Object recognition device and method thereof

Abstract

The invention relates to an object recognition apparatus and a method thereof. The object recognition apparatus includes a LIDAR sensor and a processor. The processor may identify whether the object is a dynamic object, and identify a shape of a distribution of contour points among the first shape, the second shape, and the third shape, and assign a flag corresponding to the identified shape to a specific layer. The reference line segment may include a longer line segment of a line segment connecting the peak point and a first end point, which is one of two outermost end points of the contour points, and a line segment connecting the peak point and a second end point different from the first end point, which is one of the two outermost end points.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims the benefit of priority from Korean Patent Application No. 10-2023-0131148 filed in the Korean Intellectual Property Office on September 27, 2023, the entire contents of which are incorporated herein by reference. Technical Field

[0003] The present disclosure relates to an object recognition apparatus and method, and more particularly, to a technology for representing the shape of an object based on contour points obtained by light detection and ranging (LIDAR). Background Art

[0004] In autonomous vehicles, competition to develop technology that detects the surrounding environment through sensors is intensifying. Compared with radar, LIDAR has the advantage of being able to accurately detect the surrounding environment. Therefore, technology that can interpret data obtained through LIDAR is attracting attention.

[0005] The vehicle can obtain data indicating the position of objects around the vehicle through LIDAR. For example, the distance from the LIDAR sensor to the object can be obtained by the interval between the time when the LIDAR sensor emits laser light and the time when the laser light reflected by the object is received. The vehicle can recognize the position of a point included in the object in the space where the vehicle is located based on the angle of the emitted laser light and the distance to the object.

[0006] For precise measurement, LIDAR systems with high angular resolution have been used in autonomous vehicles, and the amount of data acquired by LIDAR has increased accordingly. The importance of technology for processing data obtained from LIDAR is increasing to recognize the shape of objects around the vehicle from the data. Summary of the invention

[0007] Aspects of the present invention provide an object recognition apparatus and method for recognizing various shapes of distribution of contour points.

[0008] Aspects of the present invention provide an object recognition apparatus and method for recognizing shapes corresponding to a plurality of layers.

[0009] Aspects of the present invention provide an object recognition apparatus and method for improving accuracy of object recognition.

[0010] According to one aspect of the present disclosure, an object recognition device includes: a LIDAR sensor configured to obtain contour points of an object; and a processor.

[0011] In some implementations, the processor may identify whether the object is a dynamic object based on at least one of a reference line segment, a non-reference line segment, and a distribution of contour points included in a specific layer among a plurality of layers representing the object, or any combination thereof, and identify a shape of the distribution of contour points among the first shape, the second shape, and the third shape based on identifying that the object is a dynamic object, the length of the reference line segment is greater than a reference length, the angle between the reference line segment and the non-reference line segment satisfies a specified first angle range, the angle between the reference line segment and the non-reference line segment does not satisfy a specified first angle range, or the length of the reference line segment is less than or equal to the reference length, or any combination thereof, and assign a flag corresponding to the identified shape to the specific layer. The reference line segment may include a longer line segment among a line segment connecting a peak point and a first endpoint and a line segment connecting a peak point and a second endpoint different from the first endpoint, the first endpoint being one of two outermost endpoints among the contour points, and the second endpoint being one of the two outermost endpoints. The non-reference line segment may include a shorter line segment among a line segment connecting the first endpoint and the peak point and a line segment connecting the second endpoint and the peak point. The peak point may include a contour point located farthest from a line segment connecting the first endpoint and the second endpoint among the contour points.

[0012] In some embodiments, the processor may identify whether the object is a dynamic object based on identifying that the size of the contour box including the contour points included in a specific layer among the multiple layers representing the object satisfies the first reference size range, and based on at least one of the reference line segments, the non-reference line segments, or the distribution of the contour points, or any combination thereof; identify whether the length of the reference line segment is greater than the reference length; assign a flag to a specific layer corresponding to the first shape based on identifying that the object is a dynamic object, the length of the reference line segment is greater than the reference length, and the angle between the reference line segment and the non-reference line segment satisfies the specified first angle range, and identifying that the shape of the distribution of the contour points is a first shape; assign a flag to a specific layer corresponding to the first shape based on identifying that the object is a dynamic object, the length of the reference line segment is greater than the reference length, and the angle between the reference line segment and the non-reference line segment satisfies the specified first angle range; The method comprises the following steps: identifying whether the shape of the distribution of the contour points is a second shape based on the recognition that the object is a dynamic object and the length of the reference line segment is less than or equal to the reference length, and assigning a flag corresponding to the second shape to a specific layer; identifying whether the shape of the distribution of the contour points is a third shape based on the recognition that the object is a dynamic object and the length of the reference line segment is less than or equal to the reference length, and assigning a flag corresponding to the third shape to a specific layer when the shape of the distribution of the contour points is recognized as the third shape; and assigning a flag corresponding to the fourth shape to a specific layer when the shape of the distribution of the contour points is not recognized as one of the first shape, the second shape, and the third shape. The reference line segment may include a longer line segment among a line segment connecting a peak point and a first endpoint and a line segment connecting a peak point and a second endpoint different from the first endpoint, the first endpoint being one of the two outermost endpoints in the contour point, and the second endpoint being one of the two outermost endpoints. The non-reference line segment may include a shorter line segment among a line segment connecting a first endpoint and a peak point and a line segment connecting a second endpoint and a peak point. The peak point may include a contour point located farthest from a line segment connecting the first endpoint and the second endpoint among the contour points.

[0013] In some implementations, when the shape of the distribution of contour points is not recognized as one of the first shape, the second shape, and the third shape, the processor may assign a flag corresponding to a fourth shape to a specific layer. The fourth shape may indicate that the shape of the distribution of contour points is not recognized as one of the first shape, the second shape, and the third shape.

[0014] In some embodiments, the processor may identify whether the object is a dynamic object based on at least one of reference line segments, non-reference line segments, and distribution of contour points, or any combination thereof, based on identifying that the size of the contour box containing the contour points satisfies a first reference size range.

[0015] In some implementations, the processor may identify whether the object is a dynamic object based on at least one of the reference line segment, the distribution of the contour points, and the length of the reference line segment, or any combination thereof, based on identifying that the size of the contour box including the contour point does not satisfy the first reference size range but satisfies the second reference size range; identify whether a first average value is less than a first average threshold, the first average value being an average value of at least one distance between the reference line segment and at least one contour point included in an area to which the reference line segment belongs, the area to which the reference line segment belongs being an area to which the reference line segment belongs in two areas separated by a straight line that bisects an angle between the reference line segment and the non-reference line segment and passes through a peak point in the contour points; and identify whether a first variance value as a variance value of the at least one distance is less than a first average threshold. is less than a first variance threshold; based on the object being a dynamic object, the first average value being less than the first average threshold, and the first variance value being less than the first variance threshold, identifying that the shape of the distribution of the contour points is a second shape, and assigning a flag corresponding to the second shape to a specific layer; based on the object being a dynamic object, and the first variance value being greater than or equal to the first variance threshold or the first average value being greater than or equal to the first average threshold, identifying whether the shape of the distribution of the contour points is a third shape; and when the shape of the distribution of the contour points is identified as the third shape, assigning a flag corresponding to the third shape to a specific layer; and when the shape of the distribution of the contour points is not identified as one of the first shape, the second shape, and the third shape, assigning a flag corresponding to the fourth shape to a specific layer.

[0016] In some embodiments, the processor may identify whether the object represented by the contour point is a dynamic object based on at least one of the reference line segment, the distribution of the contour points, and the length of the reference line segment, or any combination thereof, based on identifying that the size of the contour box does not satisfy the first reference size range but satisfies the second reference size range; identify whether a first average value is less than a first average threshold, the first average value is an average value of at least one distance between the reference line segment and at least one contour point included in the area to which the reference line segment belongs, the area to which the reference line segment belongs is the area to which the reference line segment belongs among two areas separated by a straight line that bisects the angle between the reference line segment and the non-reference line segment and passes through a peak point in the contour points; identify whether a first variance value, which is a variance value of the distance, is less than a first variance threshold. based on the first variance threshold; based on the fact that the object is a dynamic object, the first average value is less than the first average threshold, and the first variance value is less than the first variance threshold, identifying that the shape of the distribution of contour points is a second shape, and assigning a flag corresponding to the second shape to a specific layer; based on the fact that the object is a dynamic object, and the first variance value is greater than or equal to the first variance threshold or the first average value is greater than or equal to the first average threshold, identifying whether the shape of the distribution of contour points is a third shape, and when the shape of the distribution of contour points is identified as the third shape, assigning a flag corresponding to the third shape to a specific layer; and when the shape of the distribution of contour points is not identified as one of the first shape, the second shape and the third shape, assigning a flag corresponding to a fourth shape to a specific layer.

[0017] In some implementations, the first shape may represent a shape including a first L-shaped distribution identified by all or part of the contour points. The second shape may represent a shape including an I-shaped distribution identified by all or part of the contour points. The third shape may represent a shape including a second L-shaped distribution identified by all or part of the contour points, the length of a reference line segment in the second L-shaped distribution being shorter than the length of a reference line segment in the first L-shaped distribution corresponding to the first shape. The fourth shape may indicate that the distribution of the contour points is not identified as one of the first shape, the second shape, and the third shape.

[0018] In some embodiments, the processor may, based on recognizing that the object is a dynamic object and the length of the reference line segment is less than or equal to the reference length, recognize that the ratio of the length of the line segment in one direction of the clustering box identified based on the multiple layers to the length of the line segment in the one direction of the contour box containing the contour point is less than a reference ratio, recognize that the angle between the reference line segment and the non-reference line segment satisfies a second angle range, recognize that the shape of the distribution of the contour points is a third shape, and assign a marker corresponding to the third shape to a specific layer based on recognizing that the ratio is less than the reference ratio and the angle between the reference line segment and the non-reference line segment satisfies the second angle range.

[0019] In some implementations, the processor may identify whether a first average value is less than or equal to a second average threshold value based on identifying that the object is a dynamic object, the length of the reference line segment is greater than the reference length, and the angle between the reference line segment and the non-reference line segment does not satisfy the first angle range, the first average value being an average value of at least one distance between the reference line segment and at least one contour point included in the area to which the reference line segment belongs, the area to which the reference line segment belongs being an area to which the reference line segment belongs among two areas separated by a straight line that bisects the angle between the reference line segment and the non-reference line segment and passes through a peak point in the contour points, identify whether the angle between the reference line segment and the non-reference line segment satisfies a third angle range, and identify that a shape of the distribution of contour points is a second shape based on identifying that the first average value is less than or equal to the second average threshold value and that the angle between the reference line segment and the non-reference line segment satisfies the third angle range, and assign a marker corresponding to the second shape to a specific layer.

[0020] In some implementations, the processor may identify whether there is at least one contour point in a region between a straight line passing through an endpoint included in a non-reference line segment and perpendicular to the non-reference line segment and a straight line passing through a peak point and perpendicular to the non-reference line segment based on identifying that the object is a dynamic object, the first average value is greater than or equal to a first average threshold, or the first variance value is greater than or equal to a first variance threshold, identify whether the first average value is less than a third average threshold, identify whether the first variance value is less than a second variance threshold, and identify whether the second average value is less than a fourth average threshold, the second average value being an average value of at least one distance between the non-reference line segment and at least one contour point included in a region to which the non-reference line segment belongs, the region to which the non-reference line segment belongs being one of two regions separated by a straight line that bisects an angle between the reference line segment and the non-reference line segment and passes through a peak point in the contour points. an area, identifying whether a second variance value is less than a third variance threshold, the second variance value being a variance of at least one distance between the non-reference line segment and at least one contour point in the area to which an endpoint included in the non-reference line segment belongs, identifying whether an angle between the reference line segment and the non-reference line segment satisfies a second angle range, and based on identifying that at least one contour point exists in an area between a straight line passing through an endpoint included in the non-reference line segment and perpendicular to the non-reference line segment and a straight line passing through a peak point and perpendicular to the non-reference line segment, the first average value is less than a third average threshold, the second average value is less than a fourth average threshold, the first variance value is less than the second variance threshold, the second variance value is less than the third variance threshold, and the angle between the reference line segment and the non-reference line segment satisfies the second angle range, identifying that a shape of distribution of contour points is a third shape, and assigning a flag corresponding to the third shape to a specific layer.

