Feature recognition device, feature recognition method, program, and storage medium

Abstract

To provide a feature recognition device capable of efficiently recognizing features from information acquired by an external sensor.SOLUTION: A feature recognition device provided herein is configured to: acquire external information from an external detection device provided on a vehicle and acquire feature location information indicative of a location of a feature near the vehicle; extract prediction information representing information belonging to the external information predicted to contain information of the feature; and recognize the feature based on the prediction information.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a technique for recognizing ground objects.

Background Art

[0002] Conventionally, as a technique for estimating the vehicle position, for example, Patent Document 1 is known. The method of Patent Document 1 detects the positions of a plurality of landmarks with respect to a vehicle based on the distances and directions of the plurality of landmarks with respect to the vehicle, and estimates the current position of the vehicle based on the collation result between the positions of at least one pair of landmarks among the detected plurality of landmarks and the positions of the landmarks on the map.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In such a technique, when detecting the position of a landmark from the information obtained by an external sensor, it is common to perform a process of extracting and recognizing the target landmark from the entire measured information, and there is room for improvement in terms of efficiency.

[0005] Examples of the problems to be solved by the present invention include the above. An object of the present invention is to provide a ground object recognition device capable of efficiently recognizing ground objects from information acquired by an external sensor.

Means for Solving the Problems

[0006] The invention according to the claim is a ground object recognition device, comprising: a first acquisition unit that acquires external information generated by a sensor disposed on a moving body; a second acquisition unit that acquires ground object position information indicating the positions of ground objects existing around the moving body and attribute information including type information indicating the types of the ground objects; and a recognition unit that recognizes the ground objects by a recognition method for recognizing the ground objects from a ground object prediction range that is a range in the external information where the ground objects are predicted to exist and that is determined based on the ground object position information, and that is determined based on the type information.

[0007] The invention according to the claim is a ground object recognition method, comprising: a first acquisition step of acquiring external information generated by a sensor disposed on a moving body; a second acquisition step of acquiring ground object position information indicating the positions of ground objects existing around the moving body and attribute information including type information indicating the types of the ground objects; and a recognition step of recognizing the ground objects by a recognition method for recognizing the ground objects from a ground object prediction range that is a range in the external information where the ground objects are predicted to exist and that is determined based on the ground object position information, and that is determined based on the type information.

[0008] The program executed by a ground object recognition device including a computer according to the claim causes the computer to function as: a first acquisition unit that acquires external information generated by a sensor disposed on a moving body; a second acquisition unit that acquires ground object position information indicating the positions of ground objects existing around the moving body and attribute information including type information indicating the types of the ground objects; and a recognition unit that recognizes the ground objects by a recognition method for recognizing the ground objects from a ground object prediction range that is a range in the external information where the ground objects are predicted to exist and that is determined based on the ground object position information, and that is determined based on the type information.

Brief Description of the Drawings

[0009]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Embodiments for Carrying Out the Invention

[0010] In a preferred embodiment of the present invention, the ground object recognition device includes a first acquisition unit that acquires external information output by an external detection device disposed on a moving body, a second acquisition unit that acquires ground object position information indicating the positions of ground objects existing around the moving body, and a recognition unit that extracts prediction information, which is information predicted to include information indicating the ground object among the external information, based on the ground object position information, and recognizes the ground object based on the prediction information.

[0011] The above ground object recognition device acquires external information from an external detection device disposed on a moving body and acquires ground object position information indicating the positions of ground objects existing around the moving body. Then, based on the ground object position information, it extracts prediction information, which is information predicted to include information indicating the ground object among the external information, and recognizes the ground object based on the prediction information. Thereby, it is only necessary to recognize the ground object based on the prediction information predicted to include the information indicating the ground object among the external information, and it is not necessary to perform the recognition process of the ground object for all the external information, so that the ground object can be recognized efficiently.

[0012] In one aspect of the above ground object recognition device, the second acquisition unit further acquires attribute information indicating the attributes of the ground object, and the recognition unit recognizes the ground object by a recognition method determined based on the attribute information. In this aspect, the recognition method is determined based on the attribute information indicating the attributes of the ground object, and the ground object is recognized.

