Multi-core Platform Of The Vehicle Recognize Algorithm Based On CnC Parallel Optimization

As the rapid growth in the number of vehicles, traffic accidents frequently occur, people are highly concerned about vehicle safety, intelligent transportation system functions are become more and more complex, the calculate data we face also become more complex and enormous, and for higher real-time intelligent traffic system, nothing but the two methods to solve this problem:First, upgrade or add hardware; Second, optimization software (algorithms). Obviously additional hardware is bound to increase cost, this paper chooses the second approach, using the vehicle recognize algorithm module of the intelligent transportation systems as a research entry point, under the conditions of the existing multi-core platform, through the parallel optimization algorithm to solve the huge amounts of data, prepare for the future direction of the parallel optimization in embedded platform.Firstly, this paper analyzes and extracts the serial implementation process of vehicle recognize algorithm based on statistical characteristics of vehicle, vehicle recognition algorithm mainly consists of two processing steps:candidate regions identified (hypothesis generation. HG) phase and vehicle verification (hypothesis verification, HV) phase. Candidate regions identified phase is based on vehicle prior knowledge; vehicle certification phase is based knowledge and statistical pattern recognition-based method to verify the hypothesis of the previous phase. In the analysis of extraction algorithms based on the serial algorithm, this paper proposed a variety of parallel optimization program and uses the Intel CnC (Intel Concurrent Collections for C++) parallel language to optimize the algorithm based on statistical characteristics of vehicle recognize. Overall, this paper can be divided into three main parts:analyzes and extracts process of the serial algorithm; analyzes the algorithm part that can be parallel, and proposes a variety of parallel optimization program according to the analysis; under the optimization program, this paper uses Intel’s CnC parallel language to reconstruct the vehicle recognize module, and tests a variety of parallel size to verify the optimized algorithm.This paper has tested a variety of distinctive features of the picture on multiple hardware platforms, experimental results show that based the Multi-core platform, optimized algorithm reached the theoretical expectations, which to some extent enreached corresponding theory in the practical application also can promote the car driver assistance systems parallel study.