We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Efficient vehicle detection for traffic video-based intelligent transportation systems applications using recurrent architecture.
- Authors
Sankaranarayanan, Manipriya; Mala, C.; Mathew, Samson
- Abstract
The growth of vehicular population increases the need for an efficient methods for managing and regulating traffic. The effective and commonly existing solution in recent times is Intelligent Transportation Systems (ITS) that implements video-based traffic monitoring with computer vision techniques. The major challenge during such process is to achieve higher accuracy in vehicle detection with reduced computational cost. This paper proposes a solution to overcome the challenge using a Recurrent Architecture with Parallel Vehicle Detection Scheme (RAP-VDS) for a selected region of the frame of traffic video. There are two modules in the proposed RAP-VDS. Using an improved matching method, the Spatial Color Reducing over Recurrent Temporal Information (SCR2TI) module detects and removes the redundant frame. By processing the color components in parallel to increase detection accuracy, the Multi-level Parallelism with Spatial Color Information Processing (MSCIP) (MSCIP) module incorporates the spatial colour information of the non-redundant frames acquired through SCR2TI. The novel and innovative feature of RAP-VDS is that it can be incorporated by any video-based vehicle detection process. Also, RAP-VDS avoids the recurrent processing of content at the same time improves the detection without supressing the spatial characterstics of the traffic video. Hence, the proposed work offers an appropriate balance with detection and computational cost. The improvements in detections and computational costs are analysed with benchmark traffic videos usign conventional background subtraction techniques. From the results it is also seen that, the overall vehicle detection accuracy is 81.6%. The overall processing time was reduced by 37.5%. The major challenges to encounter the detection using the proposed RAP-VDS were for videos during poor illumination conditions. However, the results of the proposed work can also be improved further by using other improved video based detection technniques.
- Publication
Multimedia Tools & Applications, 2023, Vol 82, Issue 25, p39015
- ISSN
1380-7501
- Publication type
Article
- DOI
10.1007/s11042-023-14812-4