# US 12,217,492 B2 · Track segment cleaning of tracked objects > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,217,492 B2 - Document ID: US12217492B2 - Status: Granted - Kind: B2 - Application number: 18/180,732 - Filed: 2023-03-08 - Published: 2023-11-09 - Granted: 2025-02-04 - Earliest claimed: 2022-05-04 - Assignees: Motional AD LLC - Applicants: Motional AD LLC - Inventors: Lubing Zhou; Jiong Yang; Yilaun Guo - Web page: https://patentgrep.com/patent/US12217492B2 ## CPC - G06V 10/806: {of extracted features} - G06V 20/584: {of vehicle lights or traffic lights} - G06T 7/248: {involving reference images or patches} - G06V 10/26: Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion - G06V 10/774: Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting - G06V 10/82: using neural networks - G06V 10/993: {Evaluation of the quality of the acquired pattern} - G06V 20/58: Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads - G06V 20/582: {of traffic signs} - G06T 2207/20081: Training; Learning - G06T 2207/30241: Trajectory - G06T 2207/30252: Vehicle exterior; Vicinity of vehicle - G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING - G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL ## Abstract Provided are methods for track segment cleaning of tracked objects using neural networks, which can include detecting a first track segment and a second track segment. The method includes applying a machine learning model trained to determine if the first track segment and second track segment capture real objects and if the first track segment and the second track segment are representative of an identical object exterior to a vehicle. The method further includes combining the first track segment and the second track segment to form a single track segment having a single trajectory in response to the first track segment and the second track segment being determined to be representative of the identical object. Systems and computer program products are also provided. ## Claims ### Claim 1 A method comprising: detecting, using at least one processor, a first track segment and a second track segment; determining, using the at least one processor, that the first track segment and the second track segment are representative of an identical object exterior to a vehicle; determining, using the at least one processor, a stitching score between the first track segment and the second track segment, the stitching score indicative of a likelihood that combining the first track segment and the second track segment is representative of a single trajectory of the identical object exterior to the vehicle; and in response to the first track segment and the second track segment being determined to be representative of the identical object by the stitching score satisfying a stitching score threshold, combining, using the at least one processor, the first track segment and the second track segment to form a single track segment comprising the single trajectory. ### Claim 2 (depends on 1) The method of claim 1, further comprising: determining, using the at least one processor, a first trajectory feature of the first track segment and second trajectory feature of the second track segment, the first trajectory feature comprising at least one of a first center point, a first size, or a first heading angle, and the second trajectory feature comprising at least one of a second center point, a second size relative to the second center point, or a second heading angle, wherein the stitching score is based on the first trajectory feature of the first track segment and the second trajectory feature of the second track segment. ### Claim 3 (depends on 2) The method of claim 2, further comprising: concatenating, using the at least one processor, the first trajectory feature of the first track segment with the second trajectory feature of the second track segment to determine the stitching score. ### Claim 4 (depends on 1) The method of claim 1, further comprising: determining, using the at least one processor, a first cloud point feature of the first track segment and a second cloud point feature of the second track segment, the first cloud point feature configured to represent at least one of a first 3D characteristic of the identical object or a first motion of the identical object, and the second cloud point feature configured to represent at least one of a second 3D characteristic of the identical object or a second motion of the identical object, wherein the stitching score is based on first cloud point features of the first track segment and second cloud point features of the second track segment. ### Claim 5 (depends on 4) The method of claim 4, wherein the first cloud point feature comprises at least one of a first track point with first coordinates, a first intensity, or a first timestamp, and wherein the second cloud point feature comprises at least one of a second track point with second coordinates, a second intensity, or a second timestamp. ### Claim 6 (depends on 4) The method of claim 4, further comprising: concatenating, using the at least one processor, the first cloud point feature of the first track segment with the second cloud point feature of the second track segment to determine the stitching score. ### Claim 7 (depends on 1) The method of claim 1, further comprising: determining, using the at least one processor, a first image feature of the first track segment and second image feature of the second track segment, the first image feature comprises at least one of a first segmentation, a first embedded feature, or a first pixel collection related to the identical object, and the second image feature comprises at least one of a second segmentation, a second embedded feature, or a second pixel collection related to the identical object, wherein the stitching score is based on the first image feature of the first track segment and the second image feature of the second track segment. ### Claim 8 (depends on 7) The method of claim 7, further comprising: concatenating, using the at least one processor, the first image feature of the first track segment with the second image feature of the second track segment to determine the stitching score. ### Claim 9 (depends on 1) The method of claim 1, further comprising: generating, using the at least one processor, at least one training sample by