IP geolocation is vital for location-aware Internet applications, relying on high-quality network landmarks for accuracy. Most research has focused on IPv4 networks, neglecting the need for dual-type network geolocation. Current IPv6 landmarks are often limited, scarce, and at outdated risk. This paper presents a dual-stack street-level landmark mining framework called DualS-Geo, which leverages the structure of IPv6 addresses that embed IPv4 addresses, achieving the collection of numerous dual-stack nodes. DualS-Geo then geolocates the IPv4 addresses of these nodes, ultimately generating a large scale of IPv4-IPv6 dual-stack street-level landmarks. The framework successfully mined 1,110,034 dual-stack landmarks, increasing the number of IPv6 landmarks by approximately 3.69 times compared to the existing method. The landmarks are spread across 148 countries, 1,016 cities, and 1,345 autonomous systems. By utilizing these dual-stack landmarks, the accuracy of IPv6 geolocation in target cities improved by an average of 38.86% over the existing method. DualS-Geo enhances IPv6 geolocation services by mining a vast range of dual-stack landmarks while also increasing the pool of IPv4 street-level landmarks, improving IP geolocation in dual-type networks.