长短期记忆网络的林火图像分割方法
|
摘要
长短期记忆网络(Long Short-Term Memory,LSTM)是当前深度学习的网络结构之一,针对林火图像,提出一种基于因子分析(FactorAnalysis,FA)与长短期记忆网络的深度学习图像分割方法。将RGB彩色图像转换成灰度图像,对灰度图像进行分块,同时将块变换成行向量,所有行向量组成矩阵并采用FA进行维数约减。最后采用LSTM对约减后的火焰和背景特征进行训练与测试并得到分类结果。试验结果表明,提出的算法的分割效果好,能很好地提取森林火焰,性能指标SD、Dice、ER、NR平均值分别为61.84%、76.42%、59.44%、1.41%。
Long Short-Term Memory(LSTM)was one of the current network structures of deep learning.A deep learning image segmentation method based on Factor Analysis(FA)and Long Short-Term Memory network is proposed for forest fire images.The RGB color image was transformed into gray image,and the gray image was divided into blocks.At the same time,the blocks were transformed into row vectors.All row vectors form a matrix and the dimension was reduced by FA.Finally,LSTM was used to train and test the flame and background features after dimension reduction,and the classification results are obtained.Experimental results showed that the proposed algorithm had good segmentation results and could well extract forest fires.The average value of SD,Dice,ER,NR were 61.84%,76.42%,59.44%,1.41%,respectively.
Long Short-Term Memory(LSTM)was one of the current network structures of deep learning.A deep learning image segmentation method based on Factor Analysis(FA)and Long Short-Term Memory network is proposed for forest fire images.The RGB color image was transformed into gray image,and the gray image was divided into blocks.At the same time,the blocks were transformed into row vectors.All row vectors form a matrix and the dimension was reduced by FA.Finally,LSTM was used to train and test the flame and background features after dimension reduction,and the classification results are obtained.Experimental results showed that the proposed algorithm had good segmentation results and could well extract forest fires.The average value of SD,Dice,ER,NR were 61.84%,76.42%,59.44%,1.41%,respectively.
引文
[1]杨绪兵,覃欣怡,张福全.基于样条的林火图像多阈值分割算法[J].计算机应用,2017,37(11):3157-3161,3167.Yang Xubing,Tan Xinyi,Zhang Fuquan.Multi-threshold segmentation algorithm for forest fire image based on spline[J].Computer application,2017,37(11):3157-3161,3167.
[2]吴兆明,郑嫦娥,上官晓锐.基于红外图像的林火识别方法及实现[J].中国农业资源与区划,2016,37(8):21-26,83.Wu Zhaoming,Zheng Change,Shangguan Xiaorui.Forest fire identification method based on infrared image and its realization[J].Agricultural Resources and Zoning,2016,37(8):21-26,83.
[3]王淑青,朱道利,潘健,等.一种改进的Otsu红外林火图像提取方法研究[J].激光杂志,2016,37(10):99-101.Wang Shuqing,Zhu Daoli,Pan Jian,et al.An improved Otsu infrared forest fire image extraction method[J].Laser Journal,2016,37(10):99-101.
[4]张庆杰,郑二功,徐亮,等.森林防火无人机系统设计与林火识别算法研究[J].电子测量技术,2017,40(1):145-150.Zhang Qingjie,Zheng Ergong,Xu Liang,et al.Design of forest fire prevention UAV system and research on forest fire recognition algorithm[J].Electronic Measurement Technology,2017,40(1):145-150.
[5]郭伟,昂海松,张思玉.基于图像检索的林火检测技术研究[J].林业科技通讯,2015(2):61-63.Guo Wei,Ang Haisong,Zhang Siyu.Research on forest fire detection technology based on image retrieval[J].Forestry Science and Technology Newsletter,2015(2):61-63.
[6]周辉.基于视频图像的林火烟雾检测方法[J].森林防火,2014(4):30-33.Zhou Hui.Forest fire smoke detection method based on video images[J].forest fire prevention,2014(4):30-33.