dividing a known single track into at least two training track segments, wherein the known single track includes a stitching score label indicating that the at least two training track segments form the known single track. ### Claim 10 (depends on 9) The method of claim 9, further comprising: training, using the at least one processor and based on the at least one training sample, a machine learning model to recognize the at least two training track segments that form the known single track. ### Claim 11 A system comprising: at least one processor, and at least one non-transitory storage media storing instructions that, when executed by the at least one processor, cause the at least one processor to perform operations comprising: detecting a first track segment and a second track segment; determining that the first track segment and the second track segment are representative of an identical object exterior to a vehicle; determining a stitching score between the first track segment and the second track segment, the stitching score indicative of a likelihood that combining the first track segment and the second track segment is representative of a single trajectory of the identical object exterior to the vehicle; and in response to the first track segment and the second track segment being determined to be representative of the identical object by the stitching score satisfying a stitching score threshold, combining the first track segment and the second track segment to form a single track segment comprising the single trajectory and a single cloud-point feature. ### Claim 12 (depends on 11) The system of claim 11, further comprising: determining, using the at least one processor, a first trajectory feature of the first track segment and second trajectory feature of the second track segment, the first trajectory feature comprising at least one of a first center point, a first size relative to the first center point, or a first heading angle, and the second trajectory feature comprising at least one of a second center point, a second size relative to the second center point, or a second heading angle, wherein the stitching score is based on the first trajectory feature of the first track segment and the second trajectory feature of the second track segment. ### Claim 13 (depends on 12) The system of claim 12, further comprising: concatenating, using the at least one processor, the first trajectory feature of the first track segment with the second trajectory feature of the second track segment to determine the stitching score. ### Claim 14 (depends on 11) The system of claim 11, further comprising: determining, using the at least one processor, a first cloud point feature of the first track segment and a second cloud point feature of the second track segment, the first cloud point feature configured to represent at least a first 3D characteristic of the identical object or a first motion of the identical object, and the second cloud point feature configured to represent at least one of a second 3D characteristic of the identical object or a second motion of the identical object, wherein the stitching score is based on first cloud point features of the first track segment and second cloud point features of the second track segment. ### Claim 15 (depends on 14) The system of claim 14, wherein the first cloud point feature comprises at least one of a first track point with first coordinates, a first intensity, or a first timestamp, and the second cloud point feature comprises at least one of a second track point with second coordinates, a second intensity, or a second timestamp. ### Claim 16 (depends on 14) The system of claim 14, further comprising: concatenating, using the at least one processor, the first cloud point feature of the first track segment with the second cloud point feature of the second track segment to determine the stitching score. ### Claim 17 (depends on 11) The system of claim 11, further comprising: determining, using the at least one processor, a first image feature of the first track segment and second image feature of the second track segment, the first image feature comprises at least one of a first segmentation, a first embedded feature, or a first pixel collection related to the identical object, and the second image feature comprises at least one of a second segmentation, a second embedded feature, or a second pixel collection related to the identical object, wherein the stitching score is based on the first image feature of the first track segment and the second image feature of the second track segment. ### Claim 18 (depends on 17) The system of claim 17, further comprising: concatenating, using the at least one processor, the first image feature of the first track segment with the second image feature of the second track segment to determine the stitching score. ### Claim 19 (depends on 11) The system of claim 11, further comprising: generating, using the at least one processor, at least one training sample by dividing a known single track into at least two training track segments, wherein the known single track includes a stitching score label indicating that the at least two training track segments form the known single track. ### Claim 20 A non-transitory computer-readable storage medium comprising at least one program for execution by one or more processors of a first device, the at least one program including instructions which, when executed by the one or more processors, cause the first device to perform operations comprising: detecting a first track segment and a second track segment; determining that the first track segment and the second track segment are representative of an identical object exterior to a vehicle; determining a stitching score between the first track segment and the second track segment, the stitching score indicative of a likelihood that combining the first track segment and the second track segment is representative of a single trajectory of the identical object exterior to the vehicle; and in response to the first track segment and the second track segment being determined to be representative of the identical object by the stitching score satisfying a stitching score threshold, combining the first track segment and the second track segment to form a single track segment comprising the single trajectory and a single cloud-point feature.