A classification method and classification system for laser radar point cloud data

Abstract

The present invention discloses a classification method and a classification system for laser radar point cloud data. The classification method comprises: step one, selecting a number of sample data from all sample data to form a training data set, and forming the remaining other sample data into a test data set; step two, building a classification model based on the training data set, comprising using a clustering algorithm to divide all training sample data in the training data set into a plurality of small data sets of different categories, for a number of small data sets of each category, respectively selecting a small data set as a starting small data set, and aggregating the other small data sets with the starting small data set to form a large data set; step three, using the test data set to test the classification accuracy of the constructed classification model. The present invention can generate a high-precision classification model when the number of sample data is not large enough.

Description

Technical Field

[0001] The present invention belongs to the technical field of point cloud data classification, and in particular relates to a classification method and a classification system for laser radar point cloud data. Background Art

[0002] Airborne laser radar transmits laser signals to the ground and then collects laser signals reflected by the ground. Since the speed of light is known, the propagation time of the laser can be converted into a distance measurement. Combined with the height and scanning angle of the airborne laser radar, the three-dimensional coordinates of each ground measurement point can be accurately calculated. A large number of three-dimensional coordinate points constitute the laser radar point cloud data. As an important basic data for three-dimensional modeling, the laser radar point cloud data usually needs to be preprocessed before modeling. The preprocessing specifically includes classifying the laser radar point cloud data.

[0003] There are many methods in the prior art that can realize the classification of lidar point cloud data. In particular, a classification model for classifying lidar point cloud data can be constructed in advance. In order to construct a classification model with high classification accuracy, it is generally necessary to divide the sample data into a training data set, a verification data set, and a test data set. Among them, the training data set is used to train the classification model so that it can learn accurate classification rules as much as possible. Therefore, the number of samples in the training data set should be larger than the number of samples in the other two data sets. The verification data set is used to calculate the classification error and adjust the parameters of the classification model. The test data set is used to evaluate the performance of the final classification model. However, when the total number of sample data is not large enough, it is often difficult to guarantee the number of samples in the training data set, which leads to the problem of poor classification effect of the classification model. Therefore, a classification method and classification system for lidar point cloud data are urgently needed to solve this technical problem. Summary of the invention

[0004] When the total amount of sample data is not large enough, the present invention generates a training data set and a test data set only from the sample data, thereby ensuring the amount of training sample data in the training data set, and constructs a classification model for classifying lidar point cloud data based on the training data set, and finally generates a classification model with a small classification error.

[0005] In order to achieve the above-mentioned purpose of the invention, a classification method of laser radar point cloud data is provided as follows, which mainly includes the following steps:

[0006] Step 1: According to the total number of sample data acquired in advance, select a number of sample data from all the sample data to form a training data set, and form the remaining sample data into a test data set, wherein the training data set is used to construct a classification model for classifying lidar point cloud data, and the test data set is used to verify the classification accuracy of the classification model that has been constructed;

[0007] Step 2: constructing the classification model based on the training data set. First, all the training sample data in the training data set are divided into a plurality of small data sets of different categories by using a clustering algorithm. Secondly, for several small data sets of each category, one small data set is selected as a starting small data set, and other small data sets are aggregated with the starting small data set to form a large data set. Finally, when all the small data sets of all categories have been aggregated, the classification model is constructed once.

[0008] Step three: Use the test sample data in the test data set to test the classification accuracy of the classification model that has been constructed. Only when the classification accuracy of the classification model can meet the requirements, there is no need to repeatedly construct the classification model, and the classification model outputs the classification rules for the lidar point cloud data. Otherwise, continue to repeat the construction steps based on the last construction of the classification model.

[0009] As a preferred technical solution of the present invention, the process of constructing the classification model based on the training data set specifically includes the following steps:

[0010] The first step is to use a clustering algorithm for the training data set to divide all the training sample data therein into a plurality of small data sets of different categories, each small data set containing a plurality of training sample data;

[0011] The second step is to select a category from multiple small data sets of different categories;

[0012] Step 3: Determine all small data sets belonging to the selected category, and determine whether the number of all small data sets belonging to the selected category is greater than 1. If the number is greater than 1, select the small data set with the most training sample data from all small data sets belonging to the selected category as the starting small data set, and continue to the next step. Otherwise, jump to step 9.