[0021] In some embodiments, among the line segments of the outline frame that meet the first reference size range, the length of the longer line segment among the line segments in the first direction and the line segments in the second direction different from the first direction may be greater than the first length value and less than the second length value greater than the first length value. The shorter line segment among the line segments in the first direction and the line segments in the second direction may be greater than or equal to the third length value. Among the line segments of the outline frame that meet the second reference size range, the length of the longer line segment among the line segments in the third direction and the line segments in the fourth direction different from the third direction may be greater than the first length value. The shorter line segment among the line segments in the third direction and the line segments in the fourth direction may be less than the fourth length value greater than the third length value.

[0022] In some implementations, the processor may identify a layer among multiple layers that is assigned a flag corresponding to a fourth shape, identify a layer among multiple layers that is assigned a flag corresponding to the first shape, identify whether the number of layers among multiple layers that are assigned a flag corresponding to the first shape is less than the number of layers that are assigned a flag corresponding to the second shape, and assign a flag corresponding to the second shape to the multiple layers based on identifying that the layer assigned a flag corresponding to the fourth shape is not identified, the layer assigned a flag corresponding to the first shape is identified, and the number of layers among multiple layers that are assigned a flag corresponding to the first shape is less than the number of layers that are assigned a flag corresponding to the second shape.

[0023] In some embodiments, the processor may identify a layer assigned with a flag corresponding to a fourth shape among multiple layers, identify a layer assigned with a flag corresponding to the first shape among multiple layers, identify a layer assigned with a flag corresponding to the second shape among multiple layers, and assign a flag corresponding to the second shape to the multiple layers based on the fact that the layer assigned with the flag corresponding to the fourth shape is not identified, the layer assigned with the flag corresponding to the first shape is not identified, and the layer assigned with the flag corresponding to the second shape is identified.

[0024] In some implementations, the processor may identify a layer among multiple layers that is assigned a flag corresponding to a fourth shape, identify a layer among multiple layers that is assigned a flag corresponding to a first shape, identify a layer among multiple layers that is assigned a flag corresponding to a second shape, identify a layer among multiple layers that is assigned a flag corresponding to a third shape, and assign a flag corresponding to the third shape to the multiple layers based on the fact that the layer assigned a flag corresponding to the fourth shape is not identified, the layer assigned a flag corresponding to the first shape is not identified, the layer assigned a flag corresponding to the second shape is not identified, and the layer assigned a flag corresponding to the third shape is identified.

[0025] According to one aspect of the present disclosure, an object recognition method includes: a processor identifies whether an object is a dynamic object based on at least one of a reference line segment, a non-reference line segment, and a distribution of contour points contained in a specific layer of a plurality of layers representing the object, or any combination thereof, wherein the contour points of the object are obtained by a LIDAR sensor; and the processor identifies the shape of the distribution of contour points among a first shape, a second shape, and a third shape based on at least one of identifying that the object is a dynamic object, the length of the reference line segment is greater than the reference length, the angle between the reference line segment and the non-reference line segment satisfies a specified first angle range, the angle between the reference line segment and the non-reference line segment does not satisfy the specified first angle range, and the length of the reference line segment is less than or equal to the reference length, or any combination thereof. The reference line segment may include a longer line segment among a line segment connecting a peak point and a first endpoint and a line segment connecting a peak point and a second endpoint different from the first endpoint, the first endpoint being one of the two outermost endpoints among the contour points, and the second endpoint being one of the two outermost endpoints. The non-reference line segment may include a line segment connecting the first endpoint and the peak point and a shorter line segment among a line segment connecting the second endpoint and the peak point. The peak point may include a contour point located farthest from a line segment connecting the first endpoint and the second endpoint among the contour points.

[0026] In some implementations, the object recognition method may further include, by the processor, identifying whether the object is a dynamic object based on at least one of reference line segments, non-reference line segments, and distribution of contour points, or any combination thereof, based on identifying that a size of a contour box including contour points included in a specific layer among multiple layers representing the object satisfies a first reference size range; identifying by the processor whether a length of the reference line segment is greater than a reference length, and assigning, by the processor, a flag corresponding to the first shape to the specific layer based on identifying that the object is a dynamic object, that the length of the reference line segment is greater than the reference length, and that an angle between the reference line segment and the non-reference line segment satisfies a specified first angle range, and identifying that a shape of the distribution of contour points is a first shape; and identifying by the processor, by the processor, assigning, by the processor, a flag corresponding to the first shape to the specific layer, based on identifying that the object is a dynamic object, that the length of the reference line segment is greater than the reference length, and that an angle between the reference line segment and the non-reference line segment satisfies a specified first angle range. The length of the reference line segment is greater than the reference length, and the angle between the reference line segment and the non-reference line segment does not satisfy the first angle range, to identify whether the shape of the distribution of the contour points is a second shape, when the shape of the distribution of the contour points is identified as the second shape, the processor assigns a flag corresponding to the second shape to a specific layer; the processor identifies whether the shape of the distribution of the contour points is a third shape based on identifying that the object is a dynamic object and the length of the reference line segment is less than or equal to the reference length, when the shape of the distribution of the contour points is identified as the third shape, the processor assigns a flag corresponding to the third shape to a specific layer; and when the shape of the distribution of the contour points is not identified as one of the first shape, the second shape and the third shape, the processor assigns a flag corresponding to the fourth shape to a specific layer. The reference line segment may include a line segment connecting the peak point and the first endpoint and a longer line segment among the line segments connecting the peak point and the second endpoint different from the first endpoint, the first endpoint being one of the two outermost endpoints in the contour point, and the second endpoint being one of the two outermost endpoints. The non-reference line segment may include a line segment connecting the first endpoint and the peak point and a shorter line segment among the line segments connecting the second endpoint and the peak point. The peak point may include a contour point located farthest from a line segment connecting the first endpoint and the second endpoint among the contour points.

[0027] In some implementations, the object recognition method may further include assigning, by the processor, a flag corresponding to a fourth shape to a specific layer when the shape of the distribution of the contour points is not recognized as one of the first shape, the second shape, and the third shape. The fourth shape may indicate that the shape of the distribution of the contour points is not recognized as one of the first shape, the second shape, and the third shape.

[0028] In some implementations, identifying by a processor whether an object is a dynamic object based on at least one of reference line segments, non-reference line segments, and distribution of contour points included in a specific layer among multiple layers representing the object, or any combination thereof, may include: identifying whether the object is a dynamic object based on at least one of reference line segments, non-reference line segments, and distribution of contour points, or any combination thereof, based on identifying that a size of a contour box including the contour points satisfies a first reference size range.

[0029] In some implementations, the object recognition method may further include, by the processor, identifying whether the object is a dynamic object based on at least one of a reference line segment, a distribution of contour points, and a length of the reference line segment, or any combination thereof, based on identifying that a size of a contour box including contour points does not satisfy a first reference size range but satisfies a second reference size range; identifying, by the processor, whether a first average value is less than a first average threshold, the first average value being an average value of at least one distance between the reference line segment and at least one contour point included in an area to which the reference line segment belongs, the area to which the reference line segment belongs being an area to which the reference line segment belongs in two areas separated by a straight line that bisects an angle between the reference line segment and the non-reference line segment and passes through a peak point in the contour points; and identifying, by the processor, a first average value as a variance value of at least one distance. whether a variance value is less than a first variance threshold; based on the object being a dynamic object, the first average value is less than the first average threshold, and the first variance value is less than the first variance threshold, identifying that the shape of the distribution of contour points is a second shape and assigning a flag corresponding to the second shape to a specific layer; the processor identifies whether the shape of the distribution of contour points is a third shape based on the object being a dynamic object and the first variance value is greater than or equal to the first variance threshold or the first average value is greater than or equal to the first average threshold, and when the shape of the distribution of contour points is identified as the third shape, assigning a flag corresponding to the third shape to a specific layer; and when the shape of the distribution of contour points is not identified as one of the first shape, the second shape and the third shape, the processor assigns a flag corresponding to a fourth shape to a specific layer.

[0030] In some implementations, the first shape may represent a shape including a first L-shaped distribution identified by all or part of the contour points. The second shape may represent a shape including an I-shaped distribution identified by all or part of the contour points. The third shape may represent a shape including a second L-shaped distribution identified by all or part of the contour points, the length of a reference line segment in the second L-shaped distribution being shorter than the length of a reference line segment in the first L-shaped distribution corresponding to the first shape. The fourth shape may indicate that the shape of the distribution of contour points is not identified as one of the first shape, the second shape, and the third shape.

[0031] In some implementations, identifying by the processor whether the shape of the distribution of contour points is a third shape based on identifying that the object is a dynamic object and the length of the reference line segment is less than or equal to the reference length, and assigning a flag corresponding to the third shape to a specific layer when the shape of the distribution of contour points is identified as the third shape may include: identifying by the processor, based on the object being a dynamic object and the length of the reference line segment is less than or equal to the reference length, that a ratio of a length of a line segment in one direction of a clustering box identified based on multiple layers to a length of a line segment in the one direction of a contour box containing contour points is less than a reference ratio, identifying by the processor that an angle between the reference line segment and a non-reference line segment satisfies a second angle range, identifying by the processor that the shape of the distribution of contour points is the third shape based on identifying that the ratio is less than the reference ratio and the angle between the reference line segment and the non-reference line segment satisfies the second angle range, and assigning by the processor a flag corresponding to the third shape to a specific layer.

[0032] In some implementations, identifying by the processor whether the shape of the distribution of contour points is a second shape based on identifying that the object is a dynamic object, the length of the reference line segment is greater than the reference length, and the angle between the reference line segment and the non-reference line segment does not satisfy the first angle range, and assigning a flag corresponding to the second shape to a specific layer when the shape of the distribution of contour points is identified as the second shape may include: identifying by the processor whether a first average value is less than or equal to a second average threshold, the first average value being an average value of at least one distance between the reference line segment and at least one contour point included in an area to which the reference line segment belongs, the area to which the reference line segment belongs being an area to which the reference line segment belongs among two areas separated by a straight line that bisects the angle between the reference line segment and the non-reference line segment and passes through a peak point in the contour points, identifying by the processor whether the angle between the reference line segment and the non-reference line segment satisfies a third angle range, and identifying by the processor that the shape of the distribution of contour points is a third shape and assigning a flag corresponding to the third shape to a specific layer based on the first average value being less than or equal to the second average threshold and the angle between the reference line segment and the non-reference line segment satisfying the third angle range.

[0033] In some implementations, the processor identifies whether the shape of the distribution of contour points is a third shape based on identifying that the object is a dynamic object and the length of the reference line segment is less than or equal to the reference length, and assigning a flag corresponding to the third shape to a specific layer when the shape of the distribution of contour points is identified as the third shape may include: based on identifying that the object is a dynamic object, the first average value is greater than or equal to the first average threshold, or the first variance value is greater than or equal to the first variance threshold, identifying by the processor whether there is at least one contour point in an area between a straight line passing through an endpoint included in the non-reference line segment and perpendicular to the non-reference line segment and a straight line passing through a peak point and perpendicular to the non-reference line segment, identifying by the processor whether the first average value is less than a third average threshold, identifying by the processor whether the first variance value is less than a second variance threshold, and identifying by the processor whether the second average value is less than a fourth average threshold, the second average value being an average of distances between the non-reference line segment and at least one contour point included in an area to which the non-reference line segment belongs, the area to which the non-reference line segment belongs being determined by assigning the reference line segment to the reference line segment. the area to which the non-reference line segment belongs in two areas divided by a straight line passing through a peak point in the contour points and bisected by an angle between the line segment and the non-reference line segment, the processor identifying whether a second variance value is less than a third variance threshold, the second variance value being a variance of at least one distance between the non-reference line segment and at least one contour point in the area to which an endpoint included in the non-reference line segment belongs, the processor identifying whether the angle between the reference line segment and the non-reference line segment satisfies a second angle range, the processor identifying that a shape of distribution of the contour points is a third shape based on identifying that at least one contour point exists in an area between a straight line passing through an endpoint included in the non-reference line segment and perpendicular to the non-reference line segment and a straight line passing through a peak point and perpendicular to the non-reference line segment, the first average value is less than the third average threshold, the second average value is less than the fourth average threshold, the first variance value is less than the second variance threshold, the second variance value is less than the third variance threshold, and the angle between the reference line segment and the non-reference line segment satisfies the second angle range, and the processor assigning a flag corresponding to the third shape to a specific layer.

