63results about How to "Improve Change Detection Accuracy" patented technology

An image change detecting system may include an image processor cooperating with the database for generating a reference geospatial image corresponding to the collected geospatial image, an image enhancer for enhancing at least one of the reference geospatial image and the collected geospatial image based upon at least one environmental condition, and a change detector cooperating with the image processor and the image enhancer. The change detector may detect a change between the collected geospatial image and the reference geospatial image with at least one thereof enhanced by the image enhancer based upon the at least one environmental condition. The environmental condition may include a weather condition, a time of day, or a time of year. The environmental condition may be typically associated with the collected geospatial image.

The invention discloses a multi-spectral image change detection method based on a generative confrontation network, which solves the problems of low detection accuracy and sensitivity to noise in the prior method. The implementation steps are: 1) setting the structure and objective function of the discriminative classification network D and the generating network G, and the distance coefficient λ between the image generated by the generating network G and the real image; 2) obtaining the difference map of two images in different phases I D ; 3) for I D Divide and obtain the initial change detection result, and according to the result, divide the two different phases into labeled and unlabeled data to form a training set; 4) Use the discriminative classification network D and the generation network G to form a classification network W, and use The training set is trained on it, and the trained discriminant classification network D' is obtained; 5) Two different phase images are input into the discriminative classification network D' to obtain the final change detection result. The invention has the advantages of high detection accuracy and strong robustness, and can be applied to image understanding or pattern recognition.