Abstract: Objectives： Emerging mass-market applications such as autonomous mobile robots, unmanned aerial vehicle delivery, and assisted driving require accurate and low-cost positioning. However, GNSS carrier-phase observations are easily affected by signal blockage and multipath effects in complex urban environments, leading to frequent cycle slips and degraded positioning performance. INS-aided cycle slip processing and dual-frequency combined observations can improve detection and repair performance. However, when low-cost MEMS-IMUs are used, rapid error accumulation during GNSS outages significantly limits their effectiveness, thereby necessitating assistance from other sensors. Vision sensors, as low-cost sensors widely used in assisted driving and mobile robots, can effectively suppress the error divergence of MEMS-IMUs, thereby further improving the performance of cycle slip processing. However, the performance gains brought by visual constraints to cycle slip detection and repair have not been systematically evaluated, and the repair performance differences among different dual-frequency combined observations have received limited attention. Therefore, this study proposes a vision-assisted low-cost GNSS/INS cycle slip detection and repair method to improve cycle slip processing reliability, quantify the performance gains brought by visual constraints, and systematically compare the detection and repair performance of different dual-frequency combined observations. Methods： A vision-assisted low-cost GNSS/INS cycle slip detection and repair method is proposed. This method uses visual data as EKF observations to impose multi-frame constraints on the IMU, thereby improving position prediction accuracy. The cycle slip detection term is obtained from double-differenced residuals computed using the predicted position. In addition to the single-frequency cycle slip detection term, a dual-frequency combined detection term with a longer equivalent wavelength is constructed to mitigate the adverse effects of accumulated prediction errors during prolonged periods of GNSS signal loss. The detection threshold is set by analyzing the composition and statistical characteristics of the residual error terms in the detection term. When a detection term exceeds the threshold, a cycle slip is declared. Subsequently, the single-frequency cycle slip values are determined using single- or dual-frequency detection terms, and the integer repair values are obtained after rounding. This framework enables real-time, single-epoch, multi-satellite cycle slip detection and repair under kinematic conditions. Results： Theoretical analysis shows that combined observations with longer equivalent wavelengths achieve better cycle slip detection performance under degraded positioning accuracy, whereas the uncombined observation provides more reliable repair performance. Simulation results demonstrate that visual assistance significantly improves cycle slip processing performance. Compared with the INS-only mode, the proposed method reduces the cycle slip detection and repair failure rates by 30% and 54%, respectively. Under degraded visual conditions, the cycle slip processing performance decreases, yet remains better than that of the INS-only mode. In GNSS outage simulations, visual assistance extends the detectable duration of a one-cycle slip to more than twice that of the INS-only mode on average, while combined observations with longer equivalent wavelengths further improve long-term detection capability. Field experimental results show that using the uncombined observation with visual assistance, the probability of maintaining a 3D positioning error below 1 m reaches 98.6%, representing improvements of 8.2% and 18.4% over the INS-only scheme and the traditional geometry-free and Melbourne-Wübbena method, respectively. In terms of positioning accuracy and repair reliability, the uncombined observation achieves better performance than the combined observation schemes. Conclusions： The proposed vision-assisted cycle slip detection and repair method can effectively improve cycle slip processing reliability and positioning performance for low-cost GNSS/INS systems in complex environments. Visual constraints effectively suppress MEMS-IMU error accumulation and significantly enhance cycle slip processing capability, even under degraded visual conditions. Combined observations with longer equivalent wavelengths are more advantageous for cycle slip detection under degraded positioning accuracy, whereas the uncombined observation provides higher repair reliability and better positioning performance. The results provide a reference for selecting cycle slip detection and repair strategies in low-cost multi-sensor integrated navigation systems.