KR102405818B1 - Method of removing noise, apparatus for removing noise, and computer program for the method - Google Patents

Abstract

The present invention provides a noise removal method capable of effectively removing noise in point cloud information of a three-dimensional lidar sensor, an apparatus for removing noise, and a computer program stored in a recording medium for executing the method, noise removal performed by a computing device In the method, obtaining point cloud data including a plurality of points from a lidar sensor that obtains three-dimensional data, using a first artificial neural network learned in advance, a restoration error value for each point of the plurality of points Estimating, determining a noise of dividing a noise point and a valid point with respect to the plurality of points based on the restoration error value and a predetermined threshold value, and removing the noise point from the point cloud data to form the valid point It provides a method for removing noise, an apparatus for removing noise, and a computer program stored in a recording medium for executing the method, including outputting three-dimensional data.

Description

A method of removing noise, an apparatus for removing noise, and a computer program stored in a recording medium for executing the method {Method of removing noise, apparatus for removing noise, and computer program for the method}

Embodiments of the present invention relate to a method for removing noise, an apparatus for removing noise, and a computer program stored in a recording medium for executing the method, and more particularly, to effectively remove noise in point cloud information of a three-dimensional lidar sensor. It relates to a method for removing noise, an apparatus for removing noise, and a computer program stored in a recording medium for executing the method.

Autonomous moving objects and robot systems, including autonomous vehicles, widely use 3D LiDAR (Light Detection And Ranging) sensors to recognize accurate 3D information about the surrounding environment. However, while most systems are operated and utilized in an outdoor environment, sensors in the corresponding environment have a disadvantage in generating noise due to weather conditions such as snowfall and rainfall. As a result, such noise may cause unawareness or misrecognition of the surrounding environment, which may cause serious errors in the system, as well as damage to surrounding people including the operator.

To solve this problem, research has been conducted to remove noise in three-dimensional point cloud data, and representatively, there are a density-based method and a supervised learning-based method. The density-based noise removal method discriminates and removes noise by measuring the density of point cloud information in a space of a preset unit. However, the accuracy is very low because there are cases where a large number of point cloud information other than noise also has a low density. The supervised learning-based noise removal method is a method of manually classifying noise information and non-noise information in point cloud information, and then performing machine learning to distinguish the information. This technique has a disadvantage in that it requires a large amount of manually classified information. In particular, unlike image information, point cloud information is a difficult technique to use in that it is not only very difficult to manually classify information, but also requires obtaining noise information under countless conditions.

The present invention is to solve various problems including the above problems, a noise removal method capable of effectively removing noise in point cloud information of a three-dimensional lidar sensor, an apparatus for removing noise, and a recording medium for implementing the method The purpose of providing a computer program stored in However, these problems are exemplary, and the scope of the present invention is not limited thereto.

According to one aspect of the present invention, in a noise removal method performed by a computing device, the step of acquiring point cloud data including a plurality of points from a lidar sensor that acquires three-dimensional data, the first artificial neural network learned in advance estimating a restoration error value for each point of the plurality of points using and outputting 3D data including the effective points by removing the noise points from the point cloud data.

The method for removing noise according to an embodiment of the present invention may further include calculating the restoration error value for each point of the plurality of points using a second artificial neural network learned in advance.

The first artificial neural network may be trained using the restoration error value of the second artificial neural network.

Calculating the restoration error value may include distinguishing a first point to be restored from among the plurality of points and a second point used for restoration, and determining the location of the first point by using the second point. and calculating a restoration error value of the first point, wherein the second point may be a set of a predetermined number of points adjacent to the first point in a three-dimensional space.

Calculating the restoration error value may include: acquiring real coordinate information of the first point; acquiring restored coordinate information of the first point; and three-dimensionality between the real coordinate information and the restored coordinate information calculating a distance value.

The estimating of the restoration error value may include: determining a third point to be estimated; inputting three-dimensional coordinate information of the third point into the first artificial neural network; It may include obtaining the restoration error value of the third point.

The determining of the noise may include determining the third point as the noise point when the restoration error value of the third point exceeds the predetermined threshold value.

According to one aspect of the present invention, there is provided a computer program stored in a recording medium for executing the above-described method using a computer.

According to one aspect of the present invention, it includes a processor, wherein the processor acquires point cloud data including a plurality of points from a lidar sensor that acquires three-dimensional data, and uses the first artificial neural network learned in advance. Estimate a restoration error value for each point of a plurality of points, classify a noise point and an effective point with respect to the plurality of points based on the restoration error value and a predetermined threshold value, and select the noise point from the point cloud data There is provided a noise removal device for outputting three-dimensional data including the valid points by removing the data.

