Multi object personnel detection and dynamic tracking method based on improved KCF
-
摘要: 针对煤矿巷道光照不足、目标尺度变化剧烈、目标容易被遮挡和矿灯干扰等因素，导致对于井下的目标检测和跟踪存在成功率和准确度低的问题，提出一种基于改进核相关滤波（KCF）算法的多目标人员检测与动态跟踪方法，为避免井下复杂环境中由于光照不均引起检测失败，在改进的KCF算法中引入SSD检测算法，以提升对多目标人员检测能力。① 读取待跟踪视频序列，使用经过井下数据集训练后的SSD算法检测图像中的目标，若没有发现目标则继续读取下一帧。② 将检测到的目标放入跟踪器中，对图像进行预处理，通过比较将所有的检测框按照设定的阈值进行打分，并根据分值从高到低依次排列，高分的检测结果直接输出，低分的检测结果用于滤除不良信息，以提升检测速度。③ 通过KCF跟踪预测目标M帧后清空跟踪器，再重新进行目标检测。通过检测算法和跟踪算法的叠加，保证对目标的持续跟踪能力。实验结果表明：① 该方法最后的损失值稳定在1.675附近，检测结果较为稳定。② 经过训练后的SSD算法识别精度较训练前的SSD算法识别精度提高了52.7%。③ 该方法对矿井人员检测成功率、跟踪准确率分别为87.9%，88.9%，均高于其他4种算法（KCF、CSRT、TLD及MIL）的检测成功率、跟踪准确率。④ 该方法在重叠阈值较低时具有较高成功率，直至重叠阈值大于0.8时，成功率大幅下降，这是因为矿井中环境多样，想要完全符合标注的框有一定难度。实际应用结果表明：在井下煤矿巷道光照不足、目标尺度变化剧烈、容易被遮挡和受矿灯干扰等复杂环境中，该方法具有较高的适用性。Abstract: Factors such as insufficient illumination in coal mine roadways, drastic changes in object scale, easy obstruction of objects, and interference from mining lights lead to low success rate and accuracy in underground object detection and tracking. In order to solve the above problems, a multi object personnel detection and dynamic tracking method based on improved kernel correlation filter (KCF) algorithm is proposed. The method can avoid detection failure due to uneven lighting in complex underground environments. The SSD detection algorithm is introduced into the improved KCF algorithm to enhance the capability to detect multiple object personnel. ① The method reads the video sequence to be tracked, uses the SSD algorithm trained on the underground dataset to detect the object in the image. The method continues reading the next frame if no object is found. ② The method places the detected object into the tracker, preprocesses the image, scores all detection boxes according to the set threshold through comparison, and arranges them in descending order based on the score. The high score detection results are directly output, while the low score detection results are used to filter out bad information to improve detection speed. ③ The method clears the tracker after tracking and predicting object M frames through KCF, and then performs object detection again. By combining detection and tracking algorithms, the continuous tracking capability of the object is ensured. The experimental results show the following points. ① The final loss value of this method is stable around 1.675, and the detection results are relatively stable. ② The SSD recognition precision after training has improved by 52.7% compared to the SSD recognition precision before training. ③ The detection success rate and tracking accuracy of this method for mine personnel are 87.9% and 88.9%, respectively, which are higher than the detection success rate and tracking accuracy of the other four algorithms (KCF, CSRT, TLD, MIL). ④ This method has a high success rate when the overlap threshold is low, and until the overlap threshold is greater than 0.8, the success rate significantly decreases. This is because the environment in the mine is diverse, and it is difficult to fully match the labeled boxes. The practical application results show that this method has high applicability in complex environments such as insufficient lighting in underground coal mine roadways, drastic changes in object scale, easy obstruction, and interference from mining lights.
-
Key words:
- mine /
- multi object detection /
- object tracking /
- kernel correlation filter /
- SSD
-
图 11 矿灯直射时目标检测与跟踪准确度
Figure 11. The accuracy of object detection and tracking during direct mine light
表 1 训练前后算法性能对比
Table 1. Algorithm performance comparison before and after training
算法 识别精度/% 检测速度/（帧·s−1） 训练前SSD 32.6 10.91 训练后SSD 85.3 11.31 
下载: 导出CSV
表 2 5种算法性能对比
Table 2. Performance comparison of the 5 algorithms
算法 成功率/% 准确率/% 检测速度/（帧·s−1） KCF 42.6 41.1 38.49 CSRT 29.8 27.5 24.28 TLD 12.6 21.6 10.19 MIL 48.7 52.3 13.21 改进KCF算法 87.9 88.9 19.01
下载: 导出CSV
-
