激光与光电子学进展, 2019, 56 (13): 131009, 网络出版: 2019-07-11 复制并引用该论文  本文已被引 6 次  被引通知

图 & 表

图 1. 毫米波成像系统示意图

Fig. 1. Diagram of millimeter wave imaging system

下载图片 查看原文

图 2. 毫米波图像与光学图像对比。(a)毫米波图像,图中方框为真实标记;(b)光学图像,图中方框为检测结果,数字为目标置信度

Fig. 2. Comparison between millimeter wave images and optical images. (a) Millimeter wave images, boxes in figure are real markers; (b) optical images, boxes in figure are the test results, and the numbers are the target confidence

下载图片 查看原文

图 3. 反卷积示意图

Fig. 3. Deconvolution diagram

下载图片 查看原文

图 4. 两个数据集中标记框的面积统计图。(a)毫米波图像数据集的统计图;(b) VOC数据集的统计图

Fig. 4. Area statistics of label boxes in two data sets. (a) Statistical chart of millimeter wave image data sets; (b) statistical chart of VOC data sets

下载图片 查看原文

图 5. 部分训练结果。(a1)(b1)(c1)(d1)真实标记;(a2)(b2)(c2)(d2)对应的检测结果

Fig. 5. Partial training results. (a1)(b1)(c1)(d1) real marks; (a2)(b2)(c2)(d2) test results of corresponding graph

下载图片 查看原文

表 1特征提取网络结构

Table1. Feature extraction network structure

Type Layers Filters Size/stride Output Convolution1MaxPool 2 64 3×3/12×2/2 (512,256)(256,128) Convolution2MaxPool 2 128 3×3/12×2/2 (256,128)(128,64) Convolution3MaxPool 3 256 3×3/12×2/2 (128,64)(64,32) Convolution4MaxPool 3 512 3×3/12×2/2 (64,32)(32,16) Convolution5 3 512 3×3/1 (32,16) Deconvolution 1 512 2×2/2 (64,32) Concateconvolution5 (64,32) Convolution6 2 512 3×3/1 (64,32)

查看原文

表 2网络测试GPU环境为GTX 1080时的实验结果对比

Table2. Comparison of experimental results when the GPU environment for network testing is GTX 1080

Model Input size /(pixel×pixel) Sampling Layers Without RCNN pre rec F1 score Time /ms VGG-1 1000×500 16 16 × 0.859 0.744 0.797 107 VGG-2 1000×500 16 16 √ 0.864 0.785 0.822 97 ZF-Net 1000×500 16 5 √ 0.837 0.757 0.794 27 ResNet 1000×500 16 41 √ 0.867 0.766 0.813 66 DenseNet 1000×500 16 90 √ 0.850 0.540 0.660 94 SSD 300×300 8 16 √ 0.854 0.763 0.805 44 YOLOv3 416×416 8 59 √ 0.822 0.784 0.802 29 Ours 512×256 8 19 √ 0.876 0.812 0.843 36

查看原文

侯冰基, 杨明辉, 孙晓玮. 基于改进Faster RCNN的毫米波图像实时目标检测[J]. 激光与光电子学进展, 2019, 56(13): 131009. Bingji Hou, Minghui Yang, Xiaowei Sun. Real-Time Object Detection for Millimeter-Wave Images Based on Improved Faster Regions with Convolutional Neural Networks[J]. Laser & Optoelectronics Progress, 2019, 56(13): 131009.