[0034] In some implementations, the length of the longer line segment among the line segments in the first direction and the line segments in the second direction different from the first direction among the line segments of the outline frame that meet the first reference size range may be greater than the first length value and less than the second length value that is greater than the first length value. The shorter line segment among the line segments in the first direction and the line segments in the second direction may be greater than or equal to the third length value. Among the line segments of the outline frame that meet the second reference size range, the length of the longer line segment among the line segments in the third direction and the line segments in the fourth direction different from the third direction may be greater than the first length value. The shorter line segment among the line segments in the third direction and the line segments in the fourth direction may be less than the fourth length value that is greater than the third length value.

[0035] In some implementations, the object recognition method may further include identifying, by the processor, a layer among multiple layers that is assigned a marker corresponding to a fourth shape, identifying, by the processor, the number of layers among multiple layers that are assigned a marker corresponding to the first shape, identifying, by the processor, whether the number of layers among multiple layers that are assigned a marker corresponding to the first shape is less than the number of layers that are assigned a marker corresponding to the second shape, and assigning, by the processor, a marker corresponding to the second shape to the multiple layers based on identifying that the layer assigned a marker corresponding to the fourth shape is not identified, the layer assigned a marker corresponding to the first shape is identified, and the number of layers among multiple layers that are assigned a marker corresponding to the first shape is less than the number of layers that are assigned a marker corresponding to the second shape.

[0036] In some implementations, the object recognition method may further include the processor identifying, among multiple layers, a layer assigned a marker corresponding to a fourth shape, the processor identifying, among multiple layers, a layer assigned a marker corresponding to a first shape, the processor identifying, among multiple layers, a layer assigned a marker corresponding to a second shape, and the processor assigning, based on the fact that the layer assigned a marker corresponding to the fourth shape is not identified, the layer assigned a marker corresponding to the first shape is not identified, and the layer assigned a marker corresponding to the second shape is identified, the processor assigning a marker corresponding to the second shape to the multiple layers.

[0037] In some implementations, the object recognition method may further include the processor identifying, among multiple layers, a layer assigned a marker corresponding to a fourth shape, the processor identifying, among multiple layers, a layer assigned a marker corresponding to a first shape, the processor identifying, among multiple layers, a layer assigned a marker corresponding to a second shape, the processor identifying, among multiple layers, a layer assigned a marker corresponding to a third shape, and the processor assigning, based on the fact that the layer assigned a marker corresponding to the fourth shape is not identified, the layer assigned a marker corresponding to the first shape is not identified, the layer assigned a marker corresponding to the second shape is not identified, and the layer assigned a marker corresponding to the third shape is identified. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 is a block diagram illustrating an embodiment of an object recognition apparatus.

[0039] Figure 2 This is an embodiment showing a case where a third shape is recognized in an object recognition device or an object recognition method.

[0040] Figure 3 is a flowchart illustrating an exemplary operation of an object recognition device for recognizing shapes representing a plurality of layers in an object recognition device or an object recognition method.

[0041] Figure 4The present invention is an embodiment showing reference line segments, non-reference line segments, and peak points for identifying distribution of contour points in an object recognition apparatus or an object recognition method.

[0042] Figure 5 The present invention is a flowchart illustrating an exemplary operation of an object recognition apparatus for identifying whether the distribution of contour points corresponds to a shape other than a first shape when the distribution of contour points does not correspond to the first shape in an object recognition apparatus or an object recognition method.

[0043] Figure 6 is a flowchart illustrating an exemplary operation of an object recognition apparatus for identifying whether a distribution of contour points corresponds to a shape other than a second shape when the distribution does not correspond to the second shape in an object recognition apparatus or an object recognition method.

[0044] Figure 7 is a flowchart illustrating an exemplary operation of an object recognition device for assigning a flag to a layer in an object recognition device or an object recognition method.

[0045] Figure 8 is a flowchart illustrating an exemplary operation of an object recognition device for assigning flags to a plurality of layers in an object recognition device or an object recognition method.

[0046] Fig. 9 The present invention is an embodiment showing the distribution of contour points used for calculating reliability in an object recognition device or an object recognition method.

[0047] Fig.10 The present invention is an embodiment of a computing system related to an object recognition apparatus or an object recognition method. DETAILED DESCRIPTION

[0048] Figure 1 is a block diagram showing an object recognition apparatus.

[0049] See also Figure 1 , the object recognition device 101 may be implemented inside the vehicle. In this case, the object recognition device 101 may be integrally formed with an internal unit of the vehicle, or may be implemented as a separate device and connected to the unit of the vehicle through a separate connection device.

[0050] See also Figure 1 , the object recognition device 101 may include a LIDAR 103 (eg, a LIDAR sensor) and a processor 105 .

[0051] In some implementations, the processor 105 may obtain location information of points of objects around the vehicle including the object recognition device 101 through the LIDAR 103. The LIDAR 103 may acquire contour points included in multiple layers representing the object. Each contour point may indicate the location of a point on the object.

[0052] In some implementations, the contour points may be obtained by the LIDAR 103. For example, the contour points may be identified in each of the layers formed based on the z-axis among the x-axis, the y-axis, and the z-axis.

[0053] For example, the contour points may be obtained based on the representative points included in the point cloud in each layer formed on the z-axis among the x-axis, the y-axis, and the z-axis. For example, the representative points may include all or part of the points located outside among the plurality of points included in the point cloud. For example, the point cloud may be obtained by performing clustering based on each of the plurality of points acquired by the LIDAR 103 identified within a specified distance.

[0054] In some implementations, the processor of the object recognition device may obtain, through the layers, contour points representing the object and existing on a plane parallel to the specific plane. For example, the specific plane may include a plane parallel to the ground. For example, the specific plane may include a plane perpendicular to the ground and including the moving direction of the vehicle.

[0055] In some implementations, the processor 105 of the object recognition apparatus 101 may determine a shape representing an object shape based on the flags assigned to the multiple layers. The processor 105 may assign a flag to each of the multiple layers according to the distribution of contour points included in each layer.

[0056] For example, all or part of the contour points corresponding to a first shape (e.g., L-shape) included in a specific layer may represent an L-shaped distribution (or a first L-shaped distribution). For example, all or part of the contour points corresponding to a second shape (e.g., I-shape) included in a specific layer may represent an I-shaped distribution. For example, all or part of the contour points corresponding to a third shape (e.g., sL-shape) included in a specific layer may represent a second L-shaped distribution, the length of a reference line segment in the second L-shaped distribution being shorter than the length of a reference line segment in the L-shaped distribution corresponding to the first shape. For example, a fourth shape (e.g., a broken shape) may indicate that the distribution of contour points representing an object and included in a specific layer is not recognized as one of the first shape, the second shape, and the third shape. A mark corresponding to the fourth shape may be referred to as a broken mark.

[0057] For example, the contour points included in a specific layer corresponding to one of the first shape, the second shape, and the third shape include contour points corresponding to the first line segment and contour points corresponding to the second line segment. The second line segment may be connected to an endpoint of the first line segment. In the specific layer corresponding to the first shape, the angle between the first line segment and the second line segment may fall within a specified angle range. The length of the first line segment may be greater than the specified length. In the specific layer corresponding to the second shape, the angle between the first line segment and the second line segment may be outside the specified angle range. In the specific layer corresponding to the third shape, the length of the first line segment may be less than or equal to the specified length.

[0058] For example, the distribution of the contour points of the first shape may be represented by the side and rear of an object such as a vehicle. The distribution of the contour points of the second shape may be represented by the side or back of an object which is a vehicle. The distribution of the contour points of the third shape may appear in a part of the side or rear of an object which is a vehicle, or appear based on an object which is a vehicle tilted relative to the vehicle including the object recognition device 101.

[0059] Figure 2 An example is shown in which a third shape is recognized in an object recognition apparatus or an object recognition method in some implementations of the present disclosure.

[0060] See also Figure 2 , the third shape may be recognized in the first scenario 201, the second scenario 211, and the third scenario 221. In the first stage 203 of the first scenario 201, the first stage 213 of the second scenario 211, and / or the first stage 223 of the third scenario 221, the processor of the object recognition device may recognize an area including a portion of the contour points included in the specific layer and an area including another portion of the contour points included in the specific layer based on the distribution of the recognized contour points. In the second stage 205 of the first scenario 201, the second stage 215 of the second scenario 211, and / or the second stage 225 of the third scenario 221, an area of ​​the object corresponding to the contour points may be recognized by connecting the area including the portion of the contour points and the area including the other portion of the contour points.

[0061] In some implementations, in the first scenario 201, the object recognition device may recognize a front vehicle turning along a curved road in front of a vehicle including the object recognition device as an object. The processor of the object recognition device may acquire contour points representing the object (which is the front vehicle in the first scenario 201) and included in a specific layer.

[0062] In the second scenario 211, the object recognition device may recognize a vehicle traveling in front of or in a lateral direction of the vehicle including the object recognition device as an object. The processor of the object recognition device may acquire contour points representing the object (which is the front vehicle in the second scenario 211) and included in a specific layer.

[0063] In the third case 221, the object recognition device may recognize a vehicle cutting in from the front or side in the moving direction of the vehicle including the object recognition device as an object. The processor of the object recognition device may acquire contour points representing the object (which is the cutting in vehicle in the third case 221).

[0064] In the first scenario 201, the second scenario 211, or the third scenario 221, the processor of the object recognition device may recognize the shape of distribution of a portion of the contour points (e.g., a first shape, a second shape, a third shape, a fourth shape) based on the distribution of the contour points. The processor of the object recognition device may recognize the region including the portion of the contour points based on the shape of the portion of the contour points. Contour points representing an object and included in a specific layer may be divided into different regions according to the angle at which the object is tilted relative to the vehicle or the direction in which the object is traveling.

[0065] For example, in the first stage 203 of the first scenario 201, the first stage 213 of the second scenario 211, and / or the first stage 223 of the third scenario 221, the processor of the object recognition device may recognize an area representing a portion of the contour points of the object and an area representing another portion of the contour points of the object. The area recognized based on the portion of the contour points representing the object and the area recognized based on the other portion of the contour points representing the object may be separated.

[0066] In the second stage 205 of the first scenario 201, the second stage 215 of the second scenario 211 and / or the second stage 225 of the third scenario 221, the processor of the object identification device can identify the area corresponding to the object by connecting the area of ​​the part of the contour points representing the object and contained in the specific layer to the area of ​​the other part of the contour points representing the object and contained in the specific layer.

[0067] It may be easier to connect the multiple regions when there is a region including contour points distributed in the third shape among the multiple regions than to connect the multiple regions when there is no region including contour points distributed in the third shape among the multiple regions. In addition, when there is a region including contour points distributed in the fourth shape, the possibility that the object represented by the contour points in the multiple regions is a dynamic object may be greater than the possibility that the object represented by the contour points in the multiple regions is a dynamic object when there is no region including contour points recognized as the fourth shape.

[0068] In some implementations, a processor of the object recognition device may identify whether a distribution of contour points contained in a specific layer corresponds to a third shape before a flag corresponding to a fourth shape is assigned to the contour points, even though the distribution of the contour points contained in the specific layer is not identified as one of the first shape and the second shape.

[0069] Figure 3 A flowchart illustrating the operation of an object recognition device for recognizing shapes representing multiple layers in an object in an object recognition device or an object recognition method in some implementations of the present disclosure.

[0070] In the following, it is assumed Figure 1 The object recognition device 101 performs Figure 3 In addition, Figure 3 In the description of FIG. 1 , an operation described as being performed by the object recognition device 101 or an operation described as being performed by the processor 105 may be understood to be controlled by the processor 105 of the object recognition device 101 .

[0071] See also Figure 3 In a first operation 301, in some implementations, a processor of the object recognition device may assign a marker corresponding to one of a first shape, a second shape, a third shape, and a fourth shape to each of a plurality of layers.

[0072] In some implementations, the processor of the object recognition device may recognize the distribution of contour points in a specific layer included in the plurality of layers as one of a first shape, a second shape, a third shape, and a fourth shape. The processor of the object recognition device may assign a flag corresponding to the recognized shape to the layer including the contour points based on the shape of the distribution of contour points included in the specific layer. Figure 4 , Figure 5 , Figure 6 and / or Figure 7 Describes the method of assigning flags to each layer.