The processor may calculate the restoration error value for each point of the plurality of points by using the second artificial neural network learned in advance.

The first artificial neural network may be trained using the restoration error value of the second artificial neural network.

The processor distinguishes a first point to be restored from among the plurality of points and a second point used for restoration, and restores the position of the first point by using the second point, and the first point calculates a restoration error value of , and the second point may be a set of a predetermined number of points adjacent to the first point in a three-dimensional space.

The processor may obtain real coordinate information of the first point, obtain restored coordinate information of the first point, and calculate a three-dimensional distance value between the real coordinate information and the restored coordinate information.

The processor determines a third point to be estimated, inputs three-dimensional coordinate information of the third point to the first artificial neural network, and obtains the restoration error value of the third point from the first artificial neural network can be obtained

The processor may determine the third point as the noise point when the restoration error value of the third point exceeds the predetermined threshold value.

Other aspects, features and advantages other than those described above will become apparent from the following detailed description, claims and drawings for carrying out the invention.

According to an embodiment of the present invention made as described above, a noise removal method capable of effectively removing noise in point cloud information of a three-dimensional lidar sensor, an apparatus for removing noise, and a computer program stored in a recording medium for executing the method can be implemented. Of course, the scope of the present invention is not limited by these effects.

1 is a view for explaining the configuration and operation of a noise removing apparatus according to an embodiment of the present invention.
2 is a diagram for explaining the configuration of a processor of a noise removing apparatus according to an embodiment of the present invention.
3 is a flowchart illustrating a noise removal method according to an embodiment of the present invention.
4 is a flowchart illustrating a noise removal method according to another embodiment of the present invention.
5 is a flowchart illustrating a noise removal method according to another embodiment of the present invention.

Since the present invention can apply various transformations and can have various embodiments, specific embodiments are illustrated in the drawings and described in detail in the detailed description. Effects and features of the present invention, and a method for achieving them will become apparent with reference to the embodiments described below in detail in conjunction with the drawings. However, the present invention is not limited to the embodiments disclosed below and may be implemented in various forms.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings, and when described with reference to the drawings, the same or corresponding components are given the same reference numerals, and the overlapping description thereof will be omitted. .

In the following embodiments, terms such as first, second, etc. are used for the purpose of distinguishing one component from another without limiting the meaning. And singular expressions include plural expressions unless the context clearly dictates otherwise. In addition, the terms include or have means that a feature or element described in the specification is present, and does not exclude the possibility that one or more other features or elements may be added.

In the drawings, the size of the components may be exaggerated or reduced for convenience of description. For example, since the size and thickness of each component shown in the drawings are arbitrarily indicated for convenience of description, the present invention is not necessarily limited to the illustrated bar.

In the following embodiments, when it is said that a part such as a region, component, part, block or module is on or on another part, not only when it is directly on the other part, but also another region, component, part in the middle , blocks or modules are included. And when a region, component, part, block or module is connected, it is not only when the region, component, part, block or module is directly connected, but also another region in the middle of the region, component, part, block or module. , includes cases in which components, units, blocks or modules are interposed and indirectly connected.

Hereinafter, various embodiments of the present invention will be described in detail with reference to the accompanying drawings in order to enable those of ordinary skill in the art to easily practice the present invention.

FIG. 1 is a diagram for explaining the configuration and operation of a noise removing apparatus according to an embodiment of the present invention, and FIG. 2 is a diagram for explaining a processor configuration of the noise removing apparatus according to an embodiment of the present invention.

First, referring to FIG. 1 , an apparatus 100 for removing noise according to an embodiment of the present invention may include a memory 110 , a processor 120 , a communication module 130 , and a lidar sensor 140 . . However, the present invention is not limited thereto, and the noise removing apparatus 100 may further include other components or some components may be omitted. Some components of the noise removing apparatus 100 may be separated into a plurality of devices, or a plurality of components may be merged into one device.

The memory 110 is a computer-readable recording medium and may include a random access memory (RAM), a read only memory (ROM), and a permanent mass storage device such as a disk drive. In addition, a program code for controlling the noise removing apparatus 100 and a pre-learned artificial neural network may be temporarily or permanently stored in the memory 110 .