[0013] In another aspect of the above-described ground object recognition device, the second acquisition unit acquires first ground object position information indicating the position of a first ground object, first attribute information indicating the attribute of the first ground object, second ground object position information indicating the position of a second ground object different from the first ground object, and second attribute information indicating the attribute of the second ground object. The recognition unit extracts first prediction information predicted to include information indicating the first ground object among the external information based on the first ground object position information, and recognizes the first ground object by a first recognition method determined based on the first attribute information based on the first prediction information. The recognition unit extracts second prediction information predicted to include information indicating the second ground object among the external information based on the second ground object position information, and recognizes the second ground object by a second recognition method determined based on the second attribute information based on the second prediction information. In this aspect, the first recognition method and the second recognition method are determined based on the attribute information indicating the attributes of the ground objects, and the first ground object and the second ground object are recognized.

[0014] In one aspect of the above-described ground object recognition device, the recognition unit recognizes the ground object by a different recognition method for each attribute of the ground object. Thereby, it becomes possible to recognize the ground object by a recognition method suitable for the attribute of the ground object.

[0015] In another preferred embodiment of the present invention, a ground object recognition method executed by a ground object recognition device includes a first acquisition step of acquiring external information output by an external detection device disposed on a moving body, a second acquisition step of acquiring ground object position information indicating the position of a ground object existing around the moving body, and a recognition step of extracting prediction information, which is information predicted to include information indicating the ground object among the external information based on the ground object position information, and recognizing the ground object based on the prediction information. Also by this method, it is only necessary to recognize the ground object based on the prediction information predicted to include information indicating the ground object among the external information, and it is not necessary to perform the recognition process of the ground object for all the external information, so that it becomes possible to efficiently recognize the ground object.

[0016] In another preferred embodiment of the present invention, a program executed by a feature recognition device including a computer causes the computer to function as a first acquisition unit that acquires external information output by an external detection device disposed on a moving object, a second acquisition unit that acquires feature position information indicating the positions of features existing around the moving object, a recognition unit that extracts prediction information, which is information predicted to include information indicating the features among the external information, based on the feature position information, and recognizes the features based on the prediction information. By executing this program on a computer, the above-described feature recognition device can be realized. This program can be stored and handled in a storage medium.

Example

[0017] Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings. [Configuration] FIG. 1 shows a schematic configuration of a host vehicle position estimation device to which the feature recognition device of the present invention is applied. The host vehicle position estimation device 10 is mounted on a vehicle and configured to be communicable with a server 7 such as a cloud server by wireless communication. The server 7 is connected to a database 8, and the database 8 stores a high-precision map. The host vehicle position estimation device 10 communicates with the server 7 and downloads landmark information regarding landmarks around the host vehicle position of the vehicle.

[0018] The advanced map stored in the database 8 stores landmark information for each landmark. Here, the landmark information includes, for each landmark ID that identifies a landmark, a landmark map position indicating the position of the landmark on the map and a landmark attribute indicating the type and features of the landmark. The landmark attribute is prepared according to the type and features of the landmark and basically includes the type of the landmark and feature information. For example, when the landmark is a sign, the landmark attribute includes "sign" as the type of the landmark and, as feature information, the shape (circle, triangle, square, etc.) and size of the sign, the reflection intensity of the sign, and the like. When the landmark is a utility pole, the landmark attribute includes "utility pole" as the type of the landmark and, as feature information, the shape (curvature of the cross section, radius, etc.) and size of the utility pole. Other examples of landmarks include road surface markers, traffic lights, and various buildings.

[0019] On the other hand, the host vehicle position estimation device 10 includes an internal sensor 11, an external sensor 12, a host vehicle position prediction unit 13, a measurement information acquisition unit 14, a communication unit 15, a landmark information acquisition unit 16, a landmark prediction unit 17, an extraction method selection unit 18, a landmark extraction unit 19, a correlation unit 20, and a host vehicle position estimation unit 21. The host vehicle position prediction unit 13, the measurement information acquisition unit 14, the landmark information acquisition unit 16, the landmark prediction unit 17, the extraction method selection unit 18, the landmark extraction unit 19, the correlation unit 20, and the host vehicle position estimation unit 21 are actually realized by a computer such as a CPU executing a program prepared in advance.

[0020] The internal sensor 11 measures the host vehicle position of the vehicle as a GNSS (Global Navigation Satellite System) / IMU (Inertia Measurement Unit) integrated navigation system and includes a satellite positioning sensor (GPS), a gyro sensor, a vehicle speed sensor, and the like. The host vehicle position prediction unit 13 predicts the host vehicle position of the vehicle by GNSS / IMU integrated navigation based on the output of the internal sensor 11 and supplies the predicted host vehicle position to the landmark prediction unit 17.

