Non-line-of-sight (NLOS) pedestrian detection is vital for uncovering hidden pedestrians and is crucial for autonomous driving and emergency response. Traditional methods depend on high-speed lasers and precise photon detection devices, which are costly and complex. Aiming to streamline equipment and expand practicality, we explored using infrared cameras with object detection techniques on relay surfaces to detect hidden pedestrians. Despite the better reflectivity of infrared signals, such signals in the relay surface are too diffuse and vague for conventional pedestrian detection algorithms. To this end, our work introduces the Non-Lin-of-Sight Detection Net (NLOS-DetNet), which consists of a Hidden Object Enhancement Module (HOEM) to bolster relay surface features and a Hidden Object Detection Module (HODM), incorporating a Scattered Object Attention Module (SOAM) to utilize these enhanced signals for hidden pedestrian detection. With a custom long-wave infrared dataset for NLOS detection, our model pinpoints obscured pedestrians effectively and showcases the benefits of long-wave infrared for NLOS detection tasks.