Dynamic Obstacle Detection of Road Scenes
Today, many automobile companies and researchers have developed various safety systems to reduce fatalities by traffic accidents[1, 2]. In order to prevent traffic accidents by distracted driving, therefore, the proposed system presents the vehicle and pedestrian detection using a novel image representation called equi-height mosaicking system. Furthermore, the proposed system additionally suggests the partially visible vehicle detection method using equi-height peripheral mosaicking image while driving.
Dynamic Vehicle and Pedestrian Detection using Equi-Height Mosaicking Image
We present a real-time vehicle and pedestrian detection approach using a novel image representation for efficient processing called an equi-height mosaicking image (EHMI). The proposed approach uses a GPU for the real-time processing. The EHMI improves the execution time of the conventional approach without decreasing the detection accuracy. The EHMI is generated as follows. After a geometric analysis of a road scene, we crop a set of image strips by sampling several positions on the road at regular intervals. The height of each image strip is computed by projecting the predefined height of a vehicle at a distant position onto an image plane. After all the cropped images are resized to the uniform height required to build equi-height images (EHIs), we concatenate EHIs similar to a panorama image, to create the EHMI. The concatenated image has a long width but the height of the image is equivalent. The proposed approach then performs a GPU-based vehicle and pedestrian detection on the concatenated image using a 1D search based SVM classification. The proposed method has faster speed and higher accuracy than the GPU-based OpenCV HOG detector.
Flowchart of the Proposed System
Demo Video
Detection of Partially Visible Vehicle using Equi-Height Peripheral Mosaicking Image
We present a part-based detection algorithm for detecting partially visible vehicles using equi-height peripheral mosaicking image (EHPMI). A partially visible vehicle is a vehicle that quickly overtakes or suddenly cuts-in from the side of the driver’s vehicle. In order to detect this vehicle efficiently, the proposed system first acquires regions from both sides of the drivers vehicle using the image captured from a front camera. Then, it performs image warping to express the side of the vehicle and concatenates this warped image to generate a panorama-like image called EHPMI. EHPMI improves the detection performance of partially visible vehicle detection algorithm. Finally, the system performs GPU-accelerated detection using the front and rear sides of vehicle and transforms coordinates of detected vehicles to the original image. After the coordinates transform, the proposed system obtains the exact position of the detected vehicles on the road scene.
Flowchart of the Proposed System
Demo Video
Publication
- Min Woo Park, Soon Ki Jung, Dynamic Obstacle Detection of Road Scenes using Equi-Height Mosaicking Image, Electronics Letters on Computer Vision and Image Analysis, Vol.13, No.2, pp.13-14, ISSN.15775097, 2014.
- Hye Sun Park, Kwang Hee Won, Min Woo Park, Kyong Ho Kim, Soon Ki Jung, In-Vehicle AR-HUD System to Provide Driving-Safety Information, ETRI Journal , Vol.35, No.6, pp.1038-1047, ISSN.12256463, 2013.
- Min Woo Park, Soon Ki Jung, GPU-Based Real-Time Pedestrian Detection and Tracking Using Equi-Height Mosaicking Image, Lecture Notes in Computer Science, Vol.8228, pp.409-416, ISSN.03029743, 2013.
- Min Woo Park, Jung Pil Park, Soon Ki Jung, Real-time Vehicle Detection using Equi-Height Mosaicking Image, International Conference on Research in Adaptive and Convergent Systems (RACS 2013), pp.171-176, 2013.
- 박민우, 원광희, 정순기, 빌보드 스윕 스테레오 시차정합 알고리즘을 이용한 차량 검출 및 추적, 한국멀티미디어학회논문지, Vol.16, No.6, pp.758-775, ISSN.12297771, 2013.