This study presents an effective approach to crop classification by integrating multi-temporal Harmonized Landsat Sentinel-2 (HLS) imagery with machine learning models. Focusing on the MahaLakshumi-Kheda region in India, the MahaLakshumi-Kheda region in India serves as the study area, where multi-temporal Harmonized Landsat Sentinel-2 (HLS) imagery is utilized to analyze crop dynamics. MahaLakshumi-Kheda, located in the Gangapur taluka of Chhatrapati Sambhajinagar district, spans 527.65 hectares of fertile land with diverse agricultural practices, including cotton, sweet lime, and sugarcane cultivation. Our approach involves collecting ground truth data from the location and using Support Vector Machine and K-Nearest Neighbors classifiers to identify crucial crops like cotton, sugarcane, and sweet lime across different seasonal seasons. According to this study, the classification accuracy of crop varieties such as cotton, sugarcane, and sweet lime were 95%, 94%, & 83% respectively, which was pretty good for their overall class. By using advanced machine learning methods, this study can categorize crops and provide more efficient monitoring of agriculture. The KNN model achieved an overall classification accuracy of 93%, with 95% accuracy for cotton and 94% accuracy for sugarcane, superior to the SVM. This study uses satellite data and lightweight machine learning algorithms to monitor crops at high rates, providing scalability in diverse agricultural environments.