data set to obtain the flight trajectory intelligent planning thread.

8. The method according to claim 7, characterized in that The performing abnormal flight trajectory description coefficient detection on the target cruise trajectory example data in the second cruise trajectory example data subset includes: Identifying target cruise trajectory example data of the second cruise trajectory example data subset, and determining a random one of the target cruise trajectory example data as target flight trajectory data; Determine an a-difference value according to the target flight trajectory data and the ath nearest neighbor data corresponding to the target flight trajectory data, where a is a positive integer; Obtaining a real-time difference value between the target flight trajectory data and the a-th nearest neighbor data, and determining a target difference value corresponding to the target flight trajectory data according to the a-difference value and the real-time difference value; Determining a local confidence corresponding to the target flight trajectory data according to a target difference value corresponding to a nearest neighbor data of the target flight trajectory data, and determining a local influencing factor corresponding to the target flight trajectory data based on the local confidence; It is determined whether the target flight trajectory data is abnormal farmland data according to the local influencing factors and the pre-set specified values, and a directory corresponding to the flight tags covering the target flight trajectory data is determined according to the determination result.

9. The method according to claim 7, characterized in that: The training of the flight trajectory intelligent planning thread to be trained according to the unannotated data set and the annotated data set to obtain the flight trajectory intelligent planning thread includes: Actively selecting data with the largest amount of information from the unannotated data set, determining a directory corresponding to the data with the largest amount of information by manual annotation, and updating the annotated data set according to the data with the largest amount of information and the directory; Training the flight trajectory intelligent planning thread to be trained according to the updated annotation data set, and optimizing the parameters of the flight trajectory intelligent planning thread to be trained by an autonomous training unit during the training process; The above process is repeated until the pre-set training requirements are met to obtain the flight trajectory intelligent planning thread.

10. A system for intelligent planning of aerial photography routes for inland cultivated land survey, characterized in that: The invention comprises a processor and a memory communicating with each other, wherein the processor is used to read a computer program from the memory and execute the computer program to implement the method according to any one of claims 1 to 9.

Images