[0021] On the one hand, the external sensor 12 is a sensor that detects objects around the vehicle and includes a stereo camera, Lidar (Light Detection and Ranging), etc. The measurement information acquisition unit 14 acquires measurement information from the external sensor 12 and supplies it to the landmark extraction unit 19.

[0022] The communication unit 15 is a communication unit for wireless communication with the server 7. The landmark information acquisition unit 16 receives landmark information regarding landmarks existing around the vehicle from the server 7 via the communication unit 15 and supplies it to the landmark prediction unit 17. Also, the landmark information acquisition unit 16 supplies the landmark map positions included in the landmark information to the association unit 20.

[0023] Based on the landmark map positions included in the landmark information and the predicted own vehicle position acquired from the own vehicle position prediction unit 13, the landmark prediction unit 17 determines a landmark prediction range, which is a range where a landmark is predicted to exist, and supplies it to the landmark extraction unit 19. Also, the landmark prediction unit 17 supplies the landmark attributes included in the landmark information to the extraction method selection unit 18.

[0024] The extraction method selection unit 18 determines the feature extraction method that the landmark extraction unit 19 executes to extract a landmark based on the landmark attributes. As described above, landmarks include types such as signs and utility poles, and the method for extracting each type of landmark is different. That is, the feature extraction method executed by the landmark extraction unit 19 differs for each type of landmark. Specifically, the landmark extraction unit 19 extracts a feature object having the features indicated by the feature information included in the landmark attributes as a landmark. Therefore, the extraction method selection unit 18 selects the feature extraction method corresponding to the landmark based on the landmark attributes of the landmark to be extracted and instructs the landmark extraction unit 19.

[0025] The landmark extraction unit 19 extracts landmarks based on the landmark prediction range supplied from the landmark prediction unit 17 and the measurement information supplied from the measurement information acquisition unit 14. Specifically, the landmark extraction unit 19 extracts feature objects from the measurement information included in the landmark prediction range and uses these as landmarks. At this time, the landmark extraction unit 19 extracts landmarks by the feature object extraction method determined by the extraction method selection unit 18, that is, the feature object extraction method corresponding to the landmark attribute. Then, the landmark extraction unit 19 outputs the landmark measurement position of the extracted landmark to the association unit 20.

[0026] The association unit 20 stores, in association with a landmark ID, the landmark measurement position acquired from the landmark extraction unit 19 and the landmark map position acquired from the landmark information acquisition unit 16. Thereby, for each landmark ID, information in which the landmark map position and the landmark measurement position are associated (hereinafter referred to as "association information") is generated.

[0027] Then, the host vehicle position estimation unit 21 estimates the host vehicle position and the host vehicle azimuth angle of the vehicle using the landmark map position and the landmark measurement position included in the association information for at least two landmarks.

[0028] In the above configuration, the external sensor 12 is an example of the external detection device of the present invention, the measurement information acquisition unit 14 is an example of the first acquisition unit in the present invention, the landmark information acquisition unit 16 is an example of the second acquisition unit in the present invention, and the landmark extraction unit 19 is an example of the recognition unit in the present invention. Also, the landmark map position corresponds to the ground feature position information in the present invention, the measurement information corresponds to the external information in the present invention, and the measurement information in the landmark prediction range corresponds to the prediction information in the present invention.

[0029] [Determination of Landmark Prediction Range] Next, the method for determining the landmark prediction range performed by the landmark prediction unit 17 will be described in detail. FIG. 2 is a diagram for explaining the method for determining the landmark prediction range. As shown in the figure, in the map coordinate system (X m , Y m ), the vehicle 5 exists, and a vehicle coordinate system (X v , Y v ) is defined based on the position of the vehicle 5. Specifically, the traveling direction of the vehicle 5 is set as the X v axis of the vehicle coordinate system, and the direction perpendicular to it is set as the Y v axis of the vehicle coordinate system.

[0030] There are landmarks L1 and L2 around the vehicle 5. In this embodiment, it is assumed that the landmark L1 is a sign and the landmark L2 is a utility pole. The positions of the landmarks L1 and L2 in the map coordinate system, that is, the landmark map positions, are included in the enhanced map as described above and are supplied from the landmark information acquisition unit 16. In FIG. 2, the landmark map position of the landmark L1 is P LM1 (mx m1 , my m1 ), and the landmark map position of the landmark L2 is P LM2 (mx m2 , my m2 ). On the other hand, the predicted own vehicle position P' VM (x' m , y' m ) is supplied from the own vehicle position prediction unit 13.

