Infrared small target detection (IRSTD) is a challenging task and has attracted increasing attention. Due to the small size of infrared small targets, the lack of fixed features, and complex backgrounds, the results detected by existing deep learning methods often lack clear contour and texture features. To resolve the above problem, we proposed a Multi-receptive Field Feature Fusion Network for Infrared Small Target Detection (MF-Net), which leverages multi-receptive fields and effective feature fusion to improve IRSTD performance. Specifically, it consists of the Multi-Branch Perception Enhancement (MBPE) module, Cascade Deep-Layer Semantic Localization (CDSL) module and Detail-Semantic Adaptive Fusion (DSAF) module. The MBPE module adopts a multi-branch feature extraction strategy, utilizing the characteristics of multi-receptive fields to extract detailed features of small targets such as edge contours and shapes, and enhance the contrast between targets and backgrounds. The CDSL module aims to capture and enhance high-level semantic information and global spatial information by aggregating key features under different receptive fields, so as to achieve precise positioning of small targets. The DSAF module promotes effective fusion of multi-scale feature maps through attention mechanisms, enabling interaction between shallow detail features and deep semantic features. Extensive experiments on the public real-world NUAA-SIRST dataset and IRSTD-1k dataset show that our MF-Net achieves almost the best results comparison with the state-of-the-art methods.