A Parallel Framework for Object Detection and Recognition for Secure Vehicle Parking

Image analytics, biometrics access control, security, and surveillance applications utilize complex machine learningand computer vision algorithms, such as face detection andrecognition. Speedup and accuracy are two important factorsthat need to be addressed in all such complex applications. Parallel computing breaks down the complex tasks into discretefragments to be solved concurrently on multiple processors. Theparallel computing procedure significantly reduces the executiontime with improved speedup. This paper presents a parallelframework for object detection and recognition for a securevehicle parking. The proposed framework is divided in to threesteps: (1) vehicle detection at the parking entry junction, (2)driver's face detection, and (3) identification of driver's face fromthe huge database of stored facial images. On successfulidentification of authorized person, vehicle is allowed to enter inthe parking-lot. The adaptive boosting algorithm is used forvehicle and face detection, while Eigenfaces based approach isemployed for face recognition. Moreover, the scalabilitycomparison of parallely executed driver face recognitionalgorithm indicates high speedup compared to serial execution. Furthermore, the results of the proposed framework revealpromising performance and encourage outcomes to be deployedin real-time at entrance/exits of the public/private vehicle parking areas.