[0031] The landmark prediction unit 17 calculates the landmark prediction position P' VM in the vehicle coordinate system based on the predicted own vehicle position P' LM1 and the landmark map position P of the landmark L1 LV1 (l'x v1 , l'y v1 ), and determines the landmark prediction range R LV1 with the landmark prediction position P' L1 as a reference. Similarly, the landmark prediction unit 17 calculates the landmark prediction position P' VM in the vehicle coordinate system based on the predicted own vehicle position P' LM2Based on this, the predicted position P' of the landmark in the vehicle coordinate system LV2 (l’x v2 ,l’y v2 ) is calculated, and the landmark prediction range R LV2 is determined with the landmark predicted position P' L2 as a reference.

[0032] The landmark prediction range R indicates the range where the landmark L is predicted to exist. Since the predicted own vehicle position P' obtained using the internal sensor 11 contains a certain degree of error, the landmark prediction unit 17 determines the landmark prediction range R in consideration of this error. For example, the landmark prediction unit 17 determines a circle with a predetermined distance centered on the landmark predicted position P' VM as the landmark prediction range R. LV

[0033] By determining the landmark prediction range R in this way, the landmark extraction unit 19 may extract the landmark based on the measurement information belonging to the landmark prediction range R among the measurement information obtained by the external sensor 12. Generally, an external sensor such as Lidar performs measurements over a wide range, such as the entire circumference (360°) of the own vehicle position or 270° excluding the rear of the vehicle, and generates measurement information over a wide range. In this case, if the landmark extraction process is executed for all the obtained wide-range measurement information to extract the landmark, the amount of calculation will become extremely large. On the contrary, in this embodiment, based on the predicted own vehicle position obtained using the internal sensor 11, the range where the landmark is predicted to exist is determined as the landmark prediction range R, and the landmark extraction process is executed only for the measurement information belonging to that range, thereby significantly reducing the amount of calculation and enabling the landmark to be detected efficiently.

[0034] In actual processing, the external sensor 12 performs wide-range measurement as described above and outputs wide-range measurement information. The landmark extraction unit 19 may extract only the measurement information within the landmark prediction range R among them and use it as the target of the landmark extraction process. Instead, the external sensor 12 may be controlled to measure only within the landmark prediction range R.

[0035] [Landmark Extraction Process] Next, the landmark extraction process will be described. FIG. 3 shows an example of the landmark extraction process. In this embodiment, it is assumed that the landmark is a sign. In this case, the landmark attribute included in the landmark information acquired by the landmark information acquisition unit 16 from the server 7 includes "sign" as the landmark type. Based on this information, the extraction method selection unit 18 selects a feature extraction method corresponding to the sign and instructs the landmark extraction unit 19.

[0036] The landmark extraction unit 19 extracts landmarks based on the measurement information acquired from the measurement information acquisition unit 14. The Lidar as the external sensor 12 emits light pulses around, receives the reflected light pulses from the objects existing around, and generates point cloud data. The point cloud data includes the positions (3D positions) of the objects around the host vehicle and the reflection intensity data. The measurement information acquisition unit 14 outputs this point cloud data to the landmark extraction unit 19 as measurement information. Note that the interval of the point cloud data generated by the Lidar depends on the distance and direction from the Lidar to the measurement target and the angular resolution of the Lidar.

[0037] As described above, when the landmark is a sign, its landmark attribute includes "sign" as the landmark type, and includes shape (size) information and reflection intensity as feature information. Therefore, the landmark extraction unit 19 extracts feature objects based on the shape information and the reflection intensity. Specifically, the landmark extraction unit 19 extracts circular point cloud data corresponding to the sign as shown in FIG. 3 from the point cloud data acquired by the measurement information acquisition unit 14 based on the shape information and the reflection intensity of the sign. Then, the landmark extraction unit 19 calculates the centroid position (lx, ly, lz) of the extracted point cloud data of the sign, and outputs this as the landmark measurement position to the association unit 20.