The processor 120 may acquire point cloud data including a plurality of points from a lidar sensor that acquires 3D data. Also, the processor 120 may estimate a restoration error value for each point of the plurality of points by using the first artificial neural network learned in advance. Also, the processor 120 may classify a noise point and an effective point with respect to a plurality of points based on the restoration error value and a predetermined threshold value. Also, the processor 120 may output 3D data including valid points by removing noise points from the point cloud data.

The communication module 130 may provide a function for communicating with an external server through a network. For example, a request generated by the processor 120 of the noise removal apparatus 100 according to a program code stored in a recording device such as the memory 110 is transmitted to an external server through a network under the control of the communication module 130 . can Conversely, a control signal, command, content, file, etc. provided under the control of the processor of the external server may be received by the noise removing apparatus 100 through the communication module 130 through the network. For example, a control signal or command of an external server received through the communication module 130 may be transmitted to the processor 120 or the memory 110 .

The communication method is not limited, and not only a communication method using a communication network (eg, a mobile communication network, a wired Internet, a wireless Internet, a broadcasting network) that the network may include, but also short-range wireless communication between devices may be included. For example, the network includes a personal area network (PAN), a local area network (LAN), a campus area network (CAN), a metropolitan area network (MAN), a wide area network (WAN), a broadband network (BBN), the Internet, and the like. may include any one or more of the networks of Further, the network may include, but is not limited to, any one or more of a network topology including, but not limited to, a bus network, a star network, a ring network, a mesh network, a star-bus network, a tree or a hierarchical network, and the like. .

Also, the communication module 130 may communicate with an external server through a network. Although the communication method is not limited, the network may be a local area wireless network. For example, the network may be a Bluetooth (Bluetooth), BLE (Bluetooth Low Energy), or Wifi communication network.

The lidar sensor 140 may be a LiDAR (Light Detection And Ranging) sensor that acquires 3D data using a laser. In addition, the lidar sensor 140 may be provided in a vehicle or an aircraft. In this case, the lidar sensor 140 may be mounted on a vehicle or a flying vehicle to acquire 3D data, and model a 3D image based on the acquired 3D data to detect surrounding objects, topographical features, and the like.

Also, the noise removing apparatus 100 according to the present invention may include an input/output interface. The input/output interface may be a means for interfacing with an input/output device. For example, the input device may include a device such as a keyboard or mouse, and the output device may include a device such as a display for displaying a communication session of an application. As another example, the input/output interface may be a means for an interface with a device in which functions for input and output are integrated into one, such as a touch screen. As a more specific example, the processor 120 of the noise removal device 100 processes the command of the computer program loaded in the memory 110, and a service screen or content configured using data provided by an external server uses the input/output interface. can be displayed on the display.

Also, in other embodiments, the noise removing apparatus 100 may include more components than those of FIG. 1 . For example, it may be implemented to include at least a portion of the above-described input/output device, or may further include other components such as a battery and charging device for supplying power to internal components, various sensors, and a database.

Hereinafter, the internal configuration of the processor 120 of the noise removing apparatus 100 according to an embodiment of the present invention will be described in detail with reference to FIG. 2 . For ease of understanding, it is assumed that the processor 120 to be described later is the processor 120 of the noise removing apparatus 100 shown in FIG. 1 .

The processor 120 of the noise removal apparatus 100 according to an embodiment of the present invention includes a point cloud data acquisition unit 121 , a restoration error value estimation unit 122 , a noise determination unit 123 , and a data output unit 124 . ) is included. According to some embodiments, components of the processor 120 may be selectively included in or excluded from the processor 120 . In addition, according to some embodiments, the components of the processor 120 may be separated or combined to express the functions of the processor 120 .

The processor 120 and components of the processor 120 may control the noise removing apparatus 100 to perform steps S110 to S140 included in the noise removing method of FIG. 3 . For example, the processor 120 and components of the processor 120 may be implemented to execute instructions according to the code of the operating system included in the memory 110 and the code of at least one program. Here, the components of the processor 120 are expressions of different functions of the processor 120 that are performed by the processor 120 according to an instruction provided by the program code stored in the noise removal device 100 . can The internal configuration and specific operation of the processor 120 will be described with reference to the flowchart of the noise removal method of FIG. 3 .

3 is a flowchart illustrating a noise removal method according to an embodiment of the present invention.

Referring to FIG. 3 , in operation S110 , the noise removing apparatus 100 may acquire point cloud data including a plurality of points from a lidar sensor that acquires 3D data. For example, the point cloud data may be 3D data including N points. For example, the 3D data may include 3D coordinate values.

