Design And Implementation Of A Dynamic Scheduling Method Based On CnC In Vehicle Recognition

Worldwide, the rapid growth in number of vehicles increases the frequency of traffic accidents, which has attracted serious concern over vehicle safety, while the study of Driver Assistance System (DAS) has been attracting more and more attention both home and abroad. Since the location and recognition of moving vehicles under road environment is one of the most important and fundamental parts in DAS, it is necessary and of great significance to make research on its parallelism in order to improve the operational efficiencyThis paper is based on feature-fusion daytime vehicle recognition algorithm to extract the various features of a vehicle, by which to comprehensively recognize each vehicle among images. However, there are still some problems with implementation of current algorithm, relatively low operating efficiency being the top. Meanwhile, the rapid development of multi-core platform enables a more enhanced operational efficiency. But the vehicle recognition technology itself is one of the complicated feature applications, having the computing features such as input dependency, which needs runtime dynamic information and cannot be obtained only with the static and single parallel model under multi-core platform, causing the separateness between the upper algorithm and the low-level system, and the efficiency cannot be improved sufficiently; especially in the vehicle recognition algorithm which has a high demand for hard real-time, the efficiency improvement is quite important.Therefore, this paper takes CnC parallel model as abstract expression to obtain runtime information, proposes and implements a corresponding dynamic scheduling method between the upper vehicle recognition algorithm and underlying thread scheduling, optimizing the threads allocation to improve the efficiency. The desired goal has been achieved after experiment, and the research on parallelism for DAS can be promoted by applying in practice.