[7]闫建强.基于DM642和图像分析的林火烟雾检测系统研究[D].杭州:杭州电子科技大学,2011.Yan Jianqiang.Research on forest fire smoke detection system based on DM642and image analysis[D].Hangzhou:Hangzhou University of Electronic Science and Technology,2011.
[8]程玉柱,段一凡,李赵春.基于PSO与水平集的桂花图像分割方法[J].中国农机化学报,2018,39(8):81-84,91.Cheng Yuzhu,Duan Yifan,Li Zhaochun.Osmanthus fragrans image segmentation method based on PSO and level set[J].Journal of Chinese Agricultural Mechanization,2018,39(8):81-84,91.
[8]张云,许江淳,王志伟,等.基于机器视觉的棉花异性纤维检测技术优化研究[J].中国农机化学报,2018,39(9):61-65.Zhang Yun,Xu Jiangchun,Wang Zhiwei,et al.Optimization of cotton heterosexual detection technology based on machine vision[J].Journal of Chinese Agricultural Mechanization,2018,39(9):61-65.
[10]臧琛,尚士友,王志国,等.典型草原退化评价因子及其分级标准的研究[J].中国农机化学报,2016,37(10):210-213,245.Zang Chen,Shang Shiyou,Wang Zhiguo,et al.Study on degradation evaluation factors of typical steppe and its classification standard[J].Journal of Chinese Agricultural Mechanization,2016,37(10):210-213,245.
[11]束栋鑫,杨铭,梅从立,等.固态发酵过程电子鼻系统传感器阵列优化研究[J].中国农机化学报,2016,37(4):249-253.Shu Dongxin,Yang Ming,Mei Congli,et al.E-nose array optimization for monitoring straw solid-state fermentation[J].Journal of Chinese Agricultural Mechanization,2016,37(4):249-253.
[12]庄秀琴,吴爱兵,曹玉华.江苏沿海地区外向型经济发展水平测度[J].中国农机化学报,2016,37(2):276-279,284.Zhuang Xiuqin,Wu Aibing,Cao Yuhua.Measure of export-oriented economy development level in the coastal areas of Jiangsu[J].Journal of Chinese Agricultural Mechanization,2016,37(2):276-279,284.
[13]程玉柱,顾权,王众辉,等.基于深度学习的木材缺陷图像检测方法[J].林业机械与木工设备,2018,46(8):33-36.Cheng Yuzhu,Gu Quan,Wang Zhonghui,et al.Image detection of wood defects based on deep learning[J].Forestry Machinery and Woodworking Equipment,2018,46(8):33-36.
[14]Medina-Quero J,Zhang S,Nugent C,et al.Ensemble classifier of long short-term memory with fuzzy temporal windows on binary sensors for activity recognition[J].Expert Systems with Applications,2018,114:441-453.
[15]甘翔,高文华,张瑞秋.基于因子分析法的图标设计评价方法[J].图学学报,2018,39(4):706-710.Gan Xiang,Gao Wenhua,Zhang Ruichi.Evaluation meth-od of icon design based on factor analysis[J].Journal of Graphics,2018,39(4):706-710.
[16]孙奇平.基于深度学习的图像检索研究[J].景德镇学院学报,2018,33(3):15-18.Sun Qiping.Image retrieval based on deep learning[J].Journal of JingdezhenCollege,2018,33(3):15-18.
[2]吴兆明,郑嫦娥,上官晓锐.基于红外图像的林火识别方法及实现[J].中国农业资源与区划,2016,37(8):21-26,83.Wu Zhaoming,Zheng Change,Shangguan Xiaorui.Forest fire identification method based on infrared image and its realization[J].Agricultural Resources and Zoning,2016,37(8):21-26,83.
[3]王淑青,朱道利,潘健,等.一种改进的Otsu红外林火图像提取方法研究[J].激光杂志,2016,37(10):99-101.Wang Shuqing,Zhu Daoli,Pan Jian,et al.An improved Otsu infrared forest fire image extraction method[J].Laser Journal,2016,37(10):99-101.
