Research On Algae Cell Image Detection System Based On FPGA

Algae detection is one of the important means of water quality monitoring.In order to realize real-time real-time detection of the status of algae,research and development of portable instant detection equipment has important academic and application value for water algae community monitoring,water pollution prevention and algae classification research.First,a lensless algae cell image acquisition system composed of CMOS image sensor,LED light source and microfluidic chip was built.By analyzing the characteristics of lensless algae cell image and image classification and recognition algorithm,the algae cell image classification and recognition algorithm based on convolutional neural network is determined.In order to quickly build a low-resolution algae cell recognition model,first use a high-resolution algae image collected by a microscope as a data set to initially build a classification network model.On this basis,by downsampling the data set,a data set of different resolutions is constructed,and the identification effect of different resolution data sets on algae cell images is compared and analyzed under the same network model,which is a low resolution network model Provide a reference for construction.Further,according to the characteristics of the algae cell image collected under the lens,adjust the parameters of each layer of the network so that it can recognize the image at low resolution,and finally use the model to train and test the four algae images collected by the system after removing the diffraction,The accuracy rate reached 98.85%.Quantify the weight of the determined network model to reduce the size of the network,and perform circuit mapping and optimization on the NVIDIA open source hardware acceleration platform according to its composition,making it suitable for the classification of algae cell images.On this basis,the SOC platform of the classification system was built,and the overall system design was implemented on the FPGA platform.Finally,the image acquisition processing and image classification system are integrated and the test system is used to test 200 images collected by the lensless system.The test results have an accuracy rate of 95%,reaching the expected design goal.