[0013] Step 4: Select a small data set from all the small data sets belonging to the selected category as a candidate small data set. The candidate small data set refers to other small data sets except the starting small data set and has not been used as a candidate small data set.

[0014] Step 5: If the candidate small data set and the starting small data set are aggregated to form a large data set, the average classification error of the candidate small data set and the starting small data set before aggregation is calculated, and the classification error of the large data set after aggregation is calculated, and the average classification error and the classification error are compared. When the average classification error is greater than the classification error, continue to the next step, otherwise, jump to step 4;

[0015] Step 6: Continue to determine whether the classification error of the aggregated large data set is less than the minimum classification error so far. If so, proceed to the next step. Otherwise, jump to step 4.

[0016] Step 7: Use the classification error of the large data set to update the minimum classification error so far and proceed to the next step;

[0017] Step 8: Determine whether all small data sets except the starting small data set in all small data sets belonging to the selected category have been processed. If so, aggregate the small data set corresponding to the smallest classification error so far with the starting small data set, and use the formed large data set as the new starting small data set, and continue to the next step. Otherwise, jump to step 4.

[0018] Step 9: Continue to determine whether all categories of small data sets have been processed. If so, complete the construction of the classification model. Otherwise, jump to step 2.

[0019] As a preferred technical solution of the present invention, for each selected category, a minimum classification error so far is set, and the initial value of the minimum classification error so far is the classification error corresponding to the first candidate small data set that meets the condition in the fifth step that the average classification error is greater than the classification error.

[0020] As a preferred technical solution of the present invention, the average classification error and the classification error are compared, which is specifically achieved by the following formula:

[0021]

[0022] Among them, x1 is the number of training sample data in the starting small data set before aggregation, x2 is the number of training sample data in the candidate small data set before aggregation, λ1 is the classification error of the starting small data set before aggregation, λ2 is the classification error of the candidate small data set before aggregation, and x is the classification error of the large data set after aggregation.

[0023] As a preferred technical solution of the present invention, the classification model outputs classification rules for lidar point cloud data. The classification rules specifically refer to the ranges of each attribute value of different categories of training sample data determined based on the ranges occupied by different categories of training sample data in the attribute space when stopping to build the classification model, as well as the correspondence between the categories.

[0024] The present invention also provides a classification system for laser radar point cloud data, comprising the following modules:

[0025] A data set generation module is used to select a number of sample data from all the sample data to form a training data set according to the total number of sample data obtained in advance, and to form the remaining sample data into a test data set;

[0026] A classification model building module is used to build a classification model based on a training data set. All training sample data in the training data set is divided into a plurality of small data sets of different categories by using a clustering algorithm. For a number of small data sets in each category, one small data set is selected as a starting small data set, and other small data sets are aggregated with the starting small data set to form a large data set. When all small data sets of all categories have been aggregated, the classification model is built once. It is also used to repeatedly build the classification model when the classification accuracy of the built classification model cannot meet the requirements.

[0027] The classification model verification module is used to use the test sample data in the test data set to test the classification accuracy of the constructed classification model and determine whether the classification accuracy meets the requirements.

[0028] Compared with the prior art, the beneficial effects of the present invention are at least as follows:

[0029] 1. The present invention first selects several sample data from all sample data to form a training data set, and forms the remaining other sample data into a test data set; secondly, a classification model is constructed based on the training data set, including using a clustering algorithm to divide all training sample data in the training data set into a plurality of small data sets of different categories, and for a plurality of small data sets of each category, one small data set is selected from each of them as a starting small data set, and the other small data sets are aggregated with the starting small data set to form a large data set; finally, the classification accuracy of the constructed classification model is tested using the test data set, and the construction step of the classification model is stopped only when the classification accuracy meets the requirements;