[4]张庆杰,郑二功,徐亮,等.森林防火无人机系统设计与林火识别算法研究[J].电子测量技术,2017,40(1):145-150.Zhang Qingjie,Zheng Ergong,Xu Liang,et al.Design of forest fire prevention UAV system and research on forest fire recognition algorithm[J].Electronic Measurement Technology,2017,40(1):145-150.
[5]郭伟,昂海松,张思玉.基于图像检索的林火检测技术研究[J].林业科技通讯,2015(2):61-63.Guo Wei,Ang Haisong,Zhang Siyu.Research on forest fire detection technology based on image retrieval[J].Forestry Science and Technology Newsletter,2015(2):61-63.
[6]周辉.基于视频图像的林火烟雾检测方法[J].森林防火,2014(4):30-33.Zhou Hui.Forest fire smoke detection method based on video images[J].forest fire prevention,2014(4):30-33.
[7]闫建强.基于DM642和图像分析的林火烟雾检测系统研究[D].杭州:杭州电子科技大学,2011.Yan Jianqiang.Research on forest fire smoke detection system based on DM642and image analysis[D].Hangzhou:Hangzhou University of Electronic Science and Technology,2011.
[8]程玉柱,段一凡,李赵春.基于PSO与水平集的桂花图像分割方法[J].中国农机化学报,2018,39(8):81-84,91.Cheng Yuzhu,Duan Yifan,Li Zhaochun.Osmanthus fragrans image segmentation method based on PSO and level set[J].Journal of Chinese Agricultural Mechanization,2018,39(8):81-84,91.
[8]张云,许江淳,王志伟,等.基于机器视觉的棉花异性纤维检测技术优化研究[J].中国农机化学报,2018,39(9):61-65.Zhang Yun,Xu Jiangchun,Wang Zhiwei,et al.Optimization of cotton heterosexual detection technology based on machine vision[J].Journal of Chinese Agricultural Mechanization,2018,39(9):61-65.
[10]臧琛,尚士友,王志国,等.典型草原退化评价因子及其分级标准的研究[J].中国农机化学报,2016,37(10):210-213,245.Zang Chen,Shang Shiyou,Wang Zhiguo,et al.Study on degradation evaluation factors of typical steppe and its classification standard[J].Journal of Chinese Agricultural Mechanization,2016,37(10):210-213,245.
[11]束栋鑫,杨铭,梅从立,等.固态发酵过程电子鼻系统传感器阵列优化研究[J].中国农机化学报,2016,37(4):249-253.Shu Dongxin,Yang Ming,Mei Congli,et al.E-nose array optimization for monitoring straw solid-state fermentation[J].Journal of Chinese Agricultural Mechanization,2016,37(4):249-253.
[12]庄秀琴,吴爱兵,曹玉华.江苏沿海地区外向型经济发展水平测度[J].中国农机化学报,2016,37(2):276-279,284.Zhuang Xiuqin,Wu Aibing,Cao Yuhua.Measure of export-oriented economy development level in the coastal areas of Jiangsu[J].Journal of Chinese Agricultural Mechanization,2016,37(2):276-279,284.
[13]程玉柱,顾权,王众辉,等.基于深度学习的木材缺陷图像检测方法[J].林业机械与木工设备,2018,46(8):33-36.Cheng Yuzhu,Gu Quan,Wang Zhonghui,et al.Image detection of wood defects based on deep learning[J].Forestry Machinery and Woodworking Equipment,2018,46(8):33-36.
[14]Medina-Quero J,Zhang S,Nugent C,et al.Ensemble classifier of long short-term memory with fuzzy temporal windows on binary sensors for activity recognition[J].Expert Systems with Applications,2018,114:441-453.
[15]甘翔,高文华,张瑞秋.基于因子分析法的图标设计评价方法[J].图学学报,2018,39(4):706-710.Gan Xiang,Gao Wenhua,Zhang Ruichi.Evaluation meth-od of icon design based on factor analysis[J].Journal of Graphics,2018,39(4):706-710.
[16]孙奇平.基于深度学习的图像检索研究[J].景德镇学院学报,2018,33(3):15-18.Sun Qiping.Image retrieval based on deep learning[J].Journal of JingdezhenCollege,2018,33(3):15-18.