China’s BeiDou navigation satellite system (BDS) has finished global constellation construction and can achieve joint solution, simultaneously relying on the B1I + B3I signals of the BDS-2 and BDS-3 satellites. For reasons mostly related to chip shape distortions, navigation satellite observations are corrupted by receiver-dependent code biases. Those biases are brought into observation residuals and degrade the pseudorange correction accuracy. Herein, we present a code bias estimation algorithm, using what we found to be an obvious clustering code bias phenomenon between the BDS-2 and BDS-3 satellites, leading to the systematic biases existing in the BDS-2+3 joint solution. Therefore, we propose a BDS-2+3 joint solution with code bias self-calibration, which can accurately strip off clustering code biases between the BDS-2 and BDS-3 satellites, and can greatly improve precise point positioning (PPP) convergence speed and accuracy. The statistics showed that the residual biases and root mean square (RMS) improved by 36% and 15% and the convergence time improved by approximately 35%. In the convergence stage, the positioning accuracy improved by approximately 38% and 21% in the horizontal and vertical directions, respectively. Meanwhile, in the post-convergence stage, the accuracy improved by approximately 10%.

Geometric accuracy assessment is a fundamental prerequisite for the application of high-resolution optical satellite imagery, yet it remains particularly challenging in the absence of ground control points. This study proposes a consistency-driven accuracy assessment framework for ZY-3 satellite imagery that systematically bridges the conventional separation between block adjustment and accuracy evaluation. The framework is built upon a unified geometric error model that accommodates both inter-image consistency assessment and absolute accuracy evaluation within a common mathematical formulation. Recognizing the inherent uncertainties of publicly available reference data, a hierarchical validation chain is established: the Google Earth (GE) and Shuttle Radar Topography Mission (SRTM) datasets are first validated against WorldView imagery to bound their intrinsic errors, and subsequently employed as references for ZY-3 accuracy assessment. This design enables the simultaneous evaluation of planimetric accuracy, vertical accuracy, and inter-image consistency without reliance on ground control points. Experimental validation conducted on 1368 ZY-3 scenes covering a study area of 1.9 million km² yields a planimetric RMSE of 3.33 m and a vertical RMSE of 4.27 m. The results demonstrate that higher inter-image consistency empirically correlates with improved absolute positioning accuracy, and that the proposed framework not only reliably evaluates geometric quality but also provides diagnostic insights that can inform proactive adjustment strategies. The proposed method thus offers a practical and transparent solution for the geometric accuracy detection and enhancement of large-area ZY-3 satellite imagery.