Remote sensing image

Remote sensing image Overview

The remote sensing image of computer processing must be a digital image. The analog image acquired in photography must perform an image scanner or the like; the digital data acquired by the scan mode must be transferred to a general-purpose carrier such as a general digital computer to read. Computer image processing is done in the image processing system. The image processing system is composed of hardware (computers, displays, digitors, tape drives, etc.) and software (with data input, output, correction, transformation, classification, etc.). Image processing contents mainly include correction, transformation, and classification.

Resolution

space

spatial resolution, also known as ground resolution. The latter is for the floor, refers to the size of the minimum ground distance or the minimum target of the minimum object. The former is for remote sensors or images, and refers to the size or size of the minimum unit on the image, or means a minimum angle or linear distance from the inductor to distinguish between two targets. They reflect the identification, distinguishing capacity of two very close objectives, sometimes called resolution or resolution.

Spectrum

Spectral resolution refers to the minimum wavelength interval that can be distinguished when the remote sensor accepts the target radiation. The smaller the interval, the higher the resolution. The number of bands selected, the wavelength position of each wavelength, and the size of the wavelength interval, these three factors jointly determine the spectral resolution. The higher the spectral resolution, the stronger the topical research, the higher the recognition accuracy of the object, the better the effect of remote sensing application analysis. However, in the face of a large number of multi-band information and these tiny differences it provides, people should directly associate them with the object features, and integrated interpretation is more difficult, and multi-band data analysis can improve identification and Extract the probability and accuracy of the information characteristics.

Radiant Resolution refers to the sensitivity of the detector - the smallest radiation degree difference between the remote sensor sensing element can distinguish when the spectral signal is received, or The resolution of the radiation amount of two different radiation sources. Generally, the number of graded-brightened gradation value (brightness values) is used to scale the number of quantitative levels. It identifies a very meaningful element for the target.

Time

TempleR resolution is a performance indicator for remote sensing imaging intervals. Remote sensing detectors repeat data at a certain period of time, such repetition cycle, also known as the regression cycle. It is determined by the orbital height of the aircraft, the track inclination, run cycle, orbit interval, and the number of bias coefficients, etc. The minimum time interval of this repeated observation is called time resolution.

Imaging method

Aviation photography

photography imaging is an image technology for acquiring objects by imaging equipment. Traditional photographic imaging is an optical lens and a photosensitive film placed on the focal plane to record object images. Digital photography uses the photosensitive element of the placed focal plane to record the image of the object with a digital signal with a digital signal.

Aviation Scan

Scanning imaging is a point-by-line sampling of the target object in a transient field of scanning, to obtain electromagnetic radiation of the target Feature information forms an image of a certain spectrum.

Microwave radar

Microwave imaging radar working wavelengths of 1 mm-1m microwave segments, due to microwave radar is a self-propelled active sensor and microwave has the ability to penetrate the cloud Therefore, the microwal thunder achieves a full day, the weather is characterized. In urban remote sensing, this imaging is important for those identification of microwave-sensitive targets.

Application

Land coverage monitoring: land coverage is the ultimate reflection of people's interaction processes, and also the most obvious landscape sign of the earth's surface system, land coverage changes will trigger a series of environments Change. Remote Sensing Technology has become the most effective means of obtaining land coverage information because it can provide dynamic, rich and cheap data sources.

Forest coverage monitoring: Forest is the main body of land ecosystems, which is the basic resource of human beings. A type of survey of the traditional five-year-old one-year-old two-class survey has a long period of time. After a long time, the sample is easy to treat, the data is comparable, and it is difficult to scientifically, accurately assess the forest resources and ecological conditions.

Remote Sensing has the characteristics of macroscopic, objectivity, periodicity, and convenience, and has also been developed in the forest resource inventory (a type of survey) and planning and design investigation (two types of surveys).

Grassland coverage: Grassland is the land plant resources second only to forest resources. Remote sensing technology is applied in the grassland resource survey, classification, and drawing, greatly improves the development of grassland resources investigation and drawing, which has caused grassland classification to be quantified, and can complete grassland degradation monitoring and evaluation, saving human, material power and financial resources.

Wetland Resource Monitoring: Wetlands is a unique ecosystem formed by plastic land and land interactions, and is one of the most important living environments of naturally ecological diversity of nature.

Monitor the types of wetlands and its number of wetlands in real time, which provides first-hand material for wetlands. It is especially important. Remote sensing technology has a wide range of observations, large amount of information, fast acquisition, short update cycle, saving manpower and artificial interference factors, have become a strong means of wetland research. Wetland boundary can be extracted, and wetland classification, wetland dynamic change monitoring, etc.

Solution

Remote sensing image is too large, and the data type is diverse, so it is a problem, existing solution:

frame Cache technology

This scheme makes the drag are more smooth, but still seamless, if you drag, there is a black block, so that the scheme is to drag, calculate the stuff that needs to be displayed. In an object, you can exchange it. Improvement scheme, because the screen is generally only 1280 × 1024, if you read 9 times the size of the video block (the memory is not 10m) in the displayed part, this is in the image, regardless of how to drag all in the image. In the range, such dragging will appear seamless. When dragging, a cache is still required to store the area you want to use, and exchange it when dragging.