In operation S120 , the noise removal apparatus 100 may estimate a restoration error value for each point of the plurality of points using the first artificial neural network learned in advance. For example, the noise removing apparatus may estimate a restoration error value for each of the N points.

In operation S130 , the noise removing apparatus 100 may determine noise by classifying a noise point and an effective point with respect to a plurality of points based on the restoration error value and a predetermined threshold value. For example, when the restoration error value for each of the N points exceeds a predetermined threshold, the noise removing apparatus may determine the corresponding point as a noise point. Also, when the restoration error value for each of the N points is equal to or less than a predetermined threshold, the noise removing apparatus may determine the corresponding point as a valid point.

In operation S140 , the noise removal apparatus 100 may output 3D data including valid points by removing noise points from the point cloud data. For example, the noise removing apparatus may output 3D data including valid points by removing noise points among N points.

4 is a flowchart illustrating a noise removal method according to another embodiment of the present invention.

Referring to FIG. 4 , the apparatus for removing noise according to an embodiment of the present invention may reconstruct a position of a point in 3D point cloud information using the second artificial neural network F. Referring to FIG. For example, the second artificial neural network F may be an artificial neural network for estimating a location of a corresponding point by using point cloud information around one of a plurality of points. For example, the second artificial neural network F may restore the position of a corresponding point by using information about K points surrounding it based on an arbitrary point. Also, the noise removing apparatus may calculate a restoration error value of the restored point position.

The artificial neural network according to the present invention may be a Convolution Neural Network (CNN), a Recurrent Neural Network (RNN), or an Artificial Neural Network (ANN). However, the artificial neural network of the present invention is not limited to the above example, and may be implemented in various types of known artificial neural networks.

In operation S210 , the noise removing apparatus may acquire point cloud data from a lidar sensor that acquires 3D data. For example, the noise removing apparatus may acquire point cloud data including a plurality of points from the lidar sensor. For example, the plurality of points may be N points of p1 to pN. Also, the point may be a point having a three-dimensional coordinate value.

In step S220, the noise removing apparatus may select a point to be restored and a point to be used for restoration among a plurality of points. Also, the noise removing apparatus may separate a point to be restored from a point used for restoration. For example, the noise removing apparatus may select a pi point among a plurality of points as a point to be restored. Also, the noise removing apparatus may select four points around the pi point as points used for restoration. For example, the number of points used for restoration may be stored in advance.

In step S230, the noise removing apparatus may restore the position of the restoration target point. For example, the noise removal apparatus may restore a point to be restored by inputting a point used for restoration into the second artificial neural network F learned in advance. For example, the noise removing apparatus may receive four neighboring points of the pi point and restore the position of the pi point. Also, the noise removing apparatus may learn a point restoration result. More specifically, the noise removal apparatus may restore the position of the pi point by using a three-dimensional distance value between the pi point to be restored and a set of four surrounding points to be restored.

<Equation 1>

Here, the function neighbor in Equation 1 means a set of K adjacent points in a three-dimensional space. For example, neighbor(pi) may mean a set of K points adjacent to the pi point in a 3D space. For example, K may be predetermined.

<Equation 2>

Here, the function distance of Equation 2 means a three-dimensional distance value between a pi point and a set of K points adjacent to the pi point in a three-dimensional space. For example, a three-dimensional distance value may be expressed as a loss function.

In step S240 , the noise removing apparatus may calculate a restoration error value of a restoration target point. For example, the noise removal apparatus may calculate a restoration error value by comparing the actual position of the restoration target point with the restoration position.

The apparatus for removing noise according to an embodiment of the present invention may calculate a restoration error value for each point of a plurality of points by using the second artificial neural network F learned in advance.

The noise removal apparatus according to an embodiment of the present invention may distinguish a first point to be restored from among a plurality of points and a second point used for restoration. Also, the noise removing apparatus may restore the position of the first point by using the second point. Also, the noise removing apparatus may calculate a restoration error value of the first point. For example, the second point may be a set of a predetermined number of points adjacent to the first point in a three-dimensional space.

The apparatus for removing noise according to an embodiment of the present invention may acquire actual coordinate information of the first point. Also, the noise removing apparatus may obtain information on the restoration coordinates of the first point. Also, the noise removing apparatus may calculate a 3D distance value between the actual coordinate information and the restored coordinate information. Also, the noise removing apparatus may calculate a restoration error value using a 3D distance value between the actual coordinate information and the restored coordinate information.

