Method and apparatus for compensating vibration of deep-learning based washing machine

The patent creates an optical copy of vibration information by capturing images of the washing tub's movement through a camera. Instead of directly measuring vibration with sensors, the system captures visual copies of the tub's position and movement patterns, then analyzes these images to infer vibration characteristics. This approach replaces physical sensors with optical observation and image processing.

The patent replaces the mechanical sensing system (vibration sensors) with an optical system (camera). The camera captures images of the washing tub, and deep learning algorithms process these images to predict vibration values. This substitution eliminates the need for separate vibration sensors while maintaining measurement capability through image-based analysis.

The patent makes the camera serve multiple functions: it captures images for vibration analysis, tracks washing tub position, and provides visual monitoring capabilities. By making the camera multi-functional, the system achieves reliable vibration prediction without adding dedicated vibration sensors, thereby simplifying manufacturing while maintaining reliability.

The system uses the camera's own imaging capability to serve the vibration measurement function. The camera captures images that are then processed by deep learning models to predict vibration values, allowing the existing imaging system to serve dual purposes without requiring additional specialized sensors or components.

The patent performs preliminary actions by pre-training deep learning models with labeled vibration data and image data during the manufacturing or setup phase. The models are pre-learned to recognize vibration patterns from images, so during actual operation, they can quickly predict vibration values without requiring complex real-time computation. This preliminary training reduces computational complexity during runtime.

The system uses a simplified approach by focusing on specific image features that are most relevant to vibration detection, rather than processing all possible image data. The deep learning model is trained to extract only the necessary features from images for vibration prediction, reducing computational complexity while maintaining prediction precision.