KR-20260064965-A - Pedestrian Detection System Based on Driver Using Deep Learning

KR20260064965AKR 20260064965 AKR20260064965 AKR 20260064965AKR-20260064965-A

Abstract

The present invention relates to a driver-based pedestrian detection system utilizing deep learning, comprising: a camera installed in a vehicle to transmit image information capturing the area in front of and around the vehicle; and a server that receives the image information and performs image processing through a deep learning model; wherein the server comprises an image collection unit that collects image information captured through the camera; an image preprocessing unit that removes noise and normalizes image data collected from the image collection unit using a preset filter; an image processing unit that detects pedestrians and objects around the vehicle in real time using a YOLOv8-seg model based on image information provided from the image preprocessing unit; and an output unit that provides a warning signal in real time to an in-vehicle display or audio device when a pedestrian is detected.

Inventors

최원혁
정우진
한조원
임영근

Assignees

한서대학교 산학협력단

Dates

Publication Date: 20260508
Application Date: 20241030

Claims (3)

A camera installed on a vehicle to transmit video information capturing the front and surrounding areas of the vehicle; and A server that receives the above image information and performs image processing through a deep learning model; comprising, The above server is, A video collection unit that collects video information captured through a camera, and An image preprocessing unit that removes noise and normalizes collected image data using a preset filter, and An image processing unit that detects pedestrians and objects around a vehicle in real time using a YOLOv8-seg model based on preprocessed image information, and A driver-based pedestrian detection system utilizing deep learning, characterized by including a transmission unit that transmits a warning signal in real time to an in-vehicle display or audio device when a pedestrian at a crosswalk is detected.
In paragraph 1, A driver-based pedestrian detection system using deep learning, characterized in that the above-described image processing unit detects pedestrians, vehicles, and crosswalk objects by analyzing video information collected in real time using a pedestrian dataset trained through a YOLOv8-seg model.
In paragraph 1, The above YOLOv8-seg model is a driver-based pedestrian detection system utilizing deep learning, characterized by using Instance Segmentation by adding a Segmentation Branch structure to the algorithm.

Description

Driver-based Pedestrian Detection System Using Deep Learning The present invention relates to a driver-based pedestrian detection system utilizing deep learning, and more specifically, to a driver-based pedestrian detection system utilizing deep learning capable of detecting pedestrians in real time using deep learning technology. In Korea, approximately 35,000 or more pedestrian traffic accidents occur annually, with an average of 35,000 injuries. According to recent statistics, the number of pedestrian traffic accident fatalities in South Korea is about 900, which corresponds to approximately 1.7 deaths per 100,000 people. This figure is significantly high compared to the average of OECD countries. Furthermore, an average of over 500 traffic accidents involving children have occurred within school zones over the past five years, with more than 500 injuries occurring each time. These figures indicate the need for additional efforts to enhance pedestrian safety. There is a bill titled "Crime of Causing Death or Injury in Child Protection Zones," promulgated by the National Assembly on December 24, 2019. This act is an amendment composed of revisions to the Road Traffic Act and the Act on Aggravated Punishment, and is commonly known as the "Min-sik Law." The Crime of Causing Death or Injury in Child Protection Zones came into effect on March 25, 2020, strengthening safety measures to prevent accidents within child protection zones and crosswalks. In addition, the Road Traffic Act known as the Right Turn Act has been in effect since the end of January 2023. According to the enforcement rules of the Right Turn Act, when turning right when the traffic light is red, drivers must stop at the stop line, crosswalk, or immediately before the intersection before proceeding with the turn. Methods to improve pedestrian accidents in crosswalk areas are increasing, such as the installation of auxiliary traffic lights like right-turn signals. FIG. 1 is a block diagram showing a driver-based pedestrian detection system utilizing deep learning according to the present invention. Embodiments of the present invention are described below with reference to the attached drawings so that those skilled in the art can easily implement the invention. However, the present invention may be embodied in various different forms and is not limited to the embodiments described herein. Then, a preferred embodiment of a driver-based pedestrian detection system utilizing deep learning according to the present invention will be described in detail. FIG. 1 is a block diagram showing a driver-based pedestrian detection device utilizing deep learning according to the present invention. Referring to FIG. 1, a driver-based pedestrian detection system utilizing deep learning according to the present invention may include a camera (100) and a server (200). The above camera (100) can be installed on a vehicle to capture image information of the front and surrounding areas of the vehicle and transmit it to a server (200). The above server (200) can receive image information and perform image processing through a deep learning model. The above server (210) may include an image collection unit (210), an image preprocessing unit (220), an image processing unit (230), and an output unit (240). The video collection unit (210) can collect video information captured through the camera (100). The image preprocessing unit (220) can remove noise and normalize the collected image data using a preset filter. That is, the image preprocessing unit (220) can remove noise using a digital image filter and normalize the input image through white balancing, color temperature adjustment, angle correction, distortion correction, size correction, blur correction, etc. The image processing unit (230) can detect pedestrians and objects around the vehicle in real time using the YOLOv8-seg model with image information provided from the image preprocessing unit (220). Here, the YOLOv8-seg model can use Instance Segmentation by adding a Segmentation Branch structure to the algorithm. The above image processing unit (230) can detect objects such as pedestrians, cars, and crosswalks by analyzing real-time collected image information using a pedestrian dataset trained through the YOLOv8-seg model. The transmission unit (240) can transmit a warning signal in real time to a display or audio device inside the vehicle when a pedestrian on a crosswalk is detected. At this time, the transmission unit (140) can guide the driver to take an appropriate response by varying the warning level according to the distance to the detected pedestrian. Although embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improvements by those skilled in the art using the basic concept of the present invention as defined in the following claims also fall within the scope of the present invention.