5 is a flowchart illustrating a noise removal method according to another embodiment of the present invention.

Referring to FIG. 5 , an apparatus for removing noise according to an embodiment of the present invention may estimate a restoration error value of a point in 3D point cloud information using the first artificial neural network G. Referring to FIG. For example, the first artificial neural network G may be an artificial neural network capable of measuring the restoration accuracy of points. Also, the noise removing apparatus may determine whether the corresponding point is a noise point or a valid point based on the estimated restoration error value.

In operation S310, the noise removing apparatus may acquire point cloud data from a lidar sensor that acquires 3D data. For example, the noise removing apparatus may acquire point cloud data including a plurality of points from the lidar sensor. For example, the plurality of points may be N points of p1 to pN.

In steps S320 and S330, the noise removing apparatus may estimate a restoration error of each point among the plurality of points. For example, the noise removing apparatus may estimate the restoration error value of the pi point.

The first artificial neural network G according to an embodiment of the present invention may estimate the restoration error of each point. Also, the first artificial neural network G may be trained to estimate the restoration error of each point. For example, the noise removal apparatus may find a point that cannot be restored as a neighboring point by estimating a restoration error using the first artificial neural network (G), and classify the noise point and the valid point through this.

The apparatus for removing noise according to an embodiment of the present invention may estimate a restoration error value for each of a plurality of points by using the first artificial neural network (G). For example, the first artificial neural network G may receive N points of p1 to pN constituting the point cloud and estimate a restoration error value for each point.

In an embodiment of the present invention, the first artificial neural network (G) of the noise removal apparatus may be learned using the restoration error value of the second artificial neural network (F). For example, the estimation value of the restoration error of the pi point may estimate the restoration error value of the pi point calculated by the second artificial neural network (F). For example, the first artificial neural network (G) of the noise removal device learns the restoration error value of the second artificial neural network (F), and the first artificial neural network (G) learns the restoration error value of the second artificial neural network (F) It is possible to output a restoration error estimate for estimating .

<Equation 3>

Here, the loss function L in Equation 3 represents the difference between the restoration error estimation value Gi and the restoration error value. Also, the sum of the loss functions of N points in the point cloud may be expressed as a total loss function.

The noise removing apparatus according to an embodiment of the present invention may determine a third point to be estimated. Also, the noise removing apparatus may input 3D coordinate information of the third point to the first artificial neural network. Also, the noise removing apparatus may obtain a restoration error value of the third point from the first artificial neural network. For example, the restoration error value output by the first artificial neural network may be a restoration error estimation value. For example, the first artificial neural network and the second artificial neural network may be trained. In addition, the first artificial neural network may be learned by combining with the second artificial neural network.

The noise removing apparatus according to an embodiment of the present invention may determine the third point as the noise point when the restoration error value of the third point exceeds a predetermined threshold value. For example, when the output value of the first artificial neural network G with respect to the pi point exceeds a predetermined threshold, the noise removing apparatus may determine the pi point as a noise point. Also, the noise removing apparatus may remove the pi point determined as the noise point.

In the noise removal apparatus according to an embodiment of the present invention, based on the characteristic that the noise point has a distance value or location that is different from the surrounding point cloud information, when the information restored to the surrounding point cloud information has a large error from the actual point, it is determined as a noise point. can do.

The apparatus for removing noise according to the present invention may be composed of two artificial neural networks that respectively perform point position restoration and restoration error estimation using surrounding point cloud information. In addition, according to the present invention, it can be implemented as unsupervised learning that is reliable in various environments and does not require manual information.

According to the present invention, there is an advantage in that the correct answer information for distinguishing noise and non-noise is unnecessary as compared to the conventional noise removal method, and the noise detection performance is high. Therefore, when the present invention is applied to an autonomous mobile system such as an autonomous vehicle, an autonomous mobile robot, or a humanoid robot, reliable movement and operation can be ensured even in an environment where utilization is impossible.

The apparatus and/or system described above may be implemented as a hardware component, a software component, and/or a combination of the hardware component and the software component. The devices and components described in the embodiments may include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU). It may be implemented using one or more general purpose or special purpose computers, such as a logic unit, microprocessor, or any other device capable of executing and responding to instructions. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. A processing device may also access, store, manipulate, process, and generate data in response to execution of the software. For convenience of understanding, although one processing device is sometimes described as being used, one of ordinary skill in the art will recognize that the processing device includes a plurality of processing elements and/or a plurality of types of processing elements. It can be seen that can include For example, the processing device may include a plurality of processors or one processor and one controller. Other processing configurations are also possible, such as parallel processors.

