Design and Implementation of Around View Monitor System with NEON Technology

• Main Authors: LEE, CHUNG-JAN, 李中然
• Other Authors: SU, CHING-LUNG
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/gpu8ab

碩士 === 國立雲林科技大學 === 電子工程系 === 105 === Nowadays, there are traffic accidents which caused by blind spot frequently. Therefore, around view monitor (AVM) system solves the problem of blind spot when current automotive electronic industry progressing. AVM system enables drivers park on the roadside, merge on the narrow road, and avoid collison easily with the limited condition of view from A column or rearview mirror. AVM system which this paper proposes employs digital embedded development board and the fisheye camera with HD resolution (1280 × 720). Moreover, the system installs four fisheye cameras around the experimental vehicle, inputs the matching coordinates manually by features of correction cloth, and makes correcting mode synthesize AVM image via fisheye calibration and homography matrix. Then, the system establishes Look Up Table (LUT) for application of executing mode. The execution mode applies LUT to convert the read image data into AVM screen directly and optimizes AVM screen through color adjustment, linear interpolation, color mixing and other image processing. Therefore, the brightness of AVM screen reaches consistence, and definition can be smooth and seamless. HD resolution quality requires too much operand on the embedded development board and the system consequently demands to deal with image information of four HD resolution cameras. If the system computes four image information with NV12 format (4×1280×720×1.5) as 30 FPS per second, the system has to operate 158MB per second which is 3 times more than SD resolution (720 × 480). In order to improve the situation of software development having no more optimization, this paper employs NEON hardware that only ARM possesses. NEON hardware is applicable to operation of image processing. We applies NEON technology to optimize development of algorithms, and employs Pthreads to implement multi-core processing. Optimization results increase FPS of original system 1.6 times, and reduce 40% computing time approximately which improve system operation effectively.