Explore chapters and articles related to this topic
Advanced Topics
Published in Russell G. Congalton, Kass Green, Assessing the Accuracy of Remotely Sensed Data, 2019
Russell G. Congalton, Kass Green
An increasingly popular and extremely important application of remotely sensed data is for use in change detection. Change detection is the process of identifying differences in the state of an object or phenomenon by observing it at different times (Singh, 1989). Four aspects of change detection are important: (1) detecting that changes have occurred, (2) identifying the nature of the change, (3) measuring the areal extent of the change, and (4) assessing the spatial pattern of the change (Brothers and Fish, 1978; Malila, 1985; Singh, 1986). Techniques to perform change detection with digital imagery have become numerous because of an ever-increasing amount of imagery, improved versatility in manipulating digital data, better image analysis software, and significant developments in computing power. Just as assessing the accuracy of a single-date map is vital, change detection accuracy assessment is an important component of any change analysis project.
Senses in Action
Published in Haydee M. Cuevas, Jonathan Velázquez, Andrew R. Dattel, Human Factors in Practice, 2017
Lauren Reinerman-Jones, Julian Abich, Grace Teo
At times, signals may be very salient, yet they go undetected. This is especially true when trying to detect changes. Change detection is the processing involved in noticing a change, identifying the type of change, and locating the change (Rensink, 2002). Each sensory system has a different threshold at which a change can be detected (this will be discussed in more detail in the Methods section). Further, it may seem rational to believe that a change occurring to the same sensory system should be detected more often, but this is not always the case. The inability to notice change is referred to as change blindness (Simons & Levin, 1997). A very famous study required participants to view a visual scene of people passing a ball around and were asked to count how many passes occurred (Simons & Chabris, 1999). While this was occurring, a person dressed as a gorilla walks through the center of the group. When participants are asked what animal walked through the scene, most are unable to provide an answer. When the video is presented and the attention is focused on finding the animal, participants find it difficult to believe that they were unable to detect it the first time around. This emphasizes the importance of attention in detecting signals.
Design of a Local Parallel Pattern Processor for Image Processing
Published in K. S. Fu, Ichikawa Tadao, Special Computer Architectures for Pattern Processing, 1982
Ken-Ichi Mori, Masatsugu Kidode, Hidenori Shinoda, Haruo Asada
The data used in this change detection consists of two sets of time-different multispectral scanner data extracted from LANDSAT CCTs (over Kanto and Sagami Area). Preprocessing. Each MSS datum is geometrically rectified to spatially fit the image to the map in the Universal Transversal Mercator’s projection. The third order polynomial registration is employed with the aid of a set of ground control points, which are interactively defined on the color monitor and the corresponding topographic map.Classification. The maximum likelihood classification method is used for each MSS data set, whose pixels are categorized into ten classes of land cover: urban area, residential area, grass land, bare field, rice field, vegetable field, water, coniferous forest, mixed forest, and shadow part.Change detection algorithm. Changes have been found in rice fields and vegetable fields due to harvesting activities, changes in residential area along with housing development, change in shadow area due to sun-height variation, etc. Other major change detection aims are to estimate the extent of a certain crop, find a specific developed area, monitor water pollution factors, survey damaged area, and so on.
Integration of Landsat time-series vegetation indices improves consistency of change detection
Published in International Journal of Digital Earth, 2023
Mingxing Zhou, Dengqiu Li, Kuo Liao, Dengsheng Lu
The performances of five indices in detecting abrupt vegetation change were evaluated at both temporal and spatial scales. A stratified random sampling method was adopted (Schultz et al. 2016), and 100 reference points were randomly selected in each of the five abrupt change maps, with 50 samples in abrupt change areas and 50 samples in non-abrupt change areas. A total of 500 (5 × 100) reference samples were collected. The primary and optimal source for reference data is the Landsat images themselves (Cohen, Yang, and Kennedy 2010; Z. Zhu and Woodcock 2014; Wu et al. 2020). Meanwhile, high spatial resolution images from Google Earth help manual interpretation of the land cover classes and change. We manually interpreted each sample and its adjacent eight pixels using Google high-resolution images and Landsat time-series images for determining whether the samples changed or not, and the change years. The overall accuracy (OA), user’s accuracy (UA), and producer’s accuracy (PA) were calculated to assess change-detection accuracy. During visual interpretation, analysts knew in advance whether the pixel had experienced change or not in the years being examined, but did not know which index had detected these change years. The examined year was set to the actual change event if a significant change occurred within a 2-year window of it. If multiple change points were detected for a pixel, only the year that was the same or closer to the actual change year was used, and the change in this pixel was regarded as correctly detected.
cpss: an package for change-point detection by sample-splitting methods
Published in Journal of Quality Technology, 2023
Many packages are available to make existing change detection methods directly applicable in practice. Among others, the changepoint package (Killick and Eckley 2014) can be used to detect changes in the mean and/or variance of a univariate sequence, packages ecp (James and Matteson 2014) and changepoint.np (Haynes and Killick 2021) for nonparametric change detection in the distribution of a time series, and the strucchange package (Zeileis et al. 2002) with the breakpoints method for change detection in linear regression models. The dfphase1 package (Capizzi and Masarotto 2018) is very useful for Phase I analysis in SPC, which implements the recursive segmentation and permutation method (Capizzi and Masarotto 2013, 2017) together with other change detection approaches. Besides that, packages bcp (Erdman and Emerson 2007), Segmented (Muggeo 2008), cpm (Ross 2015) and chngpt (Fong et al. 2017) are available for different change models. In addition, Turner et al. (2001) developed a program to detect shifts in the mean or standard deviation of a univariate sequence in Phase I.
Multi-class change detection of remote sensing images based on class rebalancing
Published in International Journal of Digital Earth, 2022
Huakang Tang, Honglei Wang, Xiaoping Zhang
Due to the rapid development of deep learning, change detection algorithms mainly rely on deep neural networks, and there are many change detection methods. However, there are little research on change detection from the perspective of semantic segmentation. If the change detection task is considered from this point of view, the precision of semantic segmentation determines the effect of change detection. It is easy to make the change detection task become a simple semantic segmentation task. From the perspective of semantic segmentation, change detection tasks are different. The image obtained after semantic segmentation is a two-phase semantic map. How to explain ground changes from a semantic map? In this paper, the specific ground changes in the changing area of the image of the two-time phases are used to represent the characteristics of the multi-analog changes, and the specific ground classes in the change detection task are accurately represented.