1. The concept of change detection
Change detection is to quantitatively analyze and determine the characteristics and process of surface changes from remote sensing data in different periods;
Remote sensing change detection is the process of identifying and evaluating changes in various surface phenomena over time;
Remote sensing change detection is the change of the spectral response of image pixels in two periods caused by the change of surface features in the instantaneous field of view of remote sensing over time.
2. Application of change detection
Civilian use: mainly used for obtaining information on changes in land use and cover, forest and vegetation changes, and urban expansion in resource and environmental monitoring;
Mapping: geospatial data updates;
Natural disasters: monitoring and assessment of disasters such as earthquakes, floods, mudslides and forest fires.
Military: Applied to damage effect assessment, dynamic perception of battlefield information, monitoring of military targets and troop deployment, etc.
3. The main content of change detection
It involves the description of the type of change, distribution status and change information, that is, it is necessary to determine the type of features before and after the change, the boundary and the attributes of the analysis change:
1) Determine whether a change has occurred;
2) identify areas of change;
3) Identify the nature of the change;
4) Assess temporal and spatial patterns of change.
Among them, the first two aspects are the basic goals to be achieved by change detection, so they are also the problems that need to be solved first in the research of change detection. The concept of change detection in a narrow sense is composed of these two aspects.
4. Remote sensing image change detection processing flow
1) Preprocessing
Including geometric correction, geometric registration, radiometric correction, image filtering, image enhancement
Ideally, the data used in the change detection process should be obtained with the same sensor, and they should be recorded with the same spatial resolution, the same shooting Geometry, same spectral bands, same radiometric resolution, and taken at the same time of day.
The spatial registration accuracy of different image data is also a necessary condition for effective change detection. The accuracy after registration needs to be within 1/4~1/2 pixel
2) Change area detection
Various fancy CNN networks
3) Post-processing
Due to the influence of noise, the classification results after decision-making and discrimination will contain many false alarms. The post-processing stage processes the classification results, filters out some false alarms, and eliminates uninteresting changes to meet actual needs.
The main methods include filter processing, region growing method, mathematical morphology processing, etc.
4) Performance evaluation
(1) Change detection performance evaluation can be quantitatively analyzed by calculating the number of false alarms, the number of missed pixels and the total number of error pixels;
(2) Another important aspect in evaluating the performance of change detection methods is the level of automation and effectiveness of change thresholds
5. The change detection algorithm is classified from the detection level
1. Pixel level: Calculate and process the corresponding pixel gray value of the two images to detect changes.
Advantages: retain as much original information as possible, and have detailed information that is not available on other levels
Disadvantages: low efficiency; no consideration of changes in characteristic attributes such as space; poor anti-interference ability (natural factors such as solar radiation angle and surface humidity)
2. Feature level: Use a certain algorithm to extract feature information (edge, shape, contour, texture) from the original image, and then perform comprehensive analysis and change detection on these feature information
Advantages: high operating efficiency; higher credibility and accuracy in judging feature attributes; to a certain extent, it reduces the interference of external factors on the results
Disadvantages: Part of the information will be lost during the feature extraction process, and it is difficult to provide subtle information; it depends on the result of feature extraction, but the feature extraction itself is more difficult
3. Target level: It mainly detects certain specific objects (such as roads, houses and other targets with clear meanings). It is a change detection based on image understanding and image recognition. It is a high-level analysis method based on the target model.
Advantages: It is close to the needs of users, and the detection results can be directly applied
Disadvantages: Target extraction is difficult
6. Change area detection algorithm is classified from the principle of algorithm realization
Continually updated. . .
It's not easy to organize, thank you for liking and following!