[1] 孙继平. 煤矿信息化与自动化发展趋势[J]. 工矿自动化,2015,41(4):1-5.SUN Jiping. Development trend of coal mine informatization and automation[J]. Industry and Mine Automation,2015,41(4):1-5. [2] 程德强,钱建生,郭星歌,等. 煤矿安全生产视频AI识别关键技术研究综述[J]. 煤炭科学技术,2023,51(2):349-365.CHENG Deqiang,QIAN Jiansheng,GUO Xingge,et al. Review on key technologies of AI recognition for videos in coal mine[J]. Coal Science and Technology,2023,51(2):349-365. [3] 孙继平. 煤矿监控新技术与新装备[J]. 工矿自动化,2015,41(1):1-5.SUN Jiping. New technologies and new equipments of coal mine monitoring[J]. Industry and Mine Automation,2015,41(1):1-5. [4] 孙继平. 煤矿智能化与矿用5G[J]. 工矿自动化,2020,46(8):1-7.SUN Jiping. Coal mine intelligence and mine-used 5G[J]. Industry and Mine Automation,2020,46(8):1-7. [5] 程德强,冯晨晨,唐世轩,等. 多特征融合的行人检测算法[J]. 煤炭技术,2018,37(10):254-257.CHENG Deqiang,FENG Chenchen,TANG Shixuan,et al. Multi-features fusion pedestrian detection algorithm[J]. Coal Technology,2018,37(10):254-257. [6] 孟庆勇. 5G技术在煤矿井下应用架构探讨[J]. 工矿自动化,2020,46(7):28-33.Meng Qingyong. Probe on 5G architecture applied in coal mine underground[J]. Industry and Mine Automation,2020,46(7):28-33. [7] ZHANG Tianzhu,XU Changsheng,YANG M. Multi-task correlation particle filter for robust object tracking[C]. IEEE Conference on Computer Vision and Pattern Recognition(CVPR),Honolulu,2017:4819-4827. [8] 刘海仓. 基于稀疏表示的图像超分辨率与目标跟踪方法研究[D]. 长沙:湖南大学,2015.LIU Haicang. Reseach on image super resolution and object tracking based on sparse representation[D]. Changsha:Hunan University,2015. [9] 白中浩,朱磊,李智强. 基于多模型融合和重新检测的高精度鲁棒目标跟踪[J]. 仪器仪表学报,2019,40(9):132-141.BAI Zhonghao,ZHU Lei,LI Zhiqiang. High-accuracy and robust object tracking based on multi-model fusion and re-detection[J]. Chinese Journal of Scientific Instrument,2019,40(9):132-141. [10] 郝建华. 基于CamShift和粒子滤波的煤矿变电所人员目标跟踪算法研究[J]. 工矿自动化,2015,41(11):35-388.HAO Jianhua. Research of personnel tracking algorithm for coal mine substation based on CamShift and particle filter[J]. Industry and Mine Automation,2015,41(11):35-38. [11] 邵小强,李康乐,陈熙,等. 基于改进卡尔曼滤波和参数拟合的矿井TOA定位方法[J]. 煤炭学报,2019,44(5):1616-1624.SHAO Xiaoqiang,LI Kangle,CHEN Xi,et al. TOA positioning method of coalmine based on Kalman filter and parameter fitting[J]. Journal of China Coal Society,2019,44(5):1616-1624. [12] 孙继平,邵子佩,刘毅. 采煤机视频压缩感知跟踪方法[J]. 工矿自动化,2018,44(3):8-11.SUN Jiping,SHAO Zipei,LIU Yi. Visual tracking method of shearer based on compressive sensing[J]. Industry and Mine Automation,2018,44(3):8-11. [13] 贾澎涛,贾伟. 煤矿井下视频多目标轨迹跟踪算法研究[J]. 计算机工程与应用,2018,54(2):222-227. doi: 10.3778/j.issn.1002-8331.1607-0308JIA Pengtao,JIA Wei. Recherche algorithm on coal mine multi-target trajectory tracking[J]. Computer Engineering and Applications,2018,54(2):222-227. doi: 10.3778/j.issn.1002-8331.1607-0308 [14] 陈伟,丁世飞,许新征. 基于YCbCr模型的巷道监控中矿工脸部图像识别[J]. 煤炭科学技术,2009,37(9):79-82,85.CHEN Wei,DING Shifei,XU Xinzheng. Miners' face images identification in monitor images of mine gateway based on YCbCr model[J]. Coal Science and Technology,2009,37(9):79-82,85. [15] HENRIQUES J F,CASEIRO R,MARTINS P,et al. High-speed tracking with kernelized correlation filters[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2015. DOI: 10.1109/TPAMI.2014.2345390. [16] LIU Wei,ANGUELOV D,ERHAN D,et al. SSD:single shot multibox detector[J]. Springer,Cham,2016. DOI: 10.1007/978-3-319-46448-0_2. [17] KALAKE L,WAN Wanggen,HOU Li. Analysis based on recent deep learning approaches applied in real-time multi-object tracking:a review[J]. IEEE Access,2021(9):32650-32671. doi: 10.1109/ACCESS.2021.3060821 [18] LUKEZIC A,VOJIR T,CEHOVIN Z L,et al. Discriminative correlation filter with channel and spatial reliability[J]. International Journal of Computer Vision,2018. DOI: 10.1007/s11263-017-1061-3. [19] ZHAO Shuaitong,YANG Xianzhao,CHEN Yang,et al. Research on pedestrian detection algorithms combined with lightweight networks[C]. International Conference on Algorithms,High Performance Computing and Artificial Intelligence,Guangzhou,2022:62-67. [20] CHEPLYGINA V,SRENSEN L,TAX D M,et al. Label stability in multiple instance learning[C]. Medical Image Computing and Computer-Assisted Intervention,Springer,Cham,2015. DOI: 10.1007/978-3-319-24553-9_66. -
点击查看大图
图(11) / 表(2)
计量
- 文章访问数: 971
- HTML全文浏览量: 55
- PDF下载量: 44
- 被引次数: 0
