Intelligent Extraction Method of Land Type Patches Using AI Human-Computer Collaborative Technology

Classical techniques lack enough accuracy when it comes to handling complex scenes of land use patterns, while also suffering from problems like data loss and information overlap. This paper attempts to propose the AI human-computer collaborative technology-based intelligent extraction method of land type patches in order to attain higher accuracy and efficiency in extraction. First, the remote sensing image data is preprocessed, including denoising, color balancing, histogram equalization, etc., to improve the image quality. Then, CNN is used to extract features from the preprocessed image, and multi-layer convolution and pooling operations are performed to extract spatial features. The training set is expanded again using data enhancement technology. Then, based on the extracted features, a human-computer interactive feedback mechanism is introduced to allow users to adjust and optimize the model results during the extraction process to ensure the purity of the extraction. Finally, the extracted features are post-processed in combination with the segmentation algorithm (U-Net) for the creation of the final land type map. This method yields a land type extraction accuracy of 98%. AI human-machine collaboration has great potential in the intelligent extraction of land type maps. It can completely eliminate the drawbacks of traditional methods and provide more reliable data support for land management.