[0030] 2. The present invention solves the problem of difficulty in ensuring the number of samples in the training data set due to the insufficient total number of sample data, and the resulting problem of poor classification accuracy of the classification model. The present invention can avoid the formation of a verification data set, thereby ensuring the number of samples in the training data set. At the same time, the present invention can also maximize the classification accuracy of the classification model in the process of constructing the classification model. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 A flowchart of the steps of a method for classifying laser radar point cloud data according to the present invention;

[0032] Figure 2 A flow chart of the steps of constructing a classification model of the present invention;

[0033] Figure 3 This is a structural diagram of the composition of a classification system for laser radar point cloud data of the present invention. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0035] It is understood that the terms "first", "second", etc. used in this application may be used herein to describe various elements, but unless otherwise specified, these elements are not limited by these terms. These terms are only used to distinguish a first element from another element. For example, without departing from the scope of this application, a first xx script may be referred to as a second xx script, and similarly, a second xx script may be referred to as a first xx script.

[0036] The inventors have discovered that when a classification model is constructed in advance and used to classify lidar point cloud data, there may be a problem that the total number of sample data is not large enough. In this case, the number of training sample data in the training data set used to construct the classification model cannot be guaranteed. In other words, the insufficient total number of sample data may cause the classification model to be unable to learn accurate classification rules, thereby resulting in poor classification effect of the classification model for lidar point cloud data. In order to solve this technical problem, the present invention proposes a classification method for lidar point cloud data.

[0037] References Figure 1 As shown, the present invention provides a classification method for laser radar point cloud data, which is mainly achieved by executing the following steps:

[0038] Step 1: According to the total number of sample data acquired in advance, select a number of sample data from all the sample data to form a training data set, and form the remaining sample data into a test data set, wherein the training data set is used to construct a classification model for classifying lidar point cloud data, and the test data set is used to verify the classification accuracy of the classification model that has been constructed;

[0039] Specifically, for example, when the total number of sample data is 1,000, since the total number of sample data is relatively sufficient, 70% of the sample data will form a training data set, and 30% of the sample data will form a test data set. When the total number of sample data is 500, in order to ensure the number of training sample data in the training data set as much as possible, 80% of the sample data will form a training data set, and 20% of the sample data will form a test data set. In addition, since a verification data set is not formed from the sample data, the number of training sample data in the training data set is further guaranteed. This embodiment does not limit the specific method of generating a training data set and a test data set from sample data.

[0040] Step 2: construct the classification model based on the training data set. First, all the training sample data in the training data set is divided into a plurality of small data sets of different categories by using a clustering algorithm. Then, for several small data sets of each category, one small data set is selected as a starting small data set, and other small data sets are aggregated with the starting small data set to form a large data set. Finally, when all the small data sets of all categories have been aggregated, the classification model is constructed once.

[0041] Specifically, the above clustering algorithm implements clustering processing on the training sample data based on the distance between adjacent training sample data calculated by each attribute value of the training sample data. In the prior art, clustering processing of lidar point cloud data often results in multiple small data sets, some of which belong to the same category, while some belong to different categories. Therefore, small data sets of the same category should be merged as much as possible to obtain classification rules with more universal and objective significance. The process of building a classification model using the training data set will be described in detail through the first to ninth steps below.

[0042] Step 3: Use the test sample data in the test data set to test the classification accuracy of the classification model that has been built. Only when the classification accuracy of the classification model can meet the requirements, there is no need to repeatedly build the classification model, and the classification model outputs the classification rules for the lidar point cloud data. Otherwise, continue to repeat the construction steps based on the last construction of the classification model.

[0043] Specifically, when the classification model is constructed for the first time, the first to ninth steps described below are executed. When the inspection results of the test data set for the constructed classification model show that the classification accuracy of the classification model cannot meet the requirements, the construction steps that are repeatedly executed refer to the second to ninth steps described below, that is, there is no need to repeat the clustering processing of the training sample data in the training data set.

