다중카메라 기반 파노라마 이미지와 거리측정에 의한 산불 감시 시스템 (Forest Fire Monitoring System using a Panoramic Image based on Multiple Cameras and Measuring of Distance)

서지정보

· 발행기관 : 한국조명.전기설비학회
· 수록지 정보 : 조명.전기설비학회논문지 / 32권 / 3호 / 44 ~ 52페이지
· 저자명 : 송현선, 이의용

초록

In case of forest fire, early detection of forest fire is the most important factor in minimizing the damages. In this paper, we suggest an effective system that detects forest fire using panoramic images from multiple cameras with PAN/TILT head. Where in the fire detection system comprises a Laser Range Finder (LRF) for calculating a distance to a location of break out a fire using a laser and a multiple cameras which have infrared camera and CCD camera for capturing each the image and transmitting the images to a control unit. It is possibile to detect the fire analyzing the shape pattern and color of infrared panoramic image. And it determines the location of break out of the fire using coordinates of an installation location of the infrared camera, a capturing angle received from the infrared camera and the distance received from the LRF. In addition, image matching and SURF were used to create the panoramic image. Image stitching for images acquired by multi-sensors, such as CCD and IR sensors consists of five steps. In the first step, we extract meaningful features using SURF algorithm. At the second step, we detect four correspondence points between two images using the proposed algorithm. Thirdly, warping operation which changes the input image into geometrically transformed image is done. In this thesis, we use perspective transformation using four correspondence points derived in the previous step. In the final step, we complete anoramic images using blending techniques which solves the problem of discrepancies in color and brightness of two neighboring images. This discrepancy problem is caused by various environmental factors such as sunlight, rotation of camera, or transformed image, etc.

영어초록

In case of forest fire, early detection of forest fire is the most important factor in minimizing the damages. In this paper, we suggest an effective system that detects forest fire using panoramic images from multiple cameras with PAN/TILT head. Where in the fire detection system comprises a Laser Range Finder (LRF) for calculating a distance to a location of break out a fire using a laser and a multiple cameras which have infrared camera and CCD camera for capturing each the image and transmitting the images to a control unit. It is possibile to detect the fire analyzing the shape pattern and color of infrared panoramic image. And it determines the location of break out of the fire using coordinates of an installation location of the infrared camera, a capturing angle received from the infrared camera and the distance received from the LRF. In addition, image matching and SURF were used to create the panoramic image. Image stitching for images acquired by multi-sensors, such as CCD and IR sensors consists of five steps. In the first step, we extract meaningful features using SURF algorithm. At the second step, we detect four correspondence points between two images using the proposed algorithm. Thirdly, warping operation which changes the input image into geometrically transformed image is done. In this thesis, we use perspective transformation using four correspondence points derived in the previous step. In the final step, we complete anoramic images using blending techniques which solves the problem of discrepancies in color and brightness of two neighboring images. This discrepancy problem is caused by various environmental factors such as sunlight, rotation of camera, or transformed image, etc.

참고자료

· 없음