Page 251 - 应用声学2019年第4期
P. 251
第 38 卷 第 4 期 荆丹翔等: 基于成像声呐 DIDSON 的水域内鱼群数量估计方法 711
置于水下0.3 m 处时,共会产生 0.72 m 的探测盲区。 fish[J]. Canadian Journal of Fisheries & Aquatic Sciences,
其次,当水深超过声呐探测极限距离时,水底也存在 1992, 49(10): 2179–2189.
[5] Belcher E O, Matsuyama G, Trimble R. Object identifi-
探测盲区。同时,鱼类对于调查船具有自动躲避特
cation with acoustic lenses[C]//Proceeding IEEE Oceans,
性 [18] ,因此通过走航采集的水下数据并不是鱼群均 IEEE, Honolulu, Hawai, 2001(1): 6–11.
匀分布下的状态。针对本试验中的声呐,鱼类体长 [6] 陈星辰, 陈斌. 双频识别声呐水下影像监测系统及其应用 [J].
大于 5 cm、小于 80 cm 时计数准确率更高,因为体 中国水能及电气化, 2015(11): 22–25, 11.
Chen Xingchen, Chen Bin. Dual-frequency identification
长过小的鱼成像在声呐图像上容易被当成噪声滤
sonar underwater image monitoring system and its appli-
除,体长过大的鱼在视场近端时,成像容易被割断造 cation[J]. China Water Power & Electrification, 2015(11):
成计数错误。 22–25, 11.
[7] Tiffan K F, Rondorf D W, Skalicky J J. Imaging fall
本论文利用声呐采集鱼群信息,通过数据后处
chinook salmon redds in the columbia river with a dual-
理估算了整片水域中的鱼群数量,对于目标跟踪及 frequency identification sonar[J]. North American Journal
计数的准确性尚缺乏更客观的评价,因此后续可以 of Fisheries Management, 2004, 24(4): 1421–1426.
设计一个基于高清水下摄像头观测族馆鱼类的试 [8] Han J, Honda N, Asada A, et al. Automated acous-
tic method for counting and sizing farmed fish during
验,进行对比跟踪与计数分析。
transfer using DIDSON[J]. Fisheries Science, 2009, 75(6):
1359–1367.
4 结论 [9] Zhang H, Wei Q, Kang M. Measurement of swimming pat-
tern and body length of cultured Chinese sturgeon by use
(1) 本文提出了一种基于成像声呐的鱼群数量 of imaging sonar[J]. Aquaculture, 2014, 434: 184–187.
估计方法,通过声呐图像处理与多目标跟踪计数算 [10] 荆丹翔. 基于识别声呐的鱼群目标检测跟踪方法 [D]. 杭州:
浙江大学, 2018.
法统计走航探测中的目标个数,计算目标数量的平
[11] Chaillan F, Fraschini C, Courmontagne P. Speckle noise
均面密度,从而获得整片水域的鱼群数量。 reduction in SAS imagery[J]. Signal Processing, 2007,
(2) 为了去除声呐系统特有的斑点噪声,设计 87(4): 762–781.
了固定数据窗口的迭代最小二乘算法。为了有效提 [12] Padmavathi G, Subashini P, Kumar M, et al. Compari-
son of filters used for underwater image pre-processing[J].
取复杂背景中的目标,设计了基于三倍标准差准则
International Journal of Computer Science & Network Se-
的自适应阈值分割算法。 curity, 2010, 10(1): 58–65.
(3) 通过对滴水湖的走航探测,估算出了整片 [13] Lee K J, Sung H, Park E, et al. Joint optimization for one
水域的鱼群数量,并选取两段数据与人工计数对比, and two-way MIMO AF multiple-relay systems[J]. IEEE
Transactions on Wireless Communications, 2010, 9(12):
结果显示目标统计偏差在 10% 左右,相对于回波探 3671–3681.
测法精度得到了很大提高。 [14] Ding F, Xiao Y. A finite-data-window least squares al-
gorithm with a forgetting factor for dynamical mod-
eling[J]. Applied Mathematics & Computation, 2007,
参 考 文 献 186(1): 184–192.
[15] Cho H, Yu S C. Real-time sonar image enhancement for
[1] Lubis M Z, Manik H M. Review: acoustic systems (split AUV-based acoustic vision[J]. Ocean Engineering, 2015,
beam echo sounder) to determine abundance of fish in ma- 104: 568–579.
rine fisheries[J]. Journal of Geoscience, Engineering, En- [16] 何世彪, 杨士中. 3σ 准则在小波消噪中的应用 [J]. 重庆大学
vironment, and Technology, 2017, 2(1): 76–83. 学报 (自然科学版), 2002, 25(12): 58–61.
[2] Misund O A, Aglen A, Frønæs E. Mapping the shape, He Shibiao, Yang Shizhong. Applied of 3σ-rule in re-
size, and density of fish schools by echo integration and a ducing noise in signal by wavelet analysis[J]. Journal of
high-resolution sonar[J]. ICES Journal of Marine Science, Chongqing University (Natural Science Edition), 2002,
1995, 52(1): 11–20. 25(12): 58–61.
[3] Zwolinski J Z, Fernandes P G F G, Marques V, et al. Es- [17] Jing D, Han J, Wang G, et al. Dense multiple-target
timating fish abundance from acoustic surveys: calculat- tracking based on dual frequency identification sonar
ing variance due to acoustic backscatter and length distri- (DIDSON) image[C]//Oceans IEEE, Shanghai, 2016:
bution error[J]. Canadian Journal of Fisheries & Aquatic 1–5.
Sciences, 2009, 66(12): 2081–2095. [18] Jůza T, Rakowitz G, Drastik V, et al. Avoidance reac-
[4] Appenzeller A R, Leggett W C. Bias in hydroacoustic tions of fish in the trawl mouth opening in a shallow and
estimates of fish abundance due to acoustic shadowing: turbid lake at night[J]. Fisheries Research, 2013, 147(10):
evidence from day–night surveys of vertically migrating 154–160.