Research on Robot Image Recognition Based on Deep Learning

In recent years, the integration of image processing and machine vision technologies to achieve automation and intelligence has become a major focus of research. Intelligent management not only reduces labor and intensity but also enhances the efficiency of planting, harvesting, and picking crops and fruits. Initially, data is preprocessed by checking the dataset, removing erroneous labels and low-quality images, adjusting image size, transforming color space, denoising, and normalizing to enhance feature extraction and increase dataset size and diversity. This paper utilizes the YOLOV8 model to identify apples in images, obtaining bounding boxes and counting them. This count represents the number of apples. Simultaneously, a histogram is created to display the distribution of apple numbers based on these counts. Subsequently, the image dataset is read, images are displayed, and preprocessing and feature extraction functions are defined. The YOLOV8 model is loaded, position transformations are performed, and color and texture features are integrated. Finally, maturity is assessed, statistical analysis is conducted, and results are displayed. Lastly, the image dataset is read and an image is displayed. The paper defines preprocessing functions for binarization and morphological processing, contour extraction functions for apple detection and fitting, and functions for estimating apple area and quality. Statistical analysis and presentation of results are then performed.