[0073] In a second operation 303, in some implementations, the processor of the object recognition device may assign a flag to the plurality of layers based on the flags assigned to the plurality of layers. The flags assigned to the plurality of layers may correspond to shapes representing the plurality of layers.

[0074] In some implementations, a processor of an object recognition device may determine a shape of an object using markers assigned to multiple layers. The shape of the object may be one of a first shape, a second shape, a third shape, and a fourth shape. The processor of the object recognition device may assign a marker to multiple layers representing the object based on the shape of the object. Since there may be multiple layers including contour points representing the object, identifying a shape representing the object and assigning a marker to the multiple layers based on the markers assigned to the multiple layers may be important for object recognition and association. In the following, reference will be made to Figure 8 A method of assigning a flag to a plurality of layers according to the flags assigned to the plurality of layers is described.

[0075] Figure 4 Embodiments of reference line segments, non-reference line segments, and peak points for identifying distribution of contour points in an object recognition apparatus or an object recognition method in some implementations of the present disclosure are shown.

[0076] See also Figure 4 , the first endpoint 401 and the second endpoint 403 may correspond to the outermost contour point. The peak point 405 may include a contour point located at the farthest position from the line segment connecting the first endpoint 401 and the second endpoint 403. The first line segment 407 may include a line segment connecting the first endpoint 401 and the peak point 405. The second line segment 409 may include a line segment connecting the second endpoint 403 and the peak point 405. The reference line segment may include a longer line segment of the first line segment 407 and the second line segment 409. The non-reference line segment may include a shorter line segment of the first line segment 407 and the second line segment 409.

[0077] In some implementations, the processor of the object recognition device may obtain points representing the object through LIDAR. The processor of the object recognition device may obtain contour points including some or all of the points located outside among the points representing the object. The processor of the object recognition device may identify whether the distribution of the contour points corresponds to one of the first shape, the second shape, the third shape, and the fourth shape.

[0078] In some implementations, in order to identify the distribution of contour points, the processor of the object recognition device may consider the following factors, such as the size of the contour box, whether the object is a dynamic object, the length of the reference line segment, the angle between the reference line segment and the non-reference line segment, the average distance and variance value of the reference line segment, the average distance and variance value of the non-reference line segment, and whether there are contour points in a specific area.

[0079] In some implementations, a contour box may be generated (or formed) based on the contour points. For example, the contour box may correspond to an external object. For example, the contour box may include a virtual box to which information related to the external object is assigned. For example, the information related to the external object may include at least one of the type of the external object, the speed of the external object, the moving direction of the external object, and the position of the external object, or any combination thereof. The contour box may represent an area including a range of some or all contour points included in a specific layer.

[0080] In some implementations, the processor of the object recognition device may recognize whether the object is a dynamic object based on recognizing that a size of a contour box including contour points included in a specific layer among multiple layers representing the object satisfies a first reference size range.

[0081] In some implementations, when the size of the contour box satisfies the first reference size range, the processor of the object recognition device can identify that the object is a dynamic object based on the recognition that: there is at least one contour point in the area between a straight line passing through an endpoint included in a reference line segment and perpendicular to the reference line segment and a straight line passing through a peak point 405 and perpendicular to the reference line segment, there is at least one contour point in the area between a straight line passing through an endpoint included in a non-reference line segment and perpendicular to the non-reference line segment and a straight line passing through a peak point 405 and perpendicular to the non-reference line segment, the first average value is less than a third average threshold, the first variance value is less than a second variance threshold, the second average value is less than a fourth average threshold, or the second variance value is less than a third variance threshold.

[0082] For example, when the size of the outline frame satisfies the first reference size range, in order to identify whether the object is a dynamic object, the processor of the object recognition device may identify whether there is at least one outline point in the area between a straight line passing through an endpoint (e.g., first endpoint 401 or second endpoint 403) included in a reference line segment (e.g., first line segment 407 or second line segment 409) and perpendicular to the reference line segment and a straight line passing through peak point 405 and perpendicular to the reference line segment. For example, in order to identify whether the object is a dynamic object, the processor of the object recognition device may identify whether there is at least one outline point in the area between a straight line passing through an endpoint included in a non-reference line segment and perpendicular to the non-reference line segment and a straight line passing through peak point 405 and perpendicular to the non-reference line segment.

[0083] For example, when the size of the outline frame satisfies the first reference size range, in order to identify whether the object is a dynamic object, the processor of the object recognition device may identify whether a first average value is less than a third average threshold value (e.g., about 0.15), the first average value being an average value of at least one distance between the reference line segment and part or all of at least one outline point included in the region to which the reference line segment belongs, the region to which the reference line segment belongs being the region to which the reference line segment belongs among two regions separated by a straight line that bisects the angle between the reference line segment and the non-reference line segment and passes through the peak point 405. The processor of the object recognition device may identify whether a first variance value is less than a second variance threshold value (e.g., about 0.55), the first variance value being a variance value of at least one distance between the reference line segment and part or all of at least one outline point included in the region to which the reference line segment belongs.

[0084] For example, when the size of the outline frame satisfies the first reference size range, in order to identify whether the object is a dynamic object, the processor of the object recognition device may identify whether a second average value is less than a fourth average threshold value (e.g., about 0.5), the second average value being an average value of at least one distance between the non-reference line segment and part or all of at least one contour point included in the region to which the reference line segment belongs, the region to which the reference line segment belongs being the region to which the reference line segment belongs among two regions separated by a straight line that bisects the angle between the reference line segment and the non-reference line segment and passes through the peak point 405. The processor of the object recognition device may identify whether a second variance value is less than a third variance threshold value (e.g., about 3.0), the second variance value being a variance value of at least one distance between the non-reference line segment and part or all of at least one contour point included in the region to which the non-reference line segment belongs.

[0085] In some implementations, a processor of an object recognition device may recognize a shape of a distribution of contour points contained in a specific layer as a first shape based on recognizing that a size of a contour box satisfies a first reference size range, that the object is a dynamic object, that a length of a reference line segment is greater than a reference length, or that an angle between a reference line segment and a non-reference line segment satisfies a specified first angle range.

[0086] When the length of a reference line segment is less than or equal to a reference length, or the angle between the reference line segment and the non-reference line segment does not satisfy a specified first angle range, an existing object recognition device may recognize the shape of the distribution of contour points contained in a specific layer as a fourth shape even though the size of the contour box satisfies the first reference size range and the object is a dynamic object.

[0087] In some implementations, when the size of the contour box satisfies the first reference size range and the object is a dynamic object, even though the length of the reference line segment is less than or equal to the reference length, the processor of the object recognition device may also identify whether the shape of the distribution of contour points contained in a specific layer is the third shape, instead of immediately identifying the shape of the distribution of the contour points as the fourth shape.

[0088] In some implementations, when the size of the contour box satisfies the first reference size range, the object is a dynamic object, and the length of the reference line segment is greater than the reference length, even though the angle between the reference line segment and the non-reference line segment does not satisfy the specified first angle range, the processor of the object recognition device can also identify whether the shape of the distribution of contour points contained in a specific layer is the second shape, instead of immediately identifying the shape of the distribution of the contour points as the fourth shape.

[0089] In some implementations, the processor of the object recognition device may recognize whether the object is a dynamic object based on recognizing that the size of the outline box representing the object satisfies a second reference size range.

[0090] For example, when the size of the outline frame satisfies the second reference size range, the processor of the object recognition device may recognize that there is at least one outline point in the area between the straight line passing through the endpoints included in the reference line segment and perpendicular to the reference line segment and the straight line passing through the peak point 405 and perpendicular to the reference line segment, to recognize whether the object is a dynamic object. It should be noted that when the size of the outline frame satisfies the second reference size range, the distribution of the outline points may correspond to the second shape, and therefore, the distribution of the non-reference line segment may not be recognized.

[0091] For example, when the size of the outline box satisfies the second reference size range, the processor of the object recognition device may identify whether the length of the reference line segment is greater than a specified length (eg, approximately 4.5 m) to identify whether the object is a dynamic object.

[0092] For example, when the size of the outline box satisfies the second reference size range, the processor of the object recognition device can identify whether the first average value is less than a first average threshold value (e.g., approximately 0.05) to identify whether the object is a dynamic object. The processor of the object recognition device can identify whether the first variance value is less than a first variance threshold value (e.g., approximately 0.2).

[0093] In some implementations, a processor of the object recognition device may identify the shape of the distribution of contour points contained in a specific layer as a second shape based on recognizing that the size of the contour box satisfies a second reference size range, the object is a dynamic object, the first average value is less than a first average threshold, or the first variance value is less than a first variance threshold.

[0094] When the first average value is greater than or equal to the first average threshold or the first variance value is greater than or equal to the first variance threshold, the existing object recognition device may recognize the shape of the distribution of contour points contained in a specific layer as a fourth shape even though the size of the contour box satisfies the second reference size range and the object is a dynamic object.

[0095] In some implementations, when the object is a dynamic object, even though the first average value is greater than or equal to the first average threshold, or the first variance value is greater than or equal to the first variance threshold, the processor of the object recognition device may also identify whether the shape of the distribution of contour points contained in a specific layer is the third shape, instead of immediately identifying the shape of the distribution of contour points as the fourth shape.

[0096] Figure 5 A flowchart showing the operation of an object recognition device for identifying whether the distribution of contour points corresponds to a shape other than the first shape when the distribution of contour points does not correspond to the first shape in an object recognition device or an object recognition method in some implementations of the present disclosure.

[0097] In the following, it is assumed Figure 1 The object recognition device 101 performs Figure 5 In addition, Figure 5 In the description of , an operation described as being performed by the object recognition device 101 or an operation described as being performed by the processor 105 may be understood to be controlled by the processor 105 of the object recognition device 101 .

[0098] See also Figure 5 In the first operation 501, in some implementations, the processor of the object recognition device may recognize that the size of the outline frame satisfies the first reference size range. This is because the shape may be roughly determined based on the size of the outline frame. The probability that the object represented by the outline points included in the outline frame that satisfies the first reference size range has a first shape may be greater than the probability that the object represented by the outline points included in the outline frame has a second shape.

[0099] For example, the length of the line segment in the first direction and the longer line segment among the line segments in the second direction different from the first direction of the contour box satisfying the first reference size range may be greater than the first length value (e.g., about 1 m) and less than the second length value (e.g., about 8 m) greater than the first length value. The length of the line segment in the first direction and the shorter line segment among the line segments in the second direction may be greater than or equal to the third length value (e.g., about 0.8 m). The contour box may represent an area including the range of some or all contour points included in a specific layer.

[0100] In a second operation 503, in some implementations, the processor of the object recognition device may recognize that the object is a dynamic object.

[0101] In some implementations, the processor of the object recognition device may recognize that the object is a dynamic object based on recognizing that at least one contour point exists in the area between a straight line passing through an endpoint included in a reference line segment and perpendicular to the reference line segment and a straight line passing through a peak point and perpendicular to the reference line segment, at least one contour point exists in the area between a straight line passing through an endpoint included in a non-reference line segment and perpendicular to the non-reference line segment and a straight line passing through a peak point and perpendicular to the non-reference line segment, a first average value is less than a third average threshold, a first variance value is less than a second variance threshold, a second average value is less than a fourth average threshold, or a second variance value is less than a third variance threshold. Figure 4 The operation of the object recognition device for recognizing whether an object is a dynamic object has been described, so a redundant description will be omitted below.

[0102] In the third operation 505, in some implementations, the processor of the object recognition device may identify whether the length of the reference line segment is greater than a reference length (e.g., about 2 m). When the length of the reference line segment is greater than the reference length, the processor of the object recognition device may perform a fourth operation 507. When the length of the reference line segment is less than or equal to the reference length, the processor of the object recognition device may perform a fifth operation 509. This is because, when the length of the reference line segment is not greater than the reference length, it is difficult to recognize the shape of the contour point as the first shape.

[0103] In the fifth operation 509, in some implementations, the processor of the object recognition device may identify whether the shape of the distribution of the contour points is the third shape. When the shape of the distribution of the contour points is the third shape, the processor of the object recognition device may perform a sixth operation 511. When the shape of the distribution of the contour points is not the third shape, the processor of the object recognition device may perform a seventh operation 513.

[0104] In some implementations, unlike the existing object recognition apparatus, when the length of the reference line segment is less than or equal to the reference length, the processor of the object recognition apparatus may instead determine whether the shape of the distribution is recognized as the third shape, instead of immediately recognizing the shape of the distribution as the fourth shape. In order to identify whether the shape of the distribution of contour points in a specific layer is the third shape, the processor of the object recognition apparatus may execute the following Figure 7 The eighth operation 721 and the ninth operation 723 in .

