Abstract: Image with insufficient illumination often suffers from severe degradations like unexpected noise, absence of natural colors, and low contrast. These drawbacks not only result in unpleasant visual effect, but also strongly affect high-level tasks such as object detection. In order to better complete the low illumination object detection task, we propose a low illuminance object detection algorithm that combines feature enhancement with multi-scale receptive field (FEMR). First, the pixel-level high-order mapping (PHM) module learns the high-order mapping relationship from low illuminance to normal illuminance, and then improves the significance of low illuminance object features, so as to obtain preliminary enhanced feature information. Then, the key information enhancement (KIE) module combines multiple attention mechanisms to highlight important features and filters noise information to obtain further enhanced feature information. In addition, the long distance feature capture (LFC) module introduces strip receptive fields of various scales to capture the long distance relations of isolated regions in low illumination scenes. Final, extensive experiments are conducted on the ExDark dataset, and the experimental results show that the proposed algorithm has better performance than other detection algorithms in low illuminance object detection accuracy, and the proposed modules have good versatility. All in all, the proposed algorithm can complete end-to-end low-illuminance object detection and output images with normal illuminance style, which is convenient for human eyes to directly evaluate the accuracy of detection results.