[0044] For further reference, see Figure 2 As shown, the process of building the above classification model based on the above training data set specifically includes the following steps:

[0045] The first step is to use a clustering algorithm for the above training data set to divide all the training sample data into multiple small data sets of different categories, each of which contains several training sample data;

[0046] The second step is to select a category from multiple small data sets of different categories;

[0047] Step 3: Determine all small data sets belonging to the selected category, and determine whether the number of all small data sets belonging to the selected category is greater than 1. If the number is greater than 1, select the small data set with the most training sample data from all small data sets belonging to the selected category as the starting small data set, and continue to the next step. Otherwise, jump to step 9.

[0048] Step 4: Select a small data set from all the small data sets belonging to the selected category as a candidate small data set. The candidate small data set refers to other small data sets except the starting small data set and has not been used as a candidate small data set.

[0049] Step 5: If the candidate small data set and the starting small data set are aggregated to form a large data set, the average classification error of the candidate small data set and the starting small data set before aggregation is calculated, and the classification error of the large data set after aggregation is calculated, and the average classification error and the classification error are compared. When the average classification error is greater than the classification error, continue to the next step, otherwise, jump to step 4;

[0050] Step 6: Continue to determine whether the classification error of the aggregated large data set is less than the minimum classification error so far. If so, proceed to the next step. Otherwise, jump to step 4.

[0051] Step 7: Use the classification error of the large data set to update the minimum classification error so far and proceed to the next step;

[0052] Step 8: Determine whether all small data sets except the starting small data set in all small data sets belonging to the selected category have been processed. If so, aggregate the small data set corresponding to the smallest classification error so far with the starting small data set, and use the formed large data set as the new starting small data set, and continue to the next step. Otherwise, jump to step 4.

[0053] Step 9: Continue to determine whether all small data sets of all categories have been processed. If so, complete the construction of the above classification model. Otherwise, jump to step 2.

[0054] Further, for each selected category, a minimum classification error is set so far, and the initial value of the minimum classification error so far is the classification error corresponding to the first candidate small data set that satisfies the condition that the average classification error is greater than the classification error in the fifth step;

[0055] Specifically, in the first step to the ninth step, the process of aggregating small data sets belonging to different categories is the same. Therefore, only the aggregation process of small data sets belonging to one category is described in detail here. First, only when the number of small data sets is greater than one, it is possible to aggregate small data sets and determine the starting small data set in all small data sets. Then, it is necessary to select a small data set from other small data sets except the starting small data set and use it as a candidate small data set. Only one candidate small data set can be aggregated with the starting small data set at a time. Secondly, considering that the classification error after the candidate small data set and the starting small data set are aggregated should be less than the average classification error before the two are aggregated, only when this condition is met can the classification accuracy of the classification rules learned by the classification model be improved while aggregating the candidate small data set and the starting small data set. Thirdly, it can be aggregated with the starting small data set. The classification error of the aggregated candidate small data set after aggregation should be smaller than the classification errors of other small data sets after aggregation, so as to ensure that the classification accuracy of the classification rules learned by the classification model is maximized in the process of building the classification model each time. Finally, when all small data sets except the starting small data set have undergone the above processing, the small data set corresponding to the smallest classification error can be aggregated with the starting small data set. It should be noted that the large data set formed by the aggregation at this time becomes a new starting small data set, so as to facilitate repeated construction steps for the classification model. Through the method described above, the present invention realizes that in the process of aggregating small data sets and starting small data sets, the classification error changes before and after aggregation are considered at the same time, and by selecting different small data sets to aggregate with the starting small data set, the classification accuracy of the classification model can be maximized, so that there is no need to adjust the classification model through the verification data set.

[0056] Furthermore, the average classification error and the classification error are compared, which is implemented by the following formula:

[0057]

[0058] Among them, x1 is the number of training sample data in the starting small data set before aggregation, x2 is the number of training sample data in the candidate small data set before aggregation, λ1 is the classification error of the starting small data set before aggregation, λ2 is the classification error of the candidate small data set before aggregation, and x is the classification error of the large data set after aggregation; specifically, the above formula can be used to determine whether the candidate small data set and the starting small data set can be aggregated to form a large data set. It should also be noted that the classification error in this embodiment refers to the ratio of the number of incorrectly classified training sample data in the data set to the total number of training sample data in the data set, and the data set can be a small data set and a large data set.