[0105] In the sixth operation 511, in some implementations, the processor of the object recognition device may assign a flag corresponding to the third shape to a specific layer. The processor of the object recognition device may indicate through the flag that the shape of the distribution of the contour points is recognized as the third shape.

[0106] In the seventh operation 513, in some implementations, the processor of the object recognition device may assign a flag corresponding to the fourth shape to a specific layer. The processor of the object recognition device may indicate through the flag that the shape of the distribution of the contour points is recognized as the fourth shape.

[0107] In the fourth operation 507, in some implementations, the processor of the object recognition device may identify whether the angle between the reference line segment and the non-reference line segment satisfies the first angle range. When the angle between the reference line segment and the non-reference line segment satisfies the first angle range, the processor of the object recognition device may perform an eighth operation 519. When the angle between the reference line segment and the non-reference line segment does not satisfy the first angle range, the processor of the object recognition device may perform a ninth operation 515. This is because when the angle between the reference line segment and the non-reference line segment satisfies the first angle range (e.g., greater than about 80° and less than about 120°), the shape of the contour point distribution is recognized as an L-shape as the first shape.

[0108] In an eighth operation 519, in some implementations, the processor of the object recognition device may assign a flag corresponding to the first shape. The processor of the object recognition device may indicate through the flag that the shape of the distribution of the contour points is recognized as the first shape.

[0109] In the ninth operation 515, in some implementations, the processor of the object recognition device may identify whether the shape of the distribution of the contour points is the second shape. When the shape of the distribution of the contour points is the second shape, the processor of the object recognition device may perform the tenth operation 517. When the shape of the distribution of the contour points is not the second shape, the processor of the object recognition device may perform the seventh operation 513.

[0110] In some implementations, unlike the existing object recognition apparatus, although the angle between the reference line segment and the non-reference line segment does not satisfy the first angle range, the processor of the object recognition apparatus may determine whether the shape of the distribution is recognized as the second shape, instead of immediately recognizing the shape of the distribution as the fourth shape. In order to identify whether the shape of the distribution of contour points in a specific layer is the second shape, the processor of the object recognition apparatus may execute the following Figure 7 The twelfth operation 715 and the thirteenth operation 717 are performed.

[0111] In a tenth operation 517, in some implementations, the processor of the object recognition device may assign a flag corresponding to the second shape to a specific layer. The processor of the object recognition device may indicate through the flag that the shape of the distribution of the contour points is recognized as the second shape.

[0112] Figure 6A flowchart showing the operation of an object recognition device for identifying whether the distribution of contour points corresponds to a shape other than the second shape when the distribution does not correspond to the second shape in an object recognition device or an object recognition method in some implementations of the present disclosure.

[0113] In the following, it is assumed Figure 1 The object recognition device 101 performs Figure 6 In addition, Figure 6 In the description of FIG. 1 , an operation described as being performed by the object recognition device 101 or an operation described as being performed by the processor 105 may be understood to be controlled by the processor 105 of the object recognition device 101 .

[0114] See also Figure 6 In the first operation 601, in some implementations, the processor of the object recognition device may recognize that the size of the outline box meets the second reference size range. Among the line segments of the outline box that meet the second reference size range, the length of the longer line segment among the line segments in the third direction and the line segments in the fourth direction different from the third direction may be greater than the first length value (for example, about 1m). The length of the shorter line segment among the line segments in the third direction and the line segments in the fourth direction may be less than the fourth length value (for example, about 1m). The fourth length value in the second reference size range (for example, about 1m) may be greater than the third length value in the first reference size range (for example, about 0.8m).

[0115] In some implementations, when the size of the outline box satisfies the first reference size range and also satisfies the second reference size range, the processor of the object recognition device may execute Figure 5 An operation is performed to identify the shape of the distribution of contour points in a particular layer.

[0116] In the second operation 603, in some implementations, the processor of the object recognition device may recognize that the object is a dynamic object. When the size of the outline box meets the second reference size range, the processor of the object recognition device may recognize that the object is a dynamic object based on recognizing that there is at least one outline point in the area between a straight line passing through an endpoint included in the reference line segment and perpendicular to the reference line segment and a straight line passing through a peak point and perpendicular to the reference line segment, or recognizing that the length of the reference line segment is greater than a specified length. Figure 4 The operation of the object recognition device for recognizing whether an object is a dynamic object has been described, so a redundant description will be omitted below.

[0117] In the third operation 605, in some implementations, the processor of the object recognition device may identify whether the first average value of the reference line segment is less than the first average threshold and whether the first variance value of the reference line segment is less than the first variance threshold. When the first average value of the reference line segment is less than the first average threshold and the first variance value of the reference line segment is less than the first variance threshold, the processor of the object recognition device may perform a fourth operation 607. When the first average value of the reference line segment is greater than or equal to the first average threshold or the first variance value of the reference line segment is greater than or equal to the first variance threshold, the processor of the object recognition device may perform a fifth operation 609. In some implementations, the processor of the object recognition device may identify how closely the contour points are distributed to the reference line segment based on the first average value and the first variance value.

[0118] In the fourth operation 607, in some implementations, the processor of the object recognition device may assign a flag corresponding to the second shape to the specific layer. When identifying whether the shape of the distribution of contour points in the specific layer is the second shape, the processor of the object recognition device may not consider the shape of the distribution of points corresponding to the non-reference line segment.

[0119] In the fifth operation 609, in some implementations, the processor of the object recognition device may identify whether the shape of the distribution of the contour points is the third shape. When the shape of the distribution of the contour points is the third shape, the processor of the object recognition device may perform a sixth operation 611. When the shape of the distribution of the contour points is not the third shape, the processor of the object recognition device may perform a seventh operation 613.

[0120] In some implementations, unlike the existing object recognition apparatus, when the first average value is greater than or equal to the first average threshold or the first variance value is greater than or equal to the first variance threshold, the processor of the object recognition apparatus may determine whether the shape of the distribution is recognized as the third shape, instead of immediately recognizing the shape of the distribution as the fourth shape. In order to identify whether the shape of the distribution of contour points in a specific layer is the third shape, the processor of the object recognition apparatus may execute the following Figure 7 An eighteenth operation 735 , a nineteenth operation 737 , and a twentieth operation 739 .

[0121] In the sixth operation 611, in some implementations, the processor of the object recognition device may assign a flag corresponding to the third shape to a specific layer. The processor of the object recognition device may indicate through the flag that the shape of the distribution of contour points included in the specific layer is recognized as the third shape.

[0122] In the seventh operation 613, in some implementations, the processor of the object recognition device may assign a flag corresponding to the fourth shape to the specific layer. The processor of the object recognition device may indicate through the flag that the shape of the distribution of contour points included in the specific layer is recognized as the fourth shape.

[0123] Figure 7 A flowchart showing the operation of an object recognition device for assigning a flag to a layer in an object recognition device or an object recognition method in some implementations of the present disclosure is shown.

[0124] In the following, it is assumed Figure 1 The object recognition device 101 performs Figure 7 In addition, Figure 7 In the description of FIG. 1 , an operation described as being performed by the object recognition device 101 or an operation described as being performed by the processor 105 may be understood to be controlled by the processor 105 of the object recognition device 101 .

[0125] See also Figure 7 In the first operation 701, the processor of the object recognition device may recognize, in some implementations, that the length of the line segment of the clustering frame in one direction is less than a specified length (e.g., about 13 m). The clustering frame may include some or all of the contour points included in the plurality of layers. In addition, in the case where the clustering frame is greater than the specified length, since the object may be a stationary object, it is almost unnecessary to recognize the shape of the distribution of the contour points in various ways.

[0126] In the second operation 703, in some implementations, the processor of the object recognition device may identify whether the size of the outline frame satisfies the first reference size range. When the size of the outline frame satisfies the first reference size range, the processor of the object recognition device may perform a third operation 705. When the size of the outline frame does not satisfy the first reference size range, the processor of the object recognition device may perform a fourth operation 727. This is because the shape may be roughly determined based on the size of the outline frame. The probability that the object represented by the outline points included in the outline frame that satisfies the first reference size range has a first shape may be greater than the probability that the object represented by the outline points included in the outline frame has a second shape. Therefore, the processor of the object recognition device may preferentially determine whether the outline frame that satisfies the first reference size range is the first shape.

[0127] In the third operation 705 , in some implementations, the processor of the object recognition device may recognize that at least one contour point exists in each of the region of the reference line segment and the region of the non-reference line segment.

[0128] For example, the area of ​​the reference line segment may include an area between a straight line passing through an endpoint included in the reference line segment and perpendicular to the reference line segment and a straight line passing through a peak point and perpendicular to the reference line segment. The area of ​​the non-reference line segment may include an area between a straight line passing through an endpoint included in the non-reference line segment and perpendicular to the non-reference line segment and a straight line passing through a peak point and perpendicular to the non-reference line segment.

[0129] In a fifth operation 707 , in some implementations, the processor of the object recognition device may recognize that the first mean condition and the first variance condition of the reference line segment and the second mean condition and the second variance condition of the non-reference line segment are satisfied.

[0130] For example, the first average condition may include a condition that a first average value is less than a third average threshold and a variance value of at least one distance for the reference line segment is less than a second variance threshold, the first average value being the average value of the at least one distance for the reference line segment.

[0131] For example, the second average condition may include a condition in which a second average value which is an average value of at least one distance to the non-reference line segment is less than a fourth average threshold and a variance value of at least one distance to the non-reference line segment is less than a fourth variance threshold.

[0132] In the sixth operation 709, in some implementations, the processor of the object recognition device may identify whether the length of the reference line segment is less than the reference length. When the length of the reference line segment is greater than or equal to the reference length, the processor of the object recognition device may perform a seventh operation 711. When the length of the reference line segment is less than the reference length (e.g., about 2 m), the processor of the object recognition device may perform an eighth operation 721.

[0133] In the eighth operation 721, in some implementations, the processor of the object recognition device may identify that a length condition of a line segment in one direction of the clustering frame and the contour frame is satisfied. The length condition may include identifying a condition that a ratio of a length of a line segment in one direction of the clustering frame to a length of a line segment in one direction of the contour frame is less than a reference ratio (e.g., about 0.65).

[0134] In the ninth operation 723, in some implementations, the processor of the object recognition device may recognize that the angle between the reference line segment and the non-reference line segment satisfies the second angle range. This is because the probability that the shape of the distribution of the contour points corresponds to the third shape when the angle satisfies the second angle range (e.g., greater than about 80° and less than about 105°) is greater than the probability that the shape of the distribution of the contour points corresponds to the third shape when the angle does not satisfy the second angle range.

[0135] In a tenth operation 725, in some implementations, the processor of the object recognition device may assign a marker corresponding to the third shape.

[0136] In the seventh operation 711, in some implementations, the processor of the object recognition device may identify whether the angle between the reference line segment and the non-reference line segment satisfies the first angle range. When the angle between the reference line segment and the non-reference line segment satisfies the first angle range, in some implementations, the processor of the object recognition device may perform the eleventh operation 713. When the angle between the reference line segment and the non-reference line segment does not satisfy the first angle range, in some implementations, the processor of the object recognition device may perform the twelfth operation 715. This is because when the angle between the reference line segment and the non-reference line segment satisfies the first angle range (e.g., greater than about 80° and less than about 120°), the shape of the distribution of the contour points is recognized as an L shape. It should be noted that in some implementations, even when the angle between the reference line segment and the non-reference line segment does not satisfy the first angle range, the processor of the object recognition device may identify whether the shape of the distribution of points in a specific layer corresponds to the second shape.

[0137] In the twelfth operation 715, in some implementations, the processor of the object recognition device may recognize that a third average condition of the reference line segment is satisfied. The third average condition may include a condition that the first average value of the reference line segment is less than or equal to a reference value (eg, about 0.1).

[0138] In the thirteenth operation 717, in some implementations, the processor of the object recognition device may recognize that the angle between the reference line segment and the non-reference line segment satisfies the third angle range. When the angle between the reference line segment and the non-reference line segment satisfies the third angle range (for example, greater than 160°), the processor of the object recognition device may recognize that the points included in the specific layer are distributed in an I-shape.

