This study presents a remote intelligent perception system for multi-object detection, which integrates semantic segmentation, feature extraction, and deep learning techniques to enhance scene recognition. The system uses UNet for semantic segmentation, discrete wavelet transform (DWT), Sobel, Laplacian, and local binary pattern (LBP) for feature extraction, and a deep belief network (DBN) for object recognition. The system then employs AlexNet to assign scene labels based on the recognized objects. The proposed system was validated using three standard datasets: PASCALVOC-12, Cityscapes, and Caltech 101, achieving high accuracy on all datasets. The system demonstrates strong performance in scene recognition, with accuracy exceeding 96% on PASCALVOC-12 and 95.90% on Cityscapes. The model also achieves 92.2% accuracy on Caltech 101, showing significant improvements over existing models. The system's architecture includes preprocessing, semantic segmentation, feature extraction, object recognition, and scene labeling. The study also discusses the challenges of scene recognition, including semantic understanding, occlusion, illumination, and object size variations. The proposed system addresses these challenges through a combination of advanced image processing techniques and deep learning models. The system's performance is evaluated on multiple datasets, demonstrating its effectiveness in real-world scenarios. The study highlights the importance of scene recognition in various applications, including autonomous driving, healthcare, and environmental monitoring. The proposed system offers a robust and efficient solution for multi-object detection and scene recognition, with potential applications in a wide range of fields.This study presents a remote intelligent perception system for multi-object detection, which integrates semantic segmentation, feature extraction, and deep learning techniques to enhance scene recognition. The system uses UNet for semantic segmentation, discrete wavelet transform (DWT), Sobel, Laplacian, and local binary pattern (LBP) for feature extraction, and a deep belief network (DBN) for object recognition. The system then employs AlexNet to assign scene labels based on the recognized objects. The proposed system was validated using three standard datasets: PASCALVOC-12, Cityscapes, and Caltech 101, achieving high accuracy on all datasets. The system demonstrates strong performance in scene recognition, with accuracy exceeding 96% on PASCALVOC-12 and 95.90% on Cityscapes. The model also achieves 92.2% accuracy on Caltech 101, showing significant improvements over existing models. The system's architecture includes preprocessing, semantic segmentation, feature extraction, object recognition, and scene labeling. The study also discusses the challenges of scene recognition, including semantic understanding, occlusion, illumination, and object size variations. The proposed system addresses these challenges through a combination of advanced image processing techniques and deep learning models. The system's performance is evaluated on multiple datasets, demonstrating its effectiveness in real-world scenarios. The study highlights the importance of scene recognition in various applications, including autonomous driving, healthcare, and environmental monitoring. The proposed system offers a robust and efficient solution for multi-object detection and scene recognition, with potential applications in a wide range of fields.