[0059] Furthermore, the above-mentioned classification model outputs classification rules for the lidar point cloud data. The above-mentioned classification rules specifically refer to the value ranges of each attribute value of the training sample data of different categories, which are determined based on the range occupied by the training sample data of different categories in the attribute space when the construction of the above-mentioned classification model is stopped, and the correspondence between the categories; specifically, when the construction of the above-mentioned classification model is stopped, the classification accuracy of the classification model after the construction step can meet the predetermined requirements. Among them, this embodiment does not limit the specific method for evaluating the classification accuracy. For example, the average accuracy index or the log-loss index can be obtained by using the evaluation criteria of the multi-classification model to evaluate the classification accuracy of the classification model.

[0060] References Figure 3 As shown, the present invention also provides a classification system for laser radar point cloud data, which is used to implement a classification method for laser radar point cloud data as described above. Specifically, the functions of each module are described as follows:

[0061] A data set generation module is used to select a number of sample data from all the sample data to form a training data set according to the total number of sample data obtained in advance, and to form the remaining sample data into a test data set;

[0062] A classification model building module is used to build a classification model based on a training data set. All training sample data in the training data set is divided into a plurality of small data sets of different categories by using a clustering algorithm. For a number of small data sets in each category, one small data set is selected as a starting small data set, and other small data sets are aggregated with the starting small data set to form a large data set. When all small data sets of all categories have been aggregated, the classification model is built once. It is also used to repeatedly build the classification model when the classification accuracy of the built classification model cannot meet the requirements.

[0063] The classification model verification module is used to use the test sample data in the test data set to test the classification accuracy of the constructed classification model and determine whether the classification accuracy meets the requirements.

[0064] In summary, the present invention first selects several sample data from all the sample data to form a training data set, and forms the remaining other sample data into a test data set; secondly, a classification model is constructed based on the training data set, including using a clustering algorithm to divide all the training sample data in the training data set into a plurality of small data sets of different categories, and for the several small data sets of each category, one small data set is selected as a starting small data set, and the other small data sets are aggregated with the starting small data set to form a large data set; finally, the classification accuracy of the constructed classification model is tested using the test data set, and only when the classification accuracy meets the requirements, the construction step of the classification model is stopped; the present invention solves the problem of difficulty in ensuring the number of samples in the training data set due to the insufficient total number of sample data, and the problem of poor classification accuracy of the classification model caused by the insufficient total number of sample data, and the present invention can avoid the formation of a verification data set, thereby ensuring the number of samples in the training data set, and at the same time, the present invention can also maximize the classification accuracy of the classification model in the process of constructing the classification model.

[0065] It should be understood that, although each step in the flow chart of each embodiment of the present invention is shown in sequence according to the indication of the arrow, these steps are not necessarily performed in sequence according to the order indicated by the arrow. Unless there is a clear explanation in this article, the execution of these steps does not have strict order restrictions, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0066] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The above-mentioned program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0067] The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0068] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

