Ground truth is a term used in various fields to refer to information provided by direct observation (ie empirical evidence) as opposed to information provided by conclusions.
Video Ground truth
Machine statistics and learning
In machine learning, the term "basic truth" refers to the accuracy of classification of training sets for supervised learning techniques. This is used in the statistical model to prove or not to prove the research hypothesis. The term "ground truthing" refers to the appropriate (verifiable) process of collecting objective data for this test. Compare with the gold standard.
Bayesian spam filtering is a common example of supervised learning. In this system, the algorithm manually teaches the difference between spam and non-spam. It depends on the basic truth of the message used to train the algorithm - the inaccuracy of the truth in the field will be correlated with the inaccuracies in the resulting spam/non-spam verdict.
Maps Ground truth
Remote Sensing
In remote sensing, "ground truth" refers to information collected at the site. However, since the information collected in situ contains measurement errors from users and instruments, the phrases are highly inappropriate and should not be used. In scientific investigation there is no "truth", only estimates of precision and precision. Unfortunately the "basic truth" has become a reckless and misleading way to refer to field survey data, field inventory, land reference data, or soil sampling. Ground truth allows image data to be associated with real features and materials in the field. Ground-truth data collection enables calibration of remote sensing data, and aids in the interpretation and analysis of what is perceived. Examples include cartography, meteorology, aerial photography, satellite imagery and other techniques in which data are collected remotely.
More specifically, ground truth may refer to a process in which [["pixel"]] on a satellite image is compared to what is in reality (at this time) to verify the contents of "pixels" in the image (noting that the concept of "pixels "rather unclear). In the case of a confidential image, this allows a supervised classification to help determine the classification accuracy performed by remote sensing software and therefore minimize errors in classification such as commission errors and omission errors.
The basic truth is usually done on the site, performing surface observations and measuring features of the soil resolution cell feature being studied on remote sensing digital imagery. It also involves taking geographic coordinates of cell resolution with GPS technology and comparing it with the coordinates of "pixels" under study provided by remote sensing software to understand and analyze location errors and how it might affect a particular study.
The fundamental truth is important in the initial classification overseen from an image. When identities and locations of land cover types are known through a combination of fieldwork, maps, and personal experience, this field is known as a training location. The spectral characteristics of this area are used to train remote sensing software using decision rules to classify the rest of the image. These decision rules such as Maximum Likely Classification, Parallelipip Classification, and Minimum Distance Classification offer different techniques for classifying an image. Additional ground truth sites allow remote sensors to create an error matrix that validates the accuracy of the classification method used. Different classification methods may have different percentage errors for a particular classification project. It is important that remote sensors select the classification method that best fits the number of classifications used while giving the least error.
The basic truth also helps with atmospheric correction. Because images from satellites must obviously pass through the atmosphere, they can be distorted due to absorption in the atmosphere. So ground truth can help identify the full object in a satellite image.
Commission error
An example of a commission error is when a pixel reports a feature (such as a tree) that, in fact, does not exist (no tree actually exists). Ground truthing ensures that the error matrix has a higher percentage of accuracy than would happen if there were no correct pixels in the ground. This value is the opposite of the user's accuracy, i.e. Commission error = 1 - user accuracy.
Negligence error
An example of an omission error is when the pixels of a particular object, such as a maple tree, are not classified as maple trees. The ground truthing process helps ensure that pixels are correctly classified and the fault matrix is ​​more accurate. This value is the opposite of the producer's accuracy, ie. Blurring Error = 1 - producer accuracy
Geographic information system
Geographic information systems such as GIS, GPS, and GNSS, have become so widespread that the term "basic truth" has taken on a special meaning in that context. If the location coordinates returned by the location method such as GPS is the approximate location, then the "basic truth" is the actual location on earth. The smartphone might return a set of approximate coordinate locations such as 43.87870, -103.45901. The basic truth assessed by these coordinates is the tip of George Washington's nose at Mt. Rushmore. The approximate accuracy is the maximum distance between the location coordinates and the basic truth. We can say in this case that the approximate accuracy is 10 meters, which means that the point on earth represented by the location coordinates is estimated to be within 10 meters of George's nose - the basic truth. In slang, the coordinates show where we think George Washington's nose is, and the truth is essentially the real place. In practice a smart phone or handheld GPS unit can routinely estimate the basic truth within 6-10 meters. Special instruments can reduce GPS measurement error to under one centimeter.
Military use
US military nagging uses ground truth to describe the reality of a tactical situation - as opposed to intelligence reports and mission plans. The term appears in the title of the Iraq War Movie documentary The Ground Truth (2006), and also in military publications, for example Stars and Stripes says: "The line decides to find out what the truth is land in Iraq. "
Etymology
The Oxford English Dictionary (sv "ground truth") records the use of the word "Groundtruth" in the sense of "fundamental truth" of Henry Ellison's poem "The Siberian Exile's Tale", published in 1833..
References
External links
- Remote Sensing Sensor Sensor Project (including example of an error matrix)
Source of the article : Wikipedia
0 Comments