[0139] In a fourteenth operation 719, in some implementations, the processor of the object recognition device may assign a marker corresponding to the second shape.

[0140] In a fourth operation 727 , in some implementations, the processor of the object recognition device may recognize that the size of the outline box satisfies a second reference size range.

[0141] In some implementations, among the line segments of the outline frame that meet the second reference size range, the length of the longer line segment among the line segments in the third direction and the line segments in the fourth direction different from the third direction may be greater than the first length value (e.g., about 1 m). The length of the shorter line segment among the line segments in the third direction and the line segments in the fourth direction may be less than the fourth length value (e.g., about 1 m). The fourth length value in the second reference size range (e.g., about 1 m) may be greater than the third length value in the first reference size range (e.g., about 0.8 m).

[0142] In a fifteenth operation 729, in some implementations, the processor of the object recognition device may recognize that at least one contour point exists in the region of the reference line segment and satisfies the length condition of the reference line segment. The region of the reference line segment may be the same as the region of the reference line segment in operation 705. The length condition of the reference line segment may include a condition that the reference line segment is greater than a specified length (e.g., about 4.5 m).

[0143] In the sixteenth operation 731, in some implementations, the processor of the object recognition device may identify whether the fourth average condition and the third difference condition of the reference line segment are satisfied. When the fourth average condition and the third difference condition of the reference line segment are satisfied, the processor of the object recognition device may perform a seventeenth operation 733. When the fourth average condition or the third difference condition of the reference line segment is not satisfied, the processor of the object recognition device may perform an eighteenth operation 735.

[0144] In some implementations, the fourth average condition may include a condition that the first average value of the reference line segment is greater than or equal to a first average threshold. The third variance condition may include a condition that the first variance value is greater than or equal to a first variance threshold.

[0145] In a seventeenth operation 733, in some implementations, the processor of the object recognition device may assign a marker corresponding to the second shape.

[0146] In an eighteenth operation 735 , in some implementations, the processor of the object recognition device may recognize that at least one contour point exists within the region of the non-reference line segment.

[0147] In operation 705 , the area of ​​the non-reference line segment may be the same as the area of ​​the non-reference line segment.

[0148] In a nineteenth operation 737 , in some implementations, the processor of the object recognition device may recognize that the first mean condition and the first variance condition of the reference line segment and the second mean condition and the second variance condition of the non-reference line segment are satisfied.

[0149] The first average condition may include a condition that the first average value of the reference line segment is greater than or equal to a third average threshold. The first variance condition may include a condition that the first variance value of the reference line segment is greater than or equal to a second variance threshold. The second average condition may include a condition that the second average value of the non-reference line segment is greater than or equal to a fourth average threshold. The second variance condition may include a condition that the second variance value of the non-reference line segment is greater than or equal to a third variance threshold.

[0150] In the twentieth operation 739, in some implementations, the processor of the object recognition device may recognize that the angle between the reference line segment and the non-reference line segment satisfies the second angle range. This is because when the angle between the reference line segment and the non-reference line segment satisfies the second angle range (e.g., greater than about 80° and less than about 105°), the shape of the distribution of the contour points is recognized as an L-shape with a short length.

[0151] In a twenty-first operation 741 , in some implementations, the processor of the object recognition device may assign a marker corresponding to the third shape.

[0152] Figure 8 A flowchart showing the operation of an object recognition device for assigning flags to multiple layers in an object recognition device or an object recognition method in some implementations of the present disclosure is shown.

[0153] In the following, it is assumed Figure 1 The object recognition device 101 performs Figure 8 In addition, Figure 8 In the description of FIG. 1 , an operation described as being performed by the object recognition device 101 or an operation described as being performed by the processor 105 may be understood to be controlled by the processor 105 of the object recognition device 101 .

[0154] See also Figure 8 In the first operation 801, in some implementations, the processor of the object recognition device may determine whether a layer assigned with the fourth shape is recognized. When the layer assigned with the fourth shape is recognized, the processor of the object recognition device may perform a second operation 803. When the layer assigned with the fourth shape is not recognized, the processor of the object recognition device may perform a third operation 805.

[0155] In some implementations, the processor of the object recognition device may identify whether there is a layer assigned with a fourth shape among the plurality of layers. The fourth shape (e.g., a broken shape) may indicate that the distribution of contour points representing the object and contained in a specific layer is not recognized as one of the first shape, the second shape, and the third shape. This is because when one of the plurality of layers is recognized as a layer assigned with the fourth shape, it is difficult to recognize the object as an object such as a vehicle.

[0156] In the second operation 803, in some implementations, the processor of the object recognition device may assign a marker corresponding to the fourth shape. The marker corresponding to the fourth shape may be assigned to a plurality of layers.

[0157] In the third operation 805, in some implementations, the processor of the object recognition device may determine whether a layer assigned with the first shape is recognized. When a layer assigned with the first shape is recognized, the processor of the object recognition device may perform a fourth operation 807. When a layer assigned with the first shape is not recognized, the processor of the object recognition device may perform a seventh operation 813.

[0158] In some implementations, when there is no layer assigned a mark corresponding to the fourth shape among multiple layers and there is a layer assigned a mark corresponding to the first shape, even when there is a layer assigned a mark corresponding to the first shape or there is a layer assigned a mark corresponding to the second shape, the processor of the object recognition device can also recognize that the distribution of contour points representing multiple layers has a high probability of having the first shape.

[0159] In the fourth operation 807, in some implementations, the processor of the object recognition device may determine whether the number of layers assigned the second shape is greater than the number of layers assigned the first shape. When the number of layers assigned the second shape is greater than the number of layers assigned the first shape, the processor of the object recognition device may perform a fifth operation 809. When the number of layers assigned the second shape is not greater than the number of layers assigned the first shape, the processor of the object recognition device may perform a sixth operation 811.

[0160] In some implementations, unlike existing object recognition devices, even if there is a layer assigned with a mark corresponding to the first shape among multiple layers, when the number of layers assigned with marks corresponding to the second shape is greater than the number of layers assigned with marks corresponding to the first shape, the processor of the object recognition device may also recognize the second shape as a shape representing the distribution of contour points of multiple layers.

[0161] In some implementations, when the reliability value of the layer corresponding to the first shape is less than the first reference reliability (e.g., 75) and the reliability value of the layer corresponding to the second shape is greater than or equal to the second reference reliability (e.g., 70), the processor of the object recognition device may assign the marker corresponding to the second shape to the plurality of layers even though the number of layers assigned the first shape is greater than the number of layers assigned the second shape. Fig. 9 A method of calculating the reliability of a layer corresponding to a first shape and a method of calculating the reliability of a layer corresponding to a second shape are described.

[0162] In a fifth operation 809, in some implementations, the processor of the object recognition device may assign a marker corresponding to the second shape. The marker corresponding to the second shape may be assigned to a plurality of layers.

[0163] In the sixth operation 811, in some implementations, the processor of the object recognition device may assign a marker corresponding to the first shape. The marker corresponding to the first shape may be assigned to a plurality of layers.

[0164] In the seventh operation 813, in some implementations, the processor of the object recognition device may determine whether the layer assigned the second shape is recognized. When the layer assigned the second shape is recognized, the processor of the object recognition device may perform an eighth operation 815. When the layer assigned the second shape is not recognized, the processor of the object recognition device may perform a ninth operation 817.

[0165] In some implementations, when a layer assigned to a first shape and a layer assigned to a fourth shape do not exist in a plurality of layers and a layer assigned to a second shape exists, a layer assigned to a third shape exists, and the processor of the object recognition device may also assign a marker corresponding to the second shape to the plurality of layers.

[0166] In an eighth operation 815, in some implementations, the processor of the object recognition device may assign a marker corresponding to the second shape. The marker corresponding to the second shape may be included in a plurality of layers.

[0167] In the ninth operation 817, in some implementations, the processor of the object recognition device may determine whether a layer assigned with the third shape is recognized. When the layer assigned with the third shape is recognized, the processor of the object recognition device may perform a tenth operation 819. When the layer assigned with the third shape is not recognized, the processor of the object recognition device may perform an eleventh operation 821.

[0168] In some implementations, when there is no layer assigned a marker corresponding to the first shape, the second shape, or the fourth shape among multiple layers, but there is a layer assigned a marker corresponding to the third shape, the processor of the object recognition device can recognize that the distribution of contour points representing the multiple layers has a high probability of being the third shape.

[0169] In some implementations, when the reliability value of the layer corresponding to the third shape is less than a third reference reliability (e.g., about 60), the processor of the object recognition device may assign the mark corresponding to the fourth shape to the layer even though there is no layer assigned with the mark corresponding to the first shape, the second shape, or the fourth shape and there is a layer assigned with the mark corresponding to the third shape. This is because, in the case of the third shape, low reliability may affect the recognition of the heading direction indicating the moving direction of the vehicle.

[0170] In a tenth operation 819, in some implementations, the processor of the object recognition device may assign a marker corresponding to the third shape. The marker corresponding to the third shape may be assigned to a plurality of layers.

[0171] In an eleventh operation 821, in some implementations, the processor of the object recognition device may assign a marker corresponding to the fourth shape. The marker corresponding to the fourth shape may be assigned to a plurality of layers.

[0172] Fig. 9 An embodiment of the distribution of contour points used for calculating reliability in an object recognition apparatus or an object recognition method in some implementations of the present disclosure is shown.

[0173] See also Fig. 9 , the first endpoint 901 and the second endpoint 903 may correspond to the outermost contour points among the contour points representing the object and included in the specific layer. The peak point 905 may include a contour point located at a position farthest from a line segment connecting the first endpoint 901 and the second endpoint 903 among the contour points representing the object and included in the specific layer.

[0174] In some implementations, the processor of the object recognition device may calculate the reliability of the layer recognized as the first shape based on at least one of the first variable value, the second variable value, the third variable value, the fourth variable value, and the fifth variable value, or any combination thereof.

[0175] The first line segment may include a line segment connecting the first endpoint 901 and the peak point 905. The second line segment may include a line segment connecting the second endpoint 903 and the peak point 905.

[0176] In some implementations, the reliability of the layer identified as the first shape can be identified based on the sum of at least one of a value obtained by multiplying a first variable value by a weight, a value obtained by multiplying a second variable value by a weight, a value obtained by multiplying a third variable value by a weight, a value obtained by multiplying a fourth variable value by a weight, and a value obtained by multiplying a fifth variable value by a weight, or any combination thereof.

[0177] In some implementations, the first variable value may include a value identified based on the distribution of contour points representing the object and included in a specific layer. For example, the processor of the object recognition device may identify the first variable value based on a variance value of the distance between a first line segment and contour points near the first line segment, and a variance value of the distance between a second line segment and contour points near the second line segment.

[0178] For example, when the variance value of the first line segment and / or the variance value of the second line segment is within a first threshold range (e.g., greater than 0 and less than or equal to 0.5), the first line segment value and / or the second line segment value may be individually identified as 50. For example, when the variance value of the first line segment and / or the variance value of the second line segment is within a second threshold range (e.g., greater than 0.5 and less than or equal to 1.5), the first line segment value and / or the second line segment value may be identified as 25. For example, when the variance value of the first line segment and / or the variance value of the second line segment is outside the second threshold range (e.g., greater than approximately 0.5 and less than or equal to approximately 1.5) and the first threshold range, the first line segment value and / or the second line segment value may be identified as 0.

[0179] In some implementations, a first variable value for identifying reliability of a layer identified as the first shape may be identified based on a value obtained by adding a first line segment value (e.g., about 50) to a second line segment value (e.g., about 50). For example, when a variance value of the first line segment and a variance value of the second line segment are within a first threshold range, the first variable value may be about 100.

[0180] In some implementations, the second variable value may be identified based on a distribution of contour points of the first line segment and a distribution of contour points of the second line segment.

[0181] For example, when contour points near a first line segment are all identified only in one direction relative to the first line segment, the first line segment value may be identified as approximately 50. For example, when contour points near a second line segment are all identified only in one direction relative to the second line segment, the second line segment value may be identified as approximately 50.

[0182] For example, when all contour points near a first line segment are identified in several directions relative to the first line segment, the first line segment value may be identified as 0. For example, when all contour points near a second line segment are identified in several directions relative to the second line segment, the second line segment value may be identified as 0.