[0069] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A classification method for laser radar point cloud data, characterized in that: The steps include: Step 1: According to the total number of sample data acquired in advance, select a number of sample data from all the sample data to form a training data set, and form the remaining sample data into a test data set, wherein the training data set is used to construct a classification model for classifying lidar point cloud data, and the test data set is used to verify the classification accuracy of the classification model that has been constructed; Step 2: constructing the classification model based on the training data set. First, all the training sample data in the training data set are divided into a plurality of small data sets of different categories by using a clustering algorithm. Secondly, for several small data sets of each category, one small data set is selected as a starting small data set, and the other small data sets are aggregated with the starting small data set to form a large data set. Finally, when all the small data sets of all categories have been aggregated, the classification model is constructed once. Step 3: Use the test sample data in the test data set to test the classification accuracy of the classification model that has been built. Only when the classification accuracy of the classification model can meet the requirements, there is no need to repeatedly build the classification model, and the classification model outputs the classification rules for the lidar point cloud data. Otherwise, continue to repeat the construction steps based on the last construction of the classification model; The process of aggregating other small data sets with the initial small data set to form a large data set includes the following steps: The first step is to select a small data set from all the small data sets belonging to the selected category as a candidate small data set. The candidate small data set refers to other small data sets except the starting small data set and has not been used as a candidate small data set. Step 2: If the candidate small data set and the starting small data set are aggregated to form a large data set, the average classification error of the candidate small data set and the starting small data set before aggregation is calculated, and the classification error of the large data set after aggregation is calculated, and the average classification error and the classification error are compared. When the average classification error is greater than the classification error, continue to the next step, otherwise, jump to the first step; Step 3: Continue to determine whether the classification error of the aggregated large data set is less than the minimum classification error so far. If so, proceed to the next step; otherwise, jump to the first step. Step 4: Use the classification error of the large data set to update the minimum classification error so far and proceed to the next step; Step 5: Determine whether all small data sets except the starting small data set in all small data sets belonging to the selected category have been processed. If so, aggregate the small data set corresponding to the smallest classification error so far with the starting small data set, and use the formed large data set as the new starting small data set. Otherwise, jump to the first step.

2. The method for classifying laser radar point cloud data according to claim 1, characterized in that: The process of constructing the classification model based on the training data set specifically includes the following steps: The first step is to use a clustering algorithm for the training data set to divide all the training sample data therein into a plurality of small data sets of different categories, each small data set containing a plurality of training sample data; The second step is to select a category from multiple small data sets of different categories; Step 3: determine all small data sets belonging to the selected category, and judge whether the number of all small data sets belonging to the selected category is greater than 1. If the number is greater than 1, select the small data set with the most training sample data from all small data sets belonging to the selected category as the starting small data set, and aggregate other small data sets with the starting small data set to form a large data set. Otherwise, jump to step 4. Step 4: Continue to determine whether all categories of small data sets have been processed. If so, complete the construction of the classification model. Otherwise, jump to step 2.

3. The method for classifying laser radar point cloud data according to claim 2, characterized in that: For each selected category, a minimum classification error is set so far. The initial value of the minimum classification error is the classification error corresponding to the first candidate small data set in the fifth step that meets the condition that the average classification error is greater than the classification error.

4. The method for classifying laser radar point cloud data according to claim 2, characterized in that: Compare the average classification error and the classification error, which is achieved through the following formula: in, is the number of training sample data in the initial small data set before aggregation, is the number of training sample data in the candidate small data set before aggregation, is the classification error of the initial small data set before aggregation, is the classification error of the candidate small data set before aggregation, is the classification error of the large data set after aggregation.

5. The method for classifying laser radar point cloud data according to claim 1, characterized in that: The classification model outputs classification rules for lidar point cloud data. The classification rules specifically refer to the ranges of attribute values ​​of different categories of training sample data determined based on the ranges occupied by different categories of training sample data in the attribute space when the construction of the classification model is stopped, as well as the correspondence between the categories.

6. A classification system for laser radar point cloud data, used to implement the method according to any one of claims 1 to 5, characterized in that: Includes the following modules: A data set generation module is used to select a number of sample data from all the sample data to form a training data set according to the total number of sample data obtained in advance, and to form the remaining sample data into a test data set; A classification model building module is used to build a classification model based on a training data set. All training sample data in the training data set is divided into a plurality of small data sets of different categories by using a clustering algorithm. For a number of small data sets in each category, one small data set is selected as a starting small data set, and other small data sets are aggregated with the starting small data set to form a large data set. When all small data sets of all categories have been aggregated, the classification model is built once. It is also used to repeatedly build the classification model when the classification accuracy of the built classification model cannot meet the requirements. The classification model verification module is used to use the test sample data in the test data set to test the classification accuracy of the constructed classification model and determine whether the classification accuracy meets the requirements.

Images