Multi-core Parallel Processing Theory And Method For Rape Leaves Image Analysis

With the rapid development of information technology,image processing has played an important role in agricultural fields,such as plant leaf image analysis,seed identification,vegetable pesticide detection and so on.However,these processing technologies,such as image feature extraction,image compression and coding,cost much time and space overhead.When performing quality inspection for a large number of agricultural products using computer image processing technology,there is also the problem that the traditional single processor processing speed cannot meet the time requirement.As one of the most important physiological organs of most crops,leaf features information can directly reflect the crop’s nutrition and pest status.In this paper,we designs multifractal feature extraction serial scheme based on C/C++ language,combining the features of multi core processor and analyzing the large calculation feature for the DF_MFA method of rapeseed leaf image.Using the instrumentation technique to analyze the computational characteristics of each part,we design an image multifractal feature extraction OpenMP parallel optimization scheme based on a homogenous multi-core processor.Taking speedup as the performance evaluation index,the performance of the parallel scheme is verified for the three task scheduling of OpenMP.The experimental results show that the parallel scheme designed in this paper has better acceleration effect,and for the irregular loop structure,the dynamic scheduling method has the best acceleration effect.Finally,this paper further studies the widely used shared cache architecture multi-core processors.Multiple threads run on multiple processing cores at the same time.The cache resource competition problem often occurs when multiple running threads access the shared cache.At present,most classical scheduling algorithms ignore the problem of cache competition.Therefore we propose a shared cache aware multi-core processor task scheduling strategy(SCAS)in this paper.Firstly,we divided the multi-core processor Cache into blocks by using the Cache space partitioning technology.Assuming that each task requires a fixed number of Cache blocks,and one Cache block is occupied by one executing task at any time,so the Cache space of the runtime task is independent.Then,we presented a task scheduling model and scheduling problem based on shared cache block constraints in multi-core systems.The simulation results show that the algorithm SCAS performs best compared with the traditional Min-Min and MSCAS algorithm adjusted based on SCAS.