[0183] In some implementations, a second variable value for identifying the reliability of the layer identified as the first shape may be identified based on a value obtained by adding a first line segment value (e.g., about 50) to a second line segment value (e.g., about 50). For example, when contour points near the first line segment are identified only in one direction relative to the first line segment and contour points near the second line segment are identified only in one direction relative to the second line segment, the second variable value may be identified as about 100.

[0184] In some implementations, the third variable value may be identified based on an average value of the angles of the contour points. The angle of each contour point may be an angle between a half line connecting the specific contour point to one of the contour points on both sides of the specific contour point and a half line connecting the specific contour point to another of the contour points on both sides of the specific contour point.

[0185] For example, when the average value of the angles of the contour points of the first line segment is within a first reference angle range (e.g., greater than 0° and less than or equal to 10°), the first line segment value may be identified as approximately 50. For example, when the average value of the angles of the contour points of the second line segment is within the first reference angle range (e.g., greater than 0° and less than or equal to 10°), the second line segment value may be identified as approximately 50.

[0186] For example, when the average value of the angles of the contour points of the first line segment is within a second reference angle range (e.g., greater than 10° and less than or equal to 25°), the first line segment value may be identified as approximately 25. For example, when the average value of the angles of the contour points of the second line segment is within a second reference angle range (e.g., greater than 10° and less than or equal to 25°), the second line segment value may be identified as approximately 25.

[0187] In some implementations, a third variable value for identifying the reliability of the layer identified as the first shape may be identified based on a value obtained by adding the first line segment value (e.g., about 50) to the second line segment value (e.g., about 50). For example, when the average value of the angles of the contour points of the first line segment is within the second reference angle range (e.g., greater than 10° and less than or equal to 25°) and the average value of the angles of the contour points of the second line segment is within the second reference angle range (e.g., greater than 10° and less than or equal to 25°), the fourth variable value may be identified as about 75.

[0188] In some implementations, the fourth variable value may be identified based on a distribution of contour points of the first line segment and a distribution of contour points of the second line segment.

[0189] For example, after dividing the first line segment into a specified number (eg, dividing the first line segment into four equal parts), when there is a contour point in the region of each divided line segment, the first line segment value may be identified as approximately 50.

[0190] For example, after dividing the second line segment into a specified number (eg, dividing the first line segment into four equal parts), the second line segment value may be identified as approximately 50 when there is a contour point in the region of each divided line segment.

[0191] In some implementations, a fourth variable value for identifying reliability of a layer identified as the first shape may be identified based on a value obtained by adding the first line segment value (eg, approximately 50) to the second line segment value (eg, approximately 50).

[0192] In some implementations, a fourth variable value for identifying the reliability of the layer identified as the first shape may be identified based on a value obtained by adding the first line segment value (e.g., about 50) to the second line segment value (e.g., about 50). For example, when there are contour points in a region of the bisecting line segment with respect to the first line segment and there are contour points in a region of the bisecting line segment with respect to the second line segment, the fourth variable value may be identified as about 100.

[0193] In some implementations, the fifth variable value may be identified based on a ratio of a length in one direction of a contour box identified as containing contour points representing an object and contained in a specific layer to a length in one direction of a cluster box identified as containing contour points contained in multiple layers.

[0194] In some implementations, when a ratio of a length of the contour box in one direction to a length of the cluster box in one direction is greater than a specified value (eg, approximately 75%), a fifth variable value may be identified as approximately 100.

[0195] Despite Fig. 9 Calculation of the reliability of a layer recognized as a first shape is described in , but the reliability of a layer recognized as a second shape may also be calculated. However, the reliability of a layer recognized as a second shape may be identified based on the first variable value and the third variable value.

[0196] Fig.10 A computing system related to an object recognition apparatus or an object recognition method in some implementations of the present disclosure is shown.

[0197] See also Fig.10 , the computing system 1000 may include at least one processor 1010 , a memory 1030 , a user interface input device 1040 , a user interface output device 1050 , a storage device 1060 , and a network interface 1070 connected to each other via a bus 1020 .

[0198] The processor 1010 may be a central processing unit (CPU) or a semiconductor device that processes instructions stored in the memory 1030 and / or the storage device 1060. The memory 1030 and the storage device 1060 may include various types of volatile or non-volatile storage media. For example, the memory 1030 may include a ROM (Read Only Memory) 1031 and a RAM (Random Access Memory) 1032.

[0199] Thus, the operations of a method or algorithm described in conjunction with the implementations disclosed herein may be implemented directly in hardware or software modules, or a combination thereof, executed by processor 1010. The software modules may reside on a storage medium (i.e., memory 1030 and / or storage device 1060), such as RAM, flash memory, ROM, EPROM, EEPROM, registers, hard disk, removable disk, and CD-ROM.

[0200] An exemplary storage medium may be coupled to the processor 1010, and the processor 1010 may read information from the storage medium and may record information in the storage medium. Alternatively, the storage medium may be integrated with the processor 1010. The processor and the storage medium may exist in an application specific integrated circuit (ASIC). The ASIC may reside in a user terminal. In another case, the processor and the storage medium may reside in a user terminal as separate components.

[0201] The above description is merely an illustration of the technical concept of the present disclosure, and those skilled in the art to which the present disclosure belongs may make various modifications and changes without departing from the essential characteristics of the present disclosure.

[0202] This technology can improve the accuracy of object recognition by identifying the distribution of contour points obtained by LIDAR of various shapes.

[0203] Furthermore, the present technology can increase recognition of a leading vehicle that is tilted relative to a host vehicle or a cutting-in vehicle by adding a shape type that represents the distribution of contour points obtained by LIDAR.

[0204] Furthermore, the present technology can improve the ability to recognize objects in a variety of situations by reducing the stringency of the criteria for determining a particular shape.

[0205] Furthermore, the present technology can improve the accuracy of object recognition by adjusting the priorities between the flags assigned to multiple layers.

[0206] In addition, this technology can reduce the risk of accidents in autonomous driving by improving the accuracy of object recognition.

[0207] In addition, the present technology can enhance the user experience by improving the accuracy of object recognition.

[0208] Furthermore, various effects directly or indirectly understood through the present disclosure can be provided.

Claims

1. An object recognition device, comprising: a light detection and ranging sensor configured to obtain contour points of an object; a memory for storing instructions; as well as A processor configured to execute the instructions to perform operations including: identifying whether the object is a dynamic object based on at least one of a reference line segment, a non-reference line segment, and a distribution of the contour points contained in a specific layer among a plurality of layers representing the object, identifying a shape of distribution of the contour points among the first shape, the second shape, and the third shape based on (i) the object being identified as the dynamic object, (ii) the length of the reference line segment being greater than a reference length, (iii) the angle between the reference line segment and the non-reference line segment being within a specified first angle range, (iv) the angle between the reference line segment and the non-reference line segment being not within the specified first angle range, or (v) the length of the reference line segment being less than or equal to the reference length, and assigning a marker corresponding to the identified shape to the particular layer, The reference line segment includes (i) a line segment connecting a peak point and a first endpoint and (ii) a first line segment among line segments connecting the peak point and a second endpoint different from the first endpoint, the first endpoint and the second endpoint being two outermost endpoints of the contour points, The non-reference line segment includes a second line segment among the line segment connecting the first endpoint and the peak point and the line segment connecting the second endpoint and the peak point, the length of the second line segment is less than the length of the first line segment, and The peak point includes a contour point located farthest from a line segment connecting the first endpoint and the second endpoint of the contour points.

2. The object recognition device according to claim 1, wherein: The operations also include: assigning a flag corresponding to a fourth shape to the specific layer based on that a shape of the distribution of the contour points is not recognized as one of the first shape, the second shape, and the third shape, The fourth shape represents that a shape of distribution of the contour points is not recognized as one of the first shape, the second shape, and the third shape.

3. The object recognition device according to claim 1, wherein: Identifying whether the object is the dynamic object includes: identifying whether the object is the dynamic object based on a size of a contour box including the contour points being within a first reference size range and based on at least one of the reference line segments, the non-reference line segments, and the distribution of the contour points.

4. The object recognition device according to claim 2, in, Identifying whether the object is a dynamic object comprises: based on the fact that the size of the contour box including the contour points is not within a first reference size range but within a second reference size range, identifying whether the object is a dynamic object based on at least one of the reference line segments, the distribution of the contour points, and the length of the reference line segments, and The operation further includes: identifying whether a first average value is less than a first average threshold value, the first average value being an average value of at least one distance between the reference line segment and at least one contour point included in a region to which the reference line segment belongs, the region to which the reference line segment belongs being a region to which the reference line segment belongs among two regions separated by a straight line that bisects an angle between the reference line segment and the non-reference line segment and passes through a peak point among the contour points, identifying whether a first variance value as a variance value of the at least one distance is less than a first variance threshold, recognizing that the shape of the distribution of the contour points is the second shape based on the object being recognized as the dynamic object, the first average value being less than the first average threshold, and the first variance value being less than the first variance threshold, and assigning a flag corresponding to the second shape to the specific layer, Based on the object being identified as the dynamic object and the first variance value being greater than or equal to the first variance threshold or the first average being greater than or equal to the first average threshold, identifying whether the shape of the distribution of the contour points is the third shape, The shape based on the distribution of the contour points is recognized as the third shape, a flag corresponding to the third shape is assigned to the specific layer, and Based on the shape of the distribution of the contour points not being recognized as one of the first shape, the second shape, and the third shape, a flag corresponding to the fourth shape is assigned to the specific layer.

5. The object recognition device according to claim 1, wherein: The first shape representation includes a shape of a first L-shaped distribution identified by all or part of the contour points, wherein the second shape representation includes an I-shaped distribution shape identified by all or part of the contour points, and The third shape represents a shape including a second L-shaped distribution identified by all or part of the contour points, wherein the length of the reference line segment in the second L-shaped distribution is shorter than the length of the reference line segment in the first L-shaped distribution corresponding to the first shape.

6. The object recognition device according to claim 1, wherein: The operations also include: Based on the object being identified as a dynamic object and the length of the reference line segment being less than or equal to the reference length, identifying that a ratio of a length of a line segment in one direction of a cluster box identified based on the plurality of layers to a length of a line segment in the one direction of a contour box containing the contour point is less than a reference ratio; identifying that the angle between the reference line segment and the non-reference line segment satisfies a second angle range; and Based on the ratio being less than the reference ratio and the angle between the reference line segment and the non-reference line segment being within the second angle range, it is identified that the shape of the distribution of the contour points is the third shape, and a marker corresponding to the third shape is assigned to the specific layer.

7. The object recognition device according to claim 1, wherein: The operations also include: Based on the fact that the object is identified as the dynamic object, the length of the reference line segment is greater than the reference length, and the angle between the reference line segment and the non-reference line segment is not within the first angle range, identifying whether a first average value is less than or equal to a second average threshold value, the first average value being an average value of at least one distance between the reference line segment and at least one contour point included in a region to which the reference line segment belongs, the region to which the reference line segment belongs being a region to which the reference line segment belongs among two regions separated by a straight line that divides the angle between the reference line segment and the non-reference line segment in half and passes through a peak point among the contour points; identifying whether the angle between the reference line segment and the non-reference line segment is within a third angle range; and Based on the first average value being less than or equal to the second average threshold and the angle between the reference line segment and the non-reference line segment being within the third angle range, the shape of the distribution of the contour points is identified as the second shape, and a marker corresponding to the second shape is assigned to the specific layer.

8. The object recognition device according to claim 4, wherein: The operations also include: Based on the object being identified as a dynamic object, the first average value being greater than or equal to the first average threshold, or the first variance value being greater than or equal to the first variance threshold, identifying whether there is at least one contour point in a region between (i) a straight line passing through an endpoint included in the non-reference line segment and perpendicular to the non-reference line segment and (ii) a straight line passing through the peak point and perpendicular to the non-reference line segment; identifying whether the first average value is less than a third average threshold value; identifying whether the first variance value is less than a second variance threshold; identifying whether a second average value is less than a fourth average threshold value, the second average value being an average value of at least one distance between the non-reference line segment and at least one contour point included in a region to which the non-reference line segment belongs, the region to which the non-reference line segment belongs being a region to which the non-reference line segment belongs among two regions separated by a straight line that bisects an angle between the reference line segment and the non-reference line segment and passes through the peak point among the contour points; identifying whether a second variance value is less than a third variance threshold, the second variance value being a variance of at least one distance between the non-reference line segment and at least one contour point in a region to which an endpoint included in the non-reference line segment belongs; identifying whether the angle between the reference line segment and the non-reference line segment is within a second angle range; and Based on the fact that at least one of the contour points is identified as existing in the area between a straight line passing through an endpoint included in the non-reference line segment and perpendicular to the non-reference line segment and a straight line passing through the peak point and perpendicular to the non-reference line segment, the first average value is less than the third average threshold, the second average value is less than the fourth average threshold, the first variance value is less than the second variance threshold, the second variance value is less than the third variance threshold, and the angle between the reference line segment and the non-reference line segment is within the second angle range, it is identified that the shape of the distribution of the contour points is the third shape, and a flag corresponding to the third shape is assigned to the specific layer.

