Flame detection is a research hotspot in industrial production,and it has been widely used in various fields.Based on the ignition and combustion video sequence,this paper aims to improve the accuracy and unintuitive ...Flame detection is a research hotspot in industrial production,and it has been widely used in various fields.Based on the ignition and combustion video sequence,this paper aims to improve the accuracy and unintuitive detection results of the current flame detection methods of gasifier and industrial boiler.A furnace flame detection model based on support vector machine convolutional neural network(SCNN)is proposed.This algorithm uses the advantages of neural networks in the field of image classification to process flame burning video sequences which needs detailed analysis.Firstly,the support vector machine(SVM)with better small sample classification effect is used to replace the Softmax classification layer of the convolutional neural network(CNN)network.Secondly,a Dropout layer is introduced to improve the generalization ability of the network.Subsequently,the area,frequency and other important parameters of the flame image are analyzed and processed.Eventually,the experimental results show that the flame detection model designed in this paper is more accurate than the CNN model,and the accuracy of the judgment on the flame data set collected in the gasifier furnace reaches 99.53%.After several ignition tests,the furnace flame of the gasifier can be detected in real time.展开更多
Neuromorphic computing is considered to be the future of machine learning,and it provides a new way of cognitive computing.Inspired by the excellent performance of spiking neural networks(SNNs)on the fields of low-pow...Neuromorphic computing is considered to be the future of machine learning,and it provides a new way of cognitive computing.Inspired by the excellent performance of spiking neural networks(SNNs)on the fields of low-power consumption and parallel computing,many groups tried to simulate the SNN with the hardware platform.However,the efficiency of training SNNs with neuromorphic algorithms is not ideal enough.Facing this,Michael et al.proposed a method which can solve the problem with the help of DNN(deep neural network).With this method,we can easily convert a well-trained DNN into an SCNN(spiking convolutional neural network).So far,there is a little of work focusing on the hardware accelerating of SCNN.The motivation of this paper is to design an SNN processor to accelerate SNN inference for SNNs obtained by this DNN-to-SNN method.We propose SIES(Spiking Neural Network Inference Engine for SCNN Accelerating).It uses a systolic array to accomplish the task of membrane potential increments computation.It integrates an optional hardware module of max-pooling to reduce additional data moving between the host and the SIES.We also design a hardware data setup mechanism for the convolutional layer on the SIES with which we can minimize the time of input spikes preparing.We implement the SIES on FPGA XCVU440.The number of neurons it supports is up to 4000 while the synapses are 256000.The SIES can run with the working frequency of 200 MHz,and its peak performance is 1.5625 TOPS.展开更多
目的激光雷达采集的室外场景点云数据规模庞大且包含丰富的空间结构细节信息,但是目前多数点云分割方法并不能很好地平衡结构细节信息的提取和计算量之间的关系。一些方法将点云变换到多视图或体素化网格等稠密表示形式进行处理,虽然极...目的激光雷达采集的室外场景点云数据规模庞大且包含丰富的空间结构细节信息,但是目前多数点云分割方法并不能很好地平衡结构细节信息的提取和计算量之间的关系。一些方法将点云变换到多视图或体素化网格等稠密表示形式进行处理,虽然极大地减少了计算量,但却忽略了由激光雷达成像特点以及点云变换引起的信息丢失和遮挡问题,导致分割性能降低,尤其是在小样本数据以及行人和骑行者等小物体场景中。针对投影过程中的空间细节信息丢失问题,根据人类观察机制提出了一种场景视点偏移方法,以改善三维(3D)激光雷达点云分割结果。方法利用球面投影将3D点云转换为2维(2D)球面正视图(spherical front view,SFV)。水平移动SFV的原始视点以生成多视点序列,解决点云变换引起的信息丢失和遮挡的问题。考虑到多视图序列中的冗余,利用卷积神经网络(convolutional neural networks,CNN)构建场景视点偏移预测模块来预测最佳场景视点偏移。结果添加场景视点偏移模块后,在小样本数据集中,行人和骑行者分割结果改善相对明显,行人和骑行者(不同偏移距离下)的交叉比相较于原方法最高提升6.5%和15.5%。添加场景视点偏移模块和偏移预测模块后,各类别的交叉比提高1.6%Institute)上与其他算法相比,行人和骑行者的分割结果取得了较大提升,其中行人交叉比最高提升9.1%。结论本文提出的结合人类观察机制和激光雷达点云成像特点的场景视点偏移与偏移预测方法易于适配不同的点云分割方法,使得点云分割结果更加准确。展开更多
