Quadruple stacked-based concept: A novel approach for change detection

• Published in: International journal of applied earth observation and geoinformation 2023-07, Vol.121, p.103361, Article 103361
• Main Authors: Jemy, G., Shokry, M.M.F., Farag, K.S.I., Abdelmalik, K.W.
• Format: Article
• Language: English
• Subjects: Change detection; PlanetScope and Sentinel-2; Quadruple-stacked based concept; Remote sensing
• ISSN: 1569-8432; 1872-826X
• DOI: 10.1016/j.jag.2023.103361

•A novel accumulated NDVI is employed for effective determination of what changes.•A novel Quadruple Stacked-Based C D (4 images/2 time periods) is proposed.•Testing of the proposed concept shows an accuracy of 97%. Developing a novel change-detection approach instead classic one is imperative for better assessment of the accuracy of evaluable results. Modern generations of optical satellites, which are characterized by their high to moderate spatial and temporal resolutions enable us to continuously develop the traditional approaches. Successive changes on land with different scales, and their environmental impact needs more accurate specific approaches, which reduce the noise in the data and increase the precision of the results. Integration between many developed methods may produce a new specific one, which solves this. The present work developed a novel approach based on time series results temporally at a very short time scale “Quadruple-stacked changes concept”. The higher weight of change probability indicates a certain change. The advantage of the developed methodology is decreasing the amount of data noise; by using the weight concept of GIS layers. The NDVI “Normalized Difference Vegetation Index” was calculated for each month all over the used two years (2016 and 2021) - using Sentinel-2 imagery data - then were subjected to reclassification and weighted sum to produce the accumulated NDVI. PlanetScope imagery data were used for Quadruple image difference (4 images/2 times). The Threshold of 1/3, ½ and 1 Standard deviation were tested - this step was essential for mapping the true changes which is considered as one of the inputs of the change detection - The resulted accuracy of the real changed areas detected by the newly developed concept was 0.9662, 0.9602, and 0.9579 respectively to the used SD.