<p>In object detection tasks, detecting small objects presents numerous challenges, including unclear positions in the image, blurry or indistinct features, and vulnerability to complex backgrounds. Commonly, existing object detectors extract features through multiple downsampling, which may result in the loss or significant attenuation of feature information for small objects, thus affecting detection accuracy and stability. This paper proposes an innovative feature enhancement method called FReID (Feature ReIntegration and Distribution) to address the challenge of small-sized object recognition in object detection. FReID incorporates a feature reintegration module designed to fully utilize multi-level features by aligning channels based on the features at different levels, thereby enhancing the retention of small-object features. Additionally, we introduce a lightweight feature distribution module based on cross-attention, which fuses the re-integrated features with those from lower downsampling layers to improve the retention of complementary information. Our experimental results on the AI-TOD and VisDrone-2019 datasets show that our method surpasses current two-stage and one-stage object detection methods with similar parameter quantities and scales. Specifically, in the improved method based on YOLOv8, the mAP@0.5:0.95 increased by up to 5.6% across the two datasets. These findings indicate the significant potential and wide applicability of our approach in enhancing detection performance.</p>