The software may include a computer program, code, instructions, or a combination of one or more thereof, which configures the processing device to operate as desired or, independently or collectively, the processing device can be ordered. The software and/or data may be any kind of machine, component, physical device, virtual equipment, computer storage medium or apparatus, to be interpreted by or to provide instructions or data to the processing device. , or may be permanently or temporarily embody in a transmitted signal wave. The software may be distributed over networked computer systems and stored or executed in a distributed manner. Software and data may be stored in one or more computer-readable recording media.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the embodiment, or may be known and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic such as floppy disks. - includes magneto-optical media, and hardware devices specially configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

As described above, although the embodiments have been described with reference to the limited embodiments and drawings, various modifications and variations are possible from the above description by those skilled in the art. For example, the described techniques are performed in an order different from the described method, and/or the described components of the system, structure, apparatus, circuit, etc. are combined or combined in a different form than the described method, or other components Or substituted or substituted by equivalents may achieve an appropriate result.

Therefore, other implementations, other embodiments, and equivalents to the claims are also within the scope of the following claims.

100: noise canceling device
110: memory
120: processor
130: communication module
140: lidar sensor

Claims (15)

A method of removing noise performed by a computing device, the method comprising:
obtaining point cloud data including a plurality of points from a lidar sensor for obtaining three-dimensional data;
estimating a restoration error value for each point of the plurality of points using a first artificial neural network learned in advance;
calculating the restoration error value for each point of the plurality of points using a second artificial neural network learned in advance;
a noise determination step of classifying a noise point and an effective point with respect to the plurality of points based on the restoration error value and a predetermined threshold value; and
removing the noise point from among the point cloud data and outputting three-dimensional data including the valid points;
The first artificial neural network is learned using the restoration error value of the second artificial neural network,
Calculating the restoration error value comprises:
distinguishing a first point to be restored from among the plurality of points and a second point used for restoration;
restoring the position of the first point by using the second point; and
calculating a restoration error value of the first point;
including,
and the second point is a set of a predetermined number of points adjacent to the first point in a three-dimensional space. delete delete delete According to claim 1,
Calculating the restoration error value comprises:
obtaining actual coordinate information of the first point;
obtaining restoration coordinate information of the first point; and
calculating a three-dimensional distance value between the actual coordinate information and the restored coordinate information;
Including, a noise removal method. According to claim 1,
The step of estimating the restoration error value,
determining a third point to be estimated;
inputting three-dimensional coordinate information of the third point into the first artificial neural network; and
obtaining the restoration error value of the third point from the first artificial neural network;
Including, a noise removal method. 7. The method of claim 6,
The noise determination step may include determining the third point as the noise point when the restoration error value of the third point exceeds the predetermined threshold value. A computer program stored in a recording medium to execute the method according to any one of claims 1 to 8 using a computing device. processor; including;
The processor acquires point cloud data including a plurality of points from a lidar sensor that acquires three-dimensional data, and estimates a restoration error value for each point of the plurality of points using a first artificial neural network learned in advance and calculating the restoration error value for each point of the plurality of points using a second artificial neural network learned in advance, and a noise point and a noise point for the plurality of points based on the restoration error value and a predetermined threshold value A valid point is distinguished, the noise point is removed from the point cloud data, and three-dimensional data composed of the valid point is output, and a first point to be restored among the plurality of points and a second point used for restoration are selected. classify, restore the position of the first point using the second point, calculate the restoration error value of the first point,
the second point is a set of a predetermined number of points adjacent to the first point in a three-dimensional space;
The first artificial neural network is learned using the restoration error value of the second artificial neural network,
noise canceling device. delete delete delete 10. The method of claim 9,
The processor obtains real coordinate information of the first point, obtains restored coordinate information of the first point, and calculates a three-dimensional distance value between the real coordinate information and the restored coordinate information, a noise removal device . 10. The method of claim 9,
The processor determines a third point to be estimated, inputs three-dimensional coordinate information of the third point to the first artificial neural network, and obtains the restoration error value of the third point from the first artificial neural network Acquiring, noise canceling device. 15. The method of claim 14,
The processor is configured to determine the third point as the noise point when the restoration error value of the third point exceeds the predetermined threshold value.

Images