Many studies have underscored the impact of depositions, soiling, and dust accumulation on the performance of photovoltaic (PV) systems, resulting in energy output reduction. Recent research studies have explored new techniques to detect soiling and dust on photovoltaic panels, integrating image processing, machine learning, and thermal imaging. This paper introduces a Real-time embedded system based on Neural Network Algorithms and image processing techniques to recognize soiling and dust depositions on PV panels. The system architecture integrates a Raspberry Pi 4 and an HD camera for PV panel image capture and acquisition. The proposed research outlines a comprehensive approach that combines Neural Network Algorithms (NNA), image processing techniques, and hardware platforms to build a high-performance, real-time embedded system for solar panels soiling and dust recognition and help to make intelligent maintenance decisions for optimal power production. According to the results, the proposed system achieves an accuracy of over 85%. The proposed framework offers a trade-off between speed and accuracy, making them suitable for Real-time processing applications compared to other object detection algorithms.