9. The object recognition device according to claim 4, wherein: Among the line segments of the outline frame that meet the first reference size range, among the line segments in a first direction and the line segments in a second direction different from the first direction, a length of a first line segment is greater than a first length value and less than a second length value that is greater than the first length value, Among the line segments in the first direction and the line segments in the second direction, the length of a second line segment is greater than or equal to a third length value, Among the line segments of the outline frame that meet the second reference size range, among the line segments in a third direction and the line segments in a fourth direction different from the third direction, the length of a third line segment is greater than the first length value, and Among the line segments in the third direction and the line segments in the fourth direction, a length of a fourth line segment is less than a fourth length value, and the fourth length value is greater than the third length value.

10. The object recognition device according to claim 2, wherein: The operations also include: identifying, among the plurality of layers, a layer to which a marker corresponding to the fourth shape is assigned; identifying, among the plurality of layers, a layer to which a marker corresponding to the first shape is assigned; identifying whether the number of layers among the plurality of layers to which the mark corresponding to the first shape is assigned is less than the number of layers to which the mark corresponding to the second shape is assigned; and Based on the fact that the layer assigned with the mark corresponding to the fourth shape is not identified, the layer assigned with the mark corresponding to the first shape is identified, and the number of the layers assigned with the mark corresponding to the first shape in the multiple layers is less than the number of the layers assigned with the mark corresponding to the second shape, the mark corresponding to the second shape is assigned to the multiple layers.

11. The object recognition device according to claim 2, wherein: The operations also include: identifying, among the plurality of layers, a layer to which a marker corresponding to the fourth shape is assigned; identifying, among the plurality of layers, a layer to which a marker corresponding to the first shape is assigned; identifying, among the plurality of layers, a layer to which a marker corresponding to the second shape is assigned; and Based on the fact that the layer assigned with the mark corresponding to the fourth shape is not recognized, the layer assigned with the mark corresponding to the first shape is not recognized, and the layer assigned with the mark corresponding to the second shape is recognized, the mark corresponding to the second shape is assigned to the multiple layers.

12. The object recognition device according to claim 2, wherein: The operations also include: identifying, among the plurality of layers, a layer to which a marker corresponding to a fourth shape is assigned; identifying, among the plurality of layers, a layer to which a marker corresponding to the first shape is assigned; identifying, among the plurality of layers, a layer to which a marker corresponding to the second shape is assigned; identifying, among the plurality of layers, a layer to which a marker corresponding to the third shape is assigned; and Based on the fact that the layer assigned with the mark corresponding to the fourth shape is not recognized, the layer assigned with the mark corresponding to the first shape is not recognized, the layer assigned with the mark corresponding to the second shape is not recognized, and the layer assigned with the mark corresponding to the third shape is recognized, the mark corresponding to the third shape is assigned to the multiple layers.

13. A method for object recognition, comprising: identifying, by a processor, whether the object is a dynamic object based on at least one of a reference line segment, a non-reference line segment, and a distribution of contour points included in a specific layer among a plurality of layers representing the object, the contour points of the object being obtained by a light detection and ranging sensor; The processor identifies, based on the object being identified as a dynamic object, the length of the reference line segment being greater than a reference length, the angle between the reference line segment and the non-reference line segment being within a specified first angle range, the angle between the reference line segment and the non-reference line segment being not within the specified first angle range, or the length of the reference line segment being less than or equal to the reference length, a shape of distribution of the contour points among a first shape, a second shape, and a third shape; as well as assigning a marker corresponding to the identified shape to the particular layer, The reference line segment includes a line segment connecting a peak point and a first endpoint and a first line segment among line segments connecting the peak point and a second endpoint different from the first endpoint, the first endpoint and the second endpoint being two outermost endpoints of the contour points, The non-reference line segment includes a second line segment among the line segment connecting the first endpoint and the peak point and the line segment connecting the second endpoint and the peak point, the length of the second line segment is less than the length of the first line segment, and The peak point includes a contour point located at a position farthest from a line segment connecting a first endpoint and a second endpoint of the contour points.

14. The object recognition method according to claim 13, further comprising: assigning, by the processor, a flag corresponding to a fourth shape to the specific layer based on the shape of the distribution of the contour points not being recognized as one of the first shape, the second shape, and the third shape, The fourth shape indicates that a shape of the distribution of the contour points is not recognized as one of the first shape, the second shape, and the third shape.

15. The object recognition method according to claim 13, wherein: Identifying, by the processor, whether the object is a dynamic object based on at least one of the reference line segments, the non-reference line segments, and the distribution of the contour points included in the specific layer among the multiple layers representing the object includes: identifying whether the object is a dynamic object based on at least one of the reference line segments, the non-reference line segments, and the distribution of the contour points, based on a size of a contour box including the contour points being within a first reference size range.

16. The object recognition method according to claim 14, further comprising: identifying, by the processor, whether the object is a dynamic object based on at least one of the reference line segments, the distribution of the contour points, and the length of the reference line segments, based on the fact that a size of a contour box including the contour points is not within a first reference size range but is within a second reference size range; identifying, by the processor, whether a first average value is less than a first average threshold value, the first average value being an average value of at least one distance between the reference line segment and at least one contour point included in a region to which the reference line segment belongs, the region to which the reference line segment belongs being a region to which the reference line segment belongs among two regions separated by a straight line that bisects an angle between the reference line segment and the non-reference line segment and passes through a peak point among the contour points; identifying, by the processor, whether a first variance value as a variance value of the at least one distance is less than a first variance threshold, recognizing that a shape of the distribution of the contour points is the second shape based on that the object is a dynamic object, the first average value is less than the first average threshold, and the first variance value is less than the first variance threshold, and assigning a flag corresponding to the second shape to the specific layer; The processor identifies, based on the fact that the object is a dynamic object and the first variance value is greater than or equal to the first variance threshold or the first average value is greater than or equal to the first average threshold, whether the shape of the distribution of the contour points is the third shape, and assigns a flag corresponding to the third shape to the specific layer based on the shape of the distribution of the contour points being identified as the third shape; as well as A flag corresponding to the fourth shape is assigned to the specific layer based on the shape of the distribution of the contour points not being recognized as one of the first shape, the second shape, and the third shape by the processor.

17. The object recognition method according to claim 13, wherein: The first shape representation includes a shape of a first L-shaped distribution identified by all or part of the contour points, wherein the second shape representation includes an I-shaped distribution shape identified by all or part of the contour points, and The third shape represents a shape including a second L-shaped distribution identified by all or part of the contour points, wherein the length of the reference line segment in the second L-shaped distribution is shorter than the length of the reference line segment in the first L-shaped distribution corresponding to the first shape.

18. The object recognition method according to claim 13, wherein: The processor identifies, based on the object being the dynamic object and the length of the reference line segment being less than or equal to the reference length, whether the shape of the distribution of the contour points is the third shape, and assigns a flag corresponding to the third shape to the specific layer based on the shape of the distribution of the contour points being identified as the third shape. The processor identifies, based on the object being the dynamic object and the length of the reference line segment being less than or equal to the reference length, whether the shape of the distribution of the contour points is the third shape, and assigns a flag corresponding to the third shape to the specific layer based on the shape of the distribution of the contour points being identified as the third shape, including: The processor identifies, based on the object being identified as a dynamic object and the length of the reference line segment being less than or equal to the reference length, that a ratio of a length of a line segment in one direction of a cluster box identified based on the plurality of layers to a length of a line segment in the one direction of a contour box containing the contour point is less than a reference ratio; identifying that the angle between the reference line segment and the non-reference line segment is within a second angle range; and Based on the ratio being less than the reference ratio and the angle between the reference line segment and the non-reference line segment being within the second angle range, it is identified that the shape of the distribution of the contour points is the third shape, and a marker corresponding to the third shape is assigned to the specific layer.

19. The object recognition method according to claim 13, wherein: The processor identifies, based on the fact that the object is the dynamic object, the length of the reference line segment is greater than the reference length, and the angle between the reference line segment and the non-reference line segment is not within the first angle range, whether the shape of the distribution of the contour points is the second shape, and assigns a flag corresponding to the second shape to the specific layer based on the fact that the shape of the distribution of the contour points is identified as the second shape. The processor identifies, based on the fact that the object is the dynamic object, the length of the reference line segment is greater than the reference length, and the angle between the reference line segment and the non-reference line segment is not within the first angle range, whether the shape of the distribution of the contour points is the second shape, and assigns a flag corresponding to the second shape to the specific layer based on the fact that the shape of the distribution of the contour points is identified as the second shape, including: identifying, by the processor, whether a first average value is less than or equal to a second average threshold value based on the object being identified as the dynamic object, the length of the reference line segment being greater than the reference length, and the angle between the reference line segment and the non-reference line segment being not within the first angle range, the first average value being an average value of at least one distance between the reference line segment and at least one contour point included in a region to which the reference line segment belongs, the region to which the reference line segment belongs being a region to which the reference line segment belongs among two regions separated by a straight line that divides the angle between the reference line segment and the non-reference line segment in half and passes through a peak point among the contour points; identifying, by the processor, whether the angle between the reference line segment and the non-reference line segment is within a third angle range; and The processor identifies that the shape of the distribution of the contour points is the third shape based on the first average value being less than or equal to the second average threshold and the angle between the reference line segment and the non-reference line segment being within the third angle range, and assigns a mark corresponding to the third shape to the specific layer.

20. The object recognition method according to claim 16, wherein: The processor identifies, based on the fact that the object is the dynamic object and the first variance value is greater than or equal to the first variance threshold or the first average value is greater than or equal to the first average threshold, whether the shape of the distribution of the contour points is the third shape, and assigns a flag corresponding to the third shape to the specific layer based on the fact that the shape of the distribution of the contour points is identified as the third shape. The processor identifies, based on the fact that the object is the dynamic object and the first variance value is greater than or equal to the first variance threshold or the first average value is greater than or equal to the first average threshold, whether the shape of the distribution of the contour points is the third shape, and assigns a flag corresponding to the third shape to the specific layer based on the fact that the shape of the distribution of the contour points is identified as the third shape, including: The processor identifies, based on the object being a dynamic object, the first average being greater than or equal to the first average threshold, or the first variance being greater than or equal to the first variance threshold, whether there is at least one contour point in a region between (i) a straight line passing through an endpoint included in the non-reference line segment and perpendicular to the non-reference line segment and (ii) a straight line passing through the peak point and perpendicular to the non-reference line segment; identifying, by the processor, whether the first average value is less than a third average threshold value; identifying, by the processor, whether the first variance value is less than a second variance threshold; identifying, by the processor, whether a second average value is less than a fourth average threshold value, the second average value being an average value of at least one distance between the non-reference line segment and at least one contour point included in a region to which the non-reference line segment belongs, the region to which the non-reference line segment belongs being a region to which the non-reference line segment belongs among two regions separated by a straight line that bisects an angle between the reference line segment and the non-reference line segment and passes through the peak point among the contour points; identifying, by the processor, whether a second variance value is less than a third variance threshold, the second variance value being a variance of at least one distance between the non-reference line segment and at least one contour point in a region to which an endpoint included in the non-reference line segment belongs; identifying, by the processor, whether the angle between the reference line segment and the non-reference line segment is within a second angle range; and The processor identifies that the shape of the distribution of the contour points is the third shape, and assigns a marker corresponding to the third shape to the specific layer based on the fact that at least one of the contour points is identified as existing in the area between a straight line passing through an endpoint included in the non-reference line segment and perpendicular to the non-reference line segment and a straight line passing through the peak point and perpendicular to the non-reference line segment, the first average value is less than the third average threshold, the second average value is less than the fourth average threshold, the first variance value is less than the second variance threshold, the second variance value is less than the third variance threshold, and the angle between the reference line segment and the non-reference line segment is within the second angle range.

Images