基金Supported by Shaanxi Province Key Research and Development Project(No.2021GY-280)Shaanxi Province Natural Science Basic ResearchProgram Project(No.2021JM-459)National Natural Science Foundation of China(No.61834005,61772417,61802304,61602377,61634004)。
文摘Flame detection is a research hotspot in industrial production,and it has been widely used in various fields.Based on the ignition and combustion video sequence,this paper aims to improve the accuracy and unintuitive detection results of the current flame detection methods of gasifier and industrial boiler.A furnace flame detection model based on support vector machine convolutional neural network(SCNN)is proposed.This algorithm uses the advantages of neural networks in the field of image classification to process flame burning video sequences which needs detailed analysis.Firstly,the support vector machine(SVM)with better small sample classification effect is used to replace the Softmax classification layer of the convolutional neural network(CNN)network.Secondly,a Dropout layer is introduced to improve the generalization ability of the network.Subsequently,the area,frequency and other important parameters of the flame image are analyzed and processed.Eventually,the experimental results show that the flame detection model designed in this paper is more accurate than the CNN model,and the accuracy of the judgment on the flame data set collected in the gasifier furnace reaches 99.53%.After several ignition tests,the furnace flame of the gasifier can be detected in real time.
基金The work was supported by the HeGaoJi Program of China under Grant Nos.2017ZX01028103-002 and 2017ZX01038104-002the National Natural Science Foundation of China under Grant No.61472432.
文摘Neuromorphic computing is considered to be the future of machine learning,and it provides a new way of cognitive computing.Inspired by the excellent performance of spiking neural networks(SNNs)on the fields of low-power consumption and parallel computing,many groups tried to simulate the SNN with the hardware platform.However,the efficiency of training SNNs with neuromorphic algorithms is not ideal enough.Facing this,Michael et al.proposed a method which can solve the problem with the help of DNN(deep neural network).With this method,we can easily convert a well-trained DNN into an SCNN(spiking convolutional neural network).So far,there is a little of work focusing on the hardware accelerating of SCNN.The motivation of this paper is to design an SNN processor to accelerate SNN inference for SNNs obtained by this DNN-to-SNN method.We propose SIES(Spiking Neural Network Inference Engine for SCNN Accelerating).It uses a systolic array to accomplish the task of membrane potential increments computation.It integrates an optional hardware module of max-pooling to reduce additional data moving between the host and the SIES.We also design a hardware data setup mechanism for the convolutional layer on the SIES with which we can minimize the time of input spikes preparing.We implement the SIES on FPGA XCVU440.The number of neurons it supports is up to 4000 while the synapses are 256000.The SIES can run with the working frequency of 200 MHz,and its peak performance is 1.5625 TOPS.
文摘目的激光雷达采集的室外场景点云数据规模庞大且包含丰富的空间结构细节信息,但是目前多数点云分割方法并不能很好地平衡结构细节信息的提取和计算量之间的关系。一些方法将点云变换到多视图或体素化网格等稠密表示形式进行处理,虽然极大地减少了计算量,但却忽略了由激光雷达成像特点以及点云变换引起的信息丢失和遮挡问题,导致分割性能降低,尤其是在小样本数据以及行人和骑行者等小物体场景中。针对投影过程中的空间细节信息丢失问题,根据人类观察机制提出了一种场景视点偏移方法,以改善三维(3D)激光雷达点云分割结果。方法利用球面投影将3D点云转换为2维(2D)球面正视图(spherical front view,SFV)。水平移动SFV的原始视点以生成多视点序列,解决点云变换引起的信息丢失和遮挡的问题。考虑到多视图序列中的冗余,利用卷积神经网络(convolutional neural networks,CNN)构建场景视点偏移预测模块来预测最佳场景视点偏移。结果添加场景视点偏移模块后,在小样本数据集中,行人和骑行者分割结果改善相对明显,行人和骑行者(不同偏移距离下)的交叉比相较于原方法最高提升6.5%和15.5%。添加场景视点偏移模块和偏移预测模块后,各类别的交叉比提高1.6%Institute)上与其他算法相比,行人和骑行者的分割结果取得了较大提升,其中行人交叉比最高提升9.1%。结论本文提出的结合人类观察机制和激光雷达点云成像特点的场景视点偏移与偏移预测方法易于适配不同的点云分割方法,使得点云分割结果更加准确。
