IP Library › Granted Patent US 12,380,706
Granted Patent B2
US 12,380,706 · App. 18/203,385 · Granted Aug 5, 2025

System for detecting obstacle state and an operation method thereof

Inventors: Jiun-In Guo (Hsinchu, TW); Zhe-Lun Hu (Tainan, TW)
Assignee: NATIONAL YANG MING CHIAO TUNG UNIVERSITY
G06V20/58G06T7/50G06V10/26G06V10/40G06V10/764G06V10/7715G06V10/774G06T2207/30261
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,380,706
App. No.
18/203,385
Filed
May 30, 2023
Granted
Aug 5, 2025
Kind
B2
Art Unit
2667
USPC
382/103
Abstract

The invention provides a system for detecting obstacle state and an operating method thereof, comprising an image capturing module, a semantic segmentation module, a feature extraction module, an object detection module, and a distance table calibration module. The invention is delivered a semantic segmentation information to a model for processing self-learning, and selected an output of an original image size, for a carrier of an attention mechanism.

Claims (18)

1. A system for detecting obstacle state, comprising:

an image capturing module, capturing an image at a fixed frequency, an image comprises a road information;

a semantic segmentation module, calculating an image and outputting a probability classification map of the road information based on a feature data;

a feature extraction module, extracting the feature data of the road information, and outputting an image information, and a probability distribution map;

an object detection module, training the object detection module by using the image, the feature data, the image information, and the road information; and

a distance table calibration module, using a distance measurement algorithm method to calculate and store a distance value for each pixel in the image, and directly querying a distance table for a distance value of each pixel in the image.

2. The system for detecting obstacle state according to claim 1 , wherein a function of a feature extraction module comprises based on an image information of a semantic segmentation module and a probability distribution map, after through extracting and selecting features of the probability classification map, changes the image information, and a size of the probability distribution map to be the same as the size of the image, and being output.

3. The system for detecting obstacle state according to claim 1 , wherein a function of a feature extraction module comprising treats an image information and a probability distribution map by using a softmax calculation method, in order to change a feature map information, and an information is normalized to between 0 and 1.

4. The system for detecting obstacle state according to claim 1 , further comprising a function of an object detection module being using a probability distribution map output by a feature extraction module and the image being merged with as an input for training.

5. The system for detecting obstacle state according to claim 1 , wherein a distance table calibration module comprising sets a road surface level in a distance table calibration module, and using a measuring tool in advance to record an actual distance of a point, and marking a pixel at a position, using a different algorithm method, calibrating and creating a distance table of a same size as an image of an image capturing module, a distance value of any pixel point being obtained by querying the distance table.

6. A method for detecting obstacle state operation, comprising:

capturing respectively an image at a fixed frequency;

calculating an image and outputting a probability classification map of road information based on a feature data;

extracting a feature data of the road information and outputting an image information, and a probability distribution map;

training an object detection module by using the image, the feature data, the image information, and the road information;

calculating and storing a distance value for each pixel in the image using a ranging algorithm method to establish a distance table; and

obtaining the image information, and querying the distance table to determine the distance value represented by a pixel in the image.

7. The system for detecting obstacle state according to claim 1 , further comprising merging a probability distribution map and an image as an input for training.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded May 30, 2023
From: GUO, JIUN-IN; HU, ZHE-LUN
To: NATIONAL YANG MING CHIAO TUNG UNIVERSITY
Reel/Frame 063797/0205 →
Priority Claims (1)
TW 112108457 · Mar 8, 2023 · national
Continuity (1)
Related Publication 20240304000A1 · Sep 12, 2024
References Cited (10)
US 10438371B2 · Xu · 2019 [cited by examiner]
US 11538185B2 · Kaiser · 2022 [cited by examiner]
US 12039788B2 · Guo · 2024 [cited by examiner]
US 12051206B2 · Chen · 2024 [cited by examiner]
US 20200293751A1 · Zeng · 2020 [cited by examiner]
US 20210110174A1 · Ohgushi · 2021 [cited by examiner]
US 20210146952A1 · Vora · 2021 [cited by examiner]
US 20210350705A1 · Guo · 2021 [cited by examiner]
Liu et al., Vision Based Environmental Perception for Autonomous Driving, arXiv2212-11453 [[cs.CV], pp. 1-39. (Year: 2022). [cited by examiner]
Ramos et al., Detecting Unexpected Obstacles for Self-Driving Cars: Fusing Deep Learning and Geometric Modeling, IEEE 978-15090-4804-5/17, IEEE Intelligent Vehicle Symposium, Jun. 11-14, 2017, Redondo Beach, pp. 1025-10… [cited by examiner]