[0038] The association unit 20 associates the landmark measurement position acquired from the landmark extraction unit 19 with the landmark map position acquired from the landmark information acquisition unit 16. Specifically, the association unit 20 mutually associates the landmark ID corresponding to the landmark, the landmark measurement position, and the landmark map position to generate and store association information. In this way, the landmark measurement position and the landmark map position are associated with respect to the landmark extracted by the landmark extraction unit 19. In this embodiment, since the corresponding landmark map position is a position in a two-dimensional map coordinate system, the association unit 20 uses the two-dimensional position (lx, ly) among the calculated three-dimensional centroid positions (lx, ly, lz) as the landmark measurement position. The association information generated in this way is used for estimating the position of the host vehicle by the host vehicle position estimation unit 21.

[0039] In the example of FIG. 3, the landmark was a sign, but landmark extraction is similarly performed for other types of landmarks. However, the feature extraction method differs for each type of landmark. For example, when the landmark is a utility pole, the landmark attributes include information on the shape (such as the arc shape, curvature, and radius of the cross-section) and size of the utility pole as feature information. Therefore, the landmark extraction unit 19 extracts a feature corresponding to the utility pole from the measurement information based on the information on the shape and size of the utility pole as a landmark, and outputs its center position as the landmark measurement position. Also, when the landmark is a road marker, the landmark attributes include the shape information and reflection intensity of the road marker as feature information. Therefore, the landmark extraction unit 19 extracts a feature corresponding to the road marker from the measurement information based on the shape information and reflection intensity as a landmark, and outputs its center position as the landmark measurement position. Even when the landmark is another object such as a traffic signal, the landmark extraction unit 19 may extract the landmark based on the type and feature information of the landmark, and output a predetermined position such as its center position as the landmark measurement position.

[0040] Thus, in this embodiment, the attribute information of each landmark is included in the landmark information included in the high-precision map, and when performing landmark extraction from the measurement information by the external sensor 12, a feature extraction method corresponding to the attribute of the target landmark is used. Thereby, landmark extraction can be performed efficiently. That is, when the attributes of the landmark are not used as in this embodiment, since it is unknown what the landmark to be extracted is, as a feature extraction method, it is necessary to sequentially perform extraction methods suitable for all assumed landmarks such as a method suitable for a sign, a method suitable for a utility pole, and a method suitable for a road marker, and extract any landmark. On the other hand, if the attributes of the landmark are known in advance as in this embodiment, only the feature extraction method suitable for that landmark needs to be implemented. For example, in the example of FIG. 2, it is known from the landmark attributes that the landmark L1 is a sign and the landmark L2 is a utility pole. Therefore, the landmark prediction range R L1For this, perform feature extraction processing suitable for the sign, and the landmark prediction range R L2 For this, perform feature extraction processing suitable for the utility pole. In this way, it is possible to reduce the processing load for landmark extraction and prevent false detection and detection omission.

[0041] [Vehicle position estimation] Next, the vehicle position estimation by the vehicle position estimation unit 21 will be described. The vehicle position estimation unit 21 estimates the vehicle's own vehicle position and own vehicle azimuth angle using the correspondence information of the two landmarks generated by the association unit 20. Hereinafter, the own vehicle position obtained by the own vehicle position estimation is referred to as the "estimated own vehicle position", and the own vehicle azimuth angle obtained by the own vehicle position estimation is referred to as the "estimated own vehicle azimuth angle".

[0042] FIG. 4 shows an example of vehicle position estimation. The vehicle 5 is located on the map coordinate system (X m , Y m ), and a vehicle coordinate system (X v , Y v ) is defined with the position of the vehicle 5 as a reference. The estimated own vehicle position of the vehicle 5 is P VM (x m , y m ), and the estimated own vehicle azimuth angle is Ψ m .

[0043] The vehicle position estimation unit 21 acquires the correspondence information about the two landmarks L1 and L2 from the association unit 20. Specifically, the vehicle position estimation unit 21 acquires the landmark map position P LM1 (mx m1 , my m1 ) and the landmark measurement position P LV1 (lx v1 , ly v1 ) for the landmark L1, and acquires the landmark map position P LM2 (mx m2 , my m2 ) and the landmark measurement position P LV2 (lx v2 , ly v2 ) for the landmark L2.

[0044] Using these landmark map positions and landmark measurement positions, the estimated own vehicle azimuth angle Ψ m is obtained by the following formula.

[0045]

Equation

[0046]

Equation

[0047] [Own Vehicle Position Estimation Process] Next, the processing flow by the own vehicle position estimation device 10 will be described. FIG. 5 is a flowchart of the processing by the own vehicle position estimation device 10. This processing is realized by a computer such as a CPU executing a program prepared in advance and functioning as each component shown in FIG. 1.

