Abstract: To accurately evaluate the resources of white shrimp（Litopenaeus vannamei）in the aquaculture water, a method for detecting and identifying and evaluating the resources of White Shrimp using Dual-Frequency Identification Sonar (DIDSON) is researched and developed. The experiment was conducted in a breeding pond (Length 170 m ⅹ Width 60 m) in Shanghai. The acoustic data of White Shrimp was collected by using an unmanned ship equipped with acoustic cameras to carry out navigation measurements in the pond, a target recognition, and counting model were constructed for its acoustic imaging to carry out the automatic tracking and counting work of White Shrimp images, and the accuracy of the recognition and counting of the model was verified by combining the fishing data and manual visual counting. The results showed that the fishery resource assessment results of White Shrimp in the pond showed that the number of large-sized individuals (body length> 10 cm) accounted for 78.2%; the average density of the water surface was 7.39 ind/m2; the total number was 75369 ind. The deviation between the artificial visual individual count of the acoustic image of white shrimp and the automatic counting using the target recognition counting model is about 8.47%; the total mass in the pond is estimated to be about 1415 kg from the relationship between the body length and weight of the White Shrimp, which is different from previous fishing. According to statistics, the body weight deviation of White Shrimp is about 8.63%. The DIDSON of this study is suitable for the evaluation of shrimp resources in ponds and behavior observation. The evaluation results provide an indicator for the management of shrimp resources in the ponds. At the same time, the evaluation research methods can be extended to the data acquisition and analysis of shrimp aquaculture management.

Key words: DIDSON, white shrimp, resource assessment；acoustic image processing, target identification 

摘要： 为准确评估养殖水域内南美白对虾资源量，研发一种利用水中声学摄像机（Dual-Frequency Identification Sonar， DIDSON），进行养殖池塘识别并评估南美白对虾（Litopenaeus vannamei）资源量的方法，并在上海某养殖池塘（长170 m  宽60 m）进行了试验。利用无人船搭载DIDSON在池塘内进行走航测量，采集南美白对虾声学数据，针对其声学成像构建目标识别计数模型，进行了南美白对虾影像的自动跟踪计数工作,并结合捕捞数据和人工目视计数验证了模型识别计数的准确性。结果显示：池塘内南美白对虾渔业资源量评估结果为，大规格个体（体长 > 10 cm）数量占比为78.2 %；水面的平均密度为7.39 ind/m2；总数量为75369 ind。南美白对虾的声学影像人工目视个体计数与使用目标识别计数模型自动计数的偏差约为8.47 %；由南美白对虾体长与体质量关系式推定池塘内总体质量约为1 415 kg，与历次捕捞统计南美白对虾体质量相比偏差约为8.63 %。本研究的水中声学摄像机适用于池塘内虾类资源量评估和行为观测，评估结果对池塘内虾类资源管理提供指示作用，同时评估研究方法可推广到虾类养殖管理的数据获取与分析中。

关键词: 声学摄像机, 南美白对虾, 资源量评估, 声学影像处理, 目标识别