To address the decline in target detection accuracy caused by blur and color deviation in underwater images and to improve the accuracy and robustness of squid detection in complex underwater environments， this study proposes an underwater squid detection algorithm named DUIE-YOLO based on image enhancement. The algorithm adopts a cascaded framework of “enhance first， detect later” consisting of the DUIE-Net enhancement module and the YOLOv8-HD detection module. The DUIE-Net module significantly improves image quality through color correction， multi-scale feature fusion， feature restoration and enhancement， and dehazing optimization. The YOLOv8-HD detection module combines the FasterNet network， a small-object detection head， the CoordAttention mechanism， and the ShapeIoU loss function to optimize feature extraction and small-object detection accuracy. Experimental results show that DUIE-YOLO outperforms the original YOLOv8n in four key metrics： Precision， Recall， F1-score， and mAP， with improvements of 4.2%， 6.8%， 5.7% and 5.5%， respectively. Joint experiments demonstrate that the combination of DUIE-Net and YOLOv8-HD achieves a 40.3% increase in mAP， a 10.5% increase in Precision， a 53% increase in Recall， and a 31% increase in F1-score compared to the baseline （Raw+YOLOv8n）， proving the algorithm's significant cascaded optimization effect. The study indicates that DUIE-YOLO effectively mitigates the performance degradation caused by poor underwater image quality through the synergistic optimization of image enhancement and detection modules. This research provides a high-precision solution for target recognition in complex underwater environments， offering significant application value for marine biological monitoring and resource development.