Observing the dynamics of regional development is a part of good governance in regional planning, to know what to preserve as a conservation area and where to optimize the potential of the development area. Historical analysis of land cover becomes an important component to observe that dynamic. This identification can be achieved by analyzing historical time-series data from remote sensing satellite imagery. Google Earth Engine (GEE) is an integrated cloud computing-based platform, capable of performing image analysis using a wide range of geospatial data, both in terms of imagery type and in terms of temporal time-series geospatial data, including Landsat-5/7/8, and Sentinel-2. This study aims to test the quality of land cover segmentation in GEE platform, observing the accuracy of pixel-based method compared to object-based image analysis (OBIA) method, in rural and semi-urban areas of Indonesia for 34 years (1988-2022), both using random forest algorithm to differentiate built-up areas, rice fields, unplanted rice fields, open land, shrubs and sparse vegetation, high density vegetation, and bodies of water from satellite imagery. Both methods resulting good information accuracy above 85% for Sentinel-2 Imagery, and OBIA is better in terms of segmentation shape and form and could differentiate unplanted rice field to open land better, but with caveat it needs a better training data especially in texture and shape of land cover for Simple Non-Iterative Clustering (SNIC) image segmentation. For 34 years Pelabuhanratu as the regional capital of Sukabumi Regency has experienced development, but it is not significant, some notable land cover change is the emergence of power plants, the addition of several residential development blocks along the coast and the expansion of rice fields and some deforestation.