Multi-sensors integration is widely used to achieve different mapping and navigation objective. Among them, vision sensor could be recognized as one of the common sensors which contain large scene and feature information. Although this type of sensor has significant advantages in mapping and navigation, there still exist big challenges such as how to detect and remove mismatches among the image dense matching and improve the reliability of the integration result. This research aims to develop a quality control procedure to eliminate mismatches or outliers in the image dense matching procedure, and to evaluate the reliability and separability of these potential mismatches or outliers, which are measured with the Minimum Detectable Bias (MDB) and the Minimum Separable Bias (MSB), respectively. The experiments have shown that when the number of images is increased, the MDB and MSB will significantly decrease, which means the reliability and separability will be improved. The numerical results from some case studies are discussed in detail.