Autonomous underwater vehicles (AUVs) equipped with online visual inspection systems can detect underwater targets during underwater operation, which is of great significance to subsea exploration. However, the undersea scene has some instinctive challenging problems, such as poor lighting conditions, sediment burial, and marine biofouling mimicry, which makes it difficult for traditional target detection algorithms to achieve online, reliable and accurate detection of underwater targets. To solve the above issues, this paper proposes a real-time object detection algorithm for underwater targets based on a lightweight convolutional neural network model. To improve the imaging quality of underwater images, the CLAHE model with fused multicolor space is designed in this paper to enhance the image quality of underwater targets. Afterwards, a spindle-shaped backbone network is designed. The inverted residual block and group convolution are used to extract depth features to ensure the target detection accuracy on the one hand and to reduce the model parameter volume on the other hand under complex scenarios. Through extensive verification, the precision, recall and mAP of the proposed algorithm reach 91.2%, 90.1% and 88.3%, respectively. It is also noticeable that this method has been integrated on the embedded GPU platform and deployed to the AUV system in the real test scenario. The average computational time is 0.053 s, which satisfies the requirements of real-time object detection.