A variety of data formats

can calculate the individual bands of the image when generating the pyramid image, obtain the gradation distribution histogram, then display the real-time calculation, Convert the original format to an 8-bit bitmap.

The current program is the most hard disk, and the bottom layer of the image is also calculated into the pyramid. And there is no frame cache technology, making the jump frame when dragging.

Correction Processing Image Correction refers to the process of eliminating a distortion from a distortion image, eliminating geometric distortion, geometric correction; eliminating radiometric distortion called radiation correction.

Related Data ​​h2>

Geometric Correction

All types of remote sensing images exist in geometric corrections. Since people are accustomed to using a topographic map of orthodontic projection, the distortion of various types of remote sensing must be geometricly corrected as a topographic map. Geometric correction steps are as follows:

1 Select Control Point: Select the same name control point on the remote sensing image and the topographic map, to establish a projection relationship between the image and the map, these control points should be obvious Located place, such as river intersections, etc.

2 Establishing an overall mapping function: Determination of the correction mathematical model based on the geometric distortion of the image and the ground control point, establish a spatial conversion relationship between the image and the map, such as polynomial method, affine transformation Method, etc.

3 heavy sample interpolation: In order to enable the corrected output image cell to correspond to the input unclear image, the data of the input image is rearranged according to the determined correction formula. In the resampling, since the coordinates of the calculated corresponding position are not an integer value, the new image element must be obtained by interpolating the surrounding pixel value.

Radiation is inconsistent with the physical quantity of the grayscale obtained from the image obtained from the remote sensor and the spectral reflectance or spectral radiation brightness of the target. This is because the remote sensor measurement The values ​​include the distortion caused by atmospheric conditions such as solar position and angle conditions, mist and 霭. In order to correctly evaluate the reflection and radiation characteristics of the target, these distortions must be eliminated. The process that eliminates image data in various distortions in radiation brightness is radiation. The results of radiation correction will change the hue and color of the image.

Image Transformation

· Image exchange generally refers to a process of generating another frame image from one frame of graphics, which mainly includes image enhancement and characteristics. Extract two aspects.

Image enhancement

image enhancement is a process of improving image visual effects. When the remote sensing image is analyzed, in order to make the analyst can easily identify the image content, the image data must be processed in accordance with the analysis purposes, and the purpose is to improve the probability of image. Image correction is a process that eliminates the error and distortion generated by the observation, making remote sensing data closer to the true value of the main purpose; while image enhancement puts the focus on the analytics to visually facilitate identifying image content, Typical images enhances grayscale exchange, color synthesis, etc.

Feature Extract

In order to use the instrument to perform image judgment and analysis processing, various parameters such as judgment marks and statistics such as analysis are required in the original image data. Transform images, highlighting the method of representative features, called feature extraction. Feature extraction can quantify the following three characteristics:

Spectrum characteristics

can extract spectral characteristics such as the brightness ratio of color or grayscale or band, such as Landsat MSS There are four bands, depending on the spectral characteristics of a certain type of substrate, it can be protruded from the specific ratio.

Spatial Features

extracts the shape, size, or edge, linear structure of the target, etc., for example, obviously highlighting the region fault.

Texture Features

refers to the characteristics of the texture of the periodic pattern and the uniformity of the region. The processing of image feature extraction is made according to the element shape, distribution density, directionality, and the like of the element shape, distribution density, and directionality, etc. are called texture analysis.

· The image classification is classified by remote sensing images, which is a name that corresponds to a single pixel or a comparison constant cell group. The principle is to utilize the image identification technology to automatically realize the remote sensing image. Classification. Computers are used to identify and classify the main markers of identification and classification. The spectral characteristics of the object, other information on the image, such as size, shape, texture, etc. have not been fully utilized.

Before the computer classification, some pretreatments are often done, such as correction, enhancement, filtering, etc., with different parts of the target characteristics or eliminate the same type of target, terrain, terrain change, scanning observation The difference in brightness is different from the angle.

Computer image classification method, there are two common, namely, monitoring classification and non-supervision classification. Sports, first, you must first select some training samples from the image area of ​​the classification. In this way, the category of the land objects in the training area is known, with it to establish a classification standard, then the computer will perform the entire image on the same criteria. Identify and classify. It is a method of extrapolate unknown area categories by known samples; non-supervisory classification is a classification method without a priori (known) category. For the objects and regions to be studied, there is no known category or training sample standard, but use the image data itself to collect the characteristics of the group in the feature measurement space, first form various data sets, and then check the representatives of these data sets. Object category.

Compared to supervisory classification, non-supervised classifications have the following advantages: do not need prior understanding of the area being studied, the same conditions for the results of the classification, the same conditions, time and cost Saving, but in fact, non-supervised classification is not as high as supervision classification, so supervising classification is more widely used.