基于边缘多通道梯度模型的多运动目标检测 下载: 857次
Multi-Moving Object Detection Based on Edge Multi-Channel Gradient Model
上海工程技术大学电子电气工程学院, 上海 201620
图 & 表
图 1. 目标检测的结果。 (a)形态学处理的结果; (b)目标块
Fig. 1. Result of object detection. (a) Result of morphological processing; (b) object block
下载图片 查看原文
图 2. 不同算法的检测结果。 (a)原始图像; (b)文献[
18]中的算法; (c)文献[
19]中的算法; (d)文献[
20]中的算法; (e)本算法
Fig. 2. Detection results of different algorithms. (a) Original image; (b) algorithm of Ref. [18]; (c) algorithm of Ref. [19]; (d) algorithm of Ref. [20]; (e) our algorithm
下载图片 查看原文
图 3. 不同算法的检测误差率
Fig. 3. Detection error rates of different algorithms
下载图片 查看原文
表 1视频序列参数
Table1. Parameter of the video sequence
Number | Video sequence | Frame | Size |
---|
1 | PETS2006 | 1200 | 720×576 | 2 | blizzard | 7000 | 720×480 | 3 | David | 770 | 320×240 |
|
查看原文
表 2不同算法的检测性能
Table2. Detection performance of different algorithms
Algorithm | Video1 | Video2 | Video3 |
---|
R/% | P/% | XF_M/% | FPS/frame | R/% | P/% | XF_M/% | FPS/frame | R/% | P/% | XF_M/% | FPS/frame |
---|
Ref.[18] | 87.3 | 90.6 | 88.9 | 2 | 79.8 | 82.1 | 80.9 | 3 | 76.9 | 70.5 | 73.6 | 14 | Ref.[19] | 98.8 | 78.9 | 87.7 | 1 | 71.5 | 85.8 | 78.0 | 5 | 78.2 | 69.9 | 73.8 | 11 | Ref. [20] | 77.8 | 85.7 | 81.6 | 4 | 79.7 | 74.8 | 77.2 | 7 | 70.5 | 73.8 | 72.1 | 13 | Ours | 93.3 | 88.1 | 90.6 | 26 | 87.4 | 83.5 | 85.4 | 32 | 83.5 | 79.8 | 81.6 | 43 |
|
查看原文
表 3各阶段计算和存储的数据量
Table3. Amount of data calculated and store in each stage
Memory consuming and computational demanding stage | Data store |
---|
Establishment of the E-McGM | (N-B)×nx×ny×nCF×nTF×nSF×mθ | Computing ATA | (N-B)×nx×ny×mθ×6 | Velocity responses in m directions around a point | (N-B)×nx×ny×mθ×4 | Computing velocity magnitude and direction | (N-Bl)×nx×ny |
|
查看原文
陈婕妤, 奚峥皓, 卢俊鑫. 基于边缘多通道梯度模型的多运动目标检测[J]. 激光与光电子学进展, 2021, 58(4): 0415002. Jieyu Chen, Zhenghao Xi, Junxin Lu. Multi-Moving Object Detection Based on Edge Multi-Channel Gradient Model[J]. Laser & Optoelectronics Progress, 2021, 58(4): 0415002.