Pomacea is an invasive species of major concern in China, which can have a negative impact on crop growth and the ecological environment. The timely acquisition of information on the distribution of eggs of Pomacea can help to prevent and control its invasion in advance. In order to improve the extraction of feature information from the Pomacea eggs in complex natural environments, this paper suggests a YOLOv5sCBAM model for eggs of Pomacea recognition based on the YOLOv5s base network model and incorporating the CBAM (Convolutional Block Attention Module) attention mechanism module. The experimental results show that the incorporation of the CBAM module provides better recognition than the introduction of the CA and SE attention modules. The YOLOv5sCBAM model, which incorporates CBAM, has a stronger recognition impact than the original YOLOv5s model and can, in some cases, overcome interference from elements like plant obscuration and reflection in the water. A 2.5 percentage point improvement over the original model, the mAP now stands at 83.8%. The method based on deep learning is feasible to identify the eggs of Pomacea in photographs obtained in complicated natural environment, which provides  a fresh perspectives for the monitoring, prevention, and management of invasive species such as  Pomacea.

福寿螺是我国重点关注的入侵物种，对农作物生长和生态环境会造成不利影响。及时获取福寿螺虫卵的分布信息，对于提前防治其入侵能起到有效的帮助作用。基于YOLOv5s基础网络模型，引入CBAM（Convolutional Block Attention Module）注意力机制模块，以提高在复杂的自然环境下对福寿螺虫卵特征信息的提取，提出YOLOv5sCBAM模型进行福寿螺虫卵识别方法。试验结果表明，引入CBAM的识别效果要好于引入CA和SE注意力模块。同时，引入CBAM的YOLOv5sCBAM模型，识别效果优于原基础YOLOv5s模型，一定程度上能够克服倒影、植物遮挡等因素干扰。且平均精度均值达到83.8%，相比原模型提升2.5个百分点。基于深度学习的方法对复杂自然环境中的福寿螺虫卵进行识别是切实可行的，为福寿螺等入侵物种的监测防控提供新的思路。