[0048] First, the own vehicle position prediction unit 13 acquires a predicted own vehicle position P’ VM based on the output from the internal sensor 11 (step S11). Next, the landmark information acquisition unit 16 connects to the server 7 through the communication unit 15 and acquires landmark information from the enhanced map stored in the database 8 (step S12). As described above, the landmark information includes the landmark map position and the landmark attribute, and the landmark information acquisition unit 16 supplies the landmark map position to the association unit 20. Note that either step S11 or S12 may be performed first.

[0049] Next, based on the landmark map position included in the landmark information obtained in step S12 and the predicted own vehicle position obtained in step S11, the landmark prediction unit 16 determines a landmark prediction range R and supplies it to the landmark extraction unit 19 (step S13).

[0050] Next, the extraction method selection unit 18 selects a feature extraction method based on the landmark attribute included in the landmark information obtained in step S12 (step S14). Specifically, the extraction method selection unit 18 selects a feature extraction method suitable for the type of landmark indicated by the landmark attribute. For example, when the landmark is a sign, the extraction method selection unit 18 selects a feature extraction method suitable for the sign as described with reference to FIG. 3 and instructs the landmark extraction unit 19.

[0051] On the other hand, the landmark extraction unit 19 acquires the measurement information output by the external sensor 12 from the measurement information acquisition unit 14 (step S15), and extracts landmarks within the landmark prediction range R obtained from the landmark prediction unit 17 by the feature extraction method instructed by the extraction method selection unit 18 (step S16). The landmark extraction unit 19 outputs the landmark measurement position to the association unit 20 as a landmark extraction result.

[0052] The association unit 20 associates the landmark map position acquired from the landmark information acquisition unit 16 and the landmark measurement position acquired from the landmark extraction unit 19 for each landmark to generate association information, and sends it to the own vehicle position estimation unit 21 (step S17). Then, the own vehicle position estimation unit 21 estimates the own vehicle position and the own vehicle azimuth angle by the method described with reference to FIG. 5 using the association information for two landmarks (step S18). In this way, the estimated own vehicle position and the estimated own vehicle azimuth angle are output.

Explanation of Signs

[0053] 5 Vehicle 7 Server 8 Database 10 Own Vehicle Position Estimation Device 11 Inner sensor 12 Outer sensor 13 Bicycle position prediction unit 14 Measurement information acquisition unit 17 Landmark prediction unit 18 Extraction method selection unit 19 Landmark extraction unit 21 Bicycle position estimation unit

Claims

1. A first acquisition unit that acquires external environment information generated by a sensor disposed in the moving object; a second acquisition unit that acquires feature position information indicating a position of a feature present around the moving object and attribute information including type information indicating a type of the feature; a recognition unit that recognizes the feature by a recognition method that recognizes the feature from a feature prediction range, which is a range in which the feature is predicted to exist in the external environment information and is determined based on the feature position information, and which is determined based on the type information; A feature recognition device comprising:

2. The attribute information further includes characteristic information determined according to the type, The feature recognition device according to claim 1 , wherein the recognition unit recognizes the feature based on the feature information.

3. 2. The feature recognition device according to claim 1, wherein the sensor generates outside world information by using light reflected by the feature.

4. 4. The feature recognition device according to claim 3, wherein the sensor has an irradiation unit that irradiates the light, and generates outside world information using light that is reflected by the feature.

5. A first acquisition step of acquiring external environment information generated by a sensor disposed in the moving object; a second acquisition step of acquiring feature position information indicating a position of a feature existing around the moving object and attribute information including type information indicating a type of the feature; a recognition step of recognizing the feature by a recognition method for recognizing the feature from a feature prediction range, which is a range in the external environment information where the feature is predicted to exist and is determined based on the feature position information, and which is determined based on the type information; A feature recognition method comprising:

6. A program executed by a feature recognition device having a computer includes: A first acquisition unit that acquires external environment information generated by a sensor disposed in the moving object; a second acquisition unit that acquires feature position information indicating a position of a feature existing around the moving object and attribute information including type information indicating a type of the feature; a recognition unit that recognizes the feature by a recognition method that recognizes the feature from a feature prediction range, which is a range in the external environment information where the feature is predicted to exist and is determined based on the feature position information, and that is determined based on the type information; A program for causing the computer to function as a

7. A storage medium storing the program according to claim 6.

Images