Parallel Design And Implementation Of Cell Image Recognition Based On TMS320C6678

With the development of computer technology and signal processing techniques,based on digital image analysis technologies automatic test equipments emerge,which promotes the development of intelligent clinical testing, improving test speedand making test results more scientific. This thesis mainly studies automatic sedimentanalyzer, based on researching the principle of cell automatic classification counts,congregating high-performance multi-core DSP technology, developing study onembedded hardware software platform design technology and multi-core parallelprocessing technology.This thesis selects TI Keystone architecture, high-performance multi-core DSPprocessors TMS320C6678as the core of cell image processing hardware platform,designed embedded hardware platforms, including chip peripheral design, GigabitEthernet interface design, expanded memory designs. Combined hardware platformdissertates software programming optimization key technologies.It analyses the characteristics of cell image processing time consumption andproposes a multi-core pipeline parallel algorithm based on feedback inter-corecommunications and queque inter-core storage, adopting feature calculation andobject recognition in the same pipeline level parallelism design and solving featurecalculation and object recognition on inter-core communications consumption andstorage constrained problems. The design is based on multi-core DSP C6678cellsimage processing pipeline parallelism methods to improve the multi-core workingefficiency.Finally, study the principle of DSP SYSBIOS programming mechanism. Usingmulti-core multi-threading, IPC inter-core synchronization, communications, andembedded TCP/IP communications technology realizes embedded multi-core cellimage processing software, building a prototype. Based prototype, we designcommunications protocol and communications data format. The experiment resultsshow that this design method is feasible, effective, reliable. What’s more,TMS320C6678-based microscope cell recognition system real-time processingcapabilities greatly improve.