Single-photon LiDAR possesses weak echo sensing capability, but random backgrounds can easily cause target response loss and fragmented false alarms. In particular, sparse return photons, background photons, and dark counts may generate stochastic responses, making weak target preservation and false-alarm suppression difficult under single-photon observations. This paper proposes STMP-UNet, which uses a structure-aware target alignment module to suppress ineffective shallow response propagation through decoder semantic constraints, and combines a morphology prior module to support the spatial continuity of coarse target-region responses. The former selectively transfers target-related shallow information under high-level semantic guidance, while the latter introduces multi-scale structural constraints to alleviate spatial fragmentation of sparse target responses. Experiments are conducted on simulated single-photon data constructed from Anti-UAV infrared sequences. A photon-counting statistical simulation is used to reproduce sparse photon responses, and an infrared-derived proximity map provides auxiliary spatial structural information. Results show that compared to Pure U-Net, STMP-UNet improves Object F1 from 0.7115 to 0.7531 and Pd from 0.6635 to 0.7074, and reduces FPPI from 0.1907 to 0.1619. Ablation and candidate-level FROC analyses further demonstrate the effectiveness of the proposed structure modeling for preserving weak target responses and improving the detection–false-alarm trade-off.

To address the bottlenecks of low optical efficiency and uneven illumination inherent to single-photon array LiDAR systems, this paper proposes a joint imaging framework combining freeform light field shaping via an orthogonal bidirectional iterative algorithm and total-variation (TV) regularized image reconstruction. The core optical contribution lies in a freeform lens design strategy based on orthogonal bidirectional iteration, which establishes a precise energy mapping between incident Gaussian beams and uniform lattice light fields matching GM-APD pixel layouts. This refractive reshaping converts raw Gaussian radiation into flat-top spot arrays with homogenized energy distribution. Numerical simulations verify that the proposed freeform modulation achieves an optical energy utilization of 89.1% and a light uniformity of 96.5%, fundamentally resolving the tradeoff between central pixel saturation and peripheral detection blind zones in array receivers. For photon-sparse, noise-contaminated raw echo data, a reconstruction workflow integrating maximum likelihood estimation (MLE) and conventional TV regularization is adopted to suppress stochastic noise while preserving sharp object geometric edges. Nighttime field experiments targeting a scene at 22.4 m demonstrate that the mean absolute error (MAE) of depth reconstruction is reduced from 2.2373 m under unoptimized flood illumination to 0.3176 m with our freeform lattice illumination. The presented scheme fulfills the large field-of-view (FOV) requirement of array-based single-photon detection and yields remarkable imaging improvements under constraints of limited laser output power and compact optical packaging. This work delivers a feasible optical route for high-efficiency, long-range single-photon 3D imaging LiDAR.