# A method for the identification of vegetation types

(57) Abstract:

The invention is intended for the rapid diagnosis of vegetation cover and stages of reforestation. Footage of the studied forest area, received on Board the orbital station, converted into digital form and calculate the two-dimensional spatial spectrum. The inverse Fourier transform to determine the autocorrelation function of the spectrum and calculate the regression function between the parameters of these functions for different vegetation types. After calibration of the results obtained by the image control plots synthesize a General picture of the vegetation over the entire area of observation. The technical result consists in the reception and transformation of the texture characteristics of images obtained by remote methods, signs sufficient to identify the type of vegetation, as well as to improve efficiency, productivity and global retrieve the target information. 5 Il. The present invention relates to forestry, in particular for surgical diagnosis vegetation clearing, the assessment of current and subsequent stages of regeneration of the forest.When Chemiczne the woodland, larch forests with high-, middle - and nizkobonitetnyh, pine forests, Gary, alasses (shrub, herbaceous). In forestry, the traditional method of determining the type of logging and stage of regeneration of the forest is their visual and instrumental inspection and mapping. For the minimum value of the allocation is made 1 ha scale 1:10000 or linear dimensions of the elementary areas of 100 x 100 m Known assessment by cutting stationary observations experimental plot at the interval of 20 years (see, for example, I. S. Melekhov. Forestry. M: Agropromizdat, 1989, S. 8, PL. 2 is similar).When the known method of forming stand on the cuts at all stages - thick calves, large pole, ripe forest estimate the number of trees per 1 ha and of a thickness of > 6 cm at a height of 1.3 meters, also Known various methods (IUFRO, Kraft, Zhilkin B. D.) classification of trees as the crown, the diameter of the shafts, vitality, position in the stand (see ibid., analog, S. 245 - 247). The disadvantages of the known analogues are great complexity associated with the need for individual conversion of each tree, neoperativnost, inaccessible mountainous and remote areas.Saava A. C. and other Aerospace monitoring of forests. M.: Nauka, 1991, S. 28 - 30.

The nearest analogue score (in conventional points) categories of forests carried out by remote method, by measuring the ratios of the spectral brightness of the sensed areas in blue, green and red parts of the visible range, the calculation of the chromatic values of the coefficients and comparing the measured values with the calibrated values of the regression function chromatic coefficients obtained by sensing the reference (test) sites.The known method is implemented by the following sequence of operations:select control sites and determine visually-instrumental method of the state stand in points;

place spectrometric module on a stand-alone rotary platform artificial satellite of the Earth;

track laid pad input settings ballistic data in the control system of the Autonomous platform;

measure the coefficients of the spectral brightness (

_{B},

_{G},

_{R}) probing pad in blue (B), green (G) and red (R) regions of the visible spectrum;

quantum continuous values of the amplitudes of the signals measured function increments kV is BR> calculate the correlation function of the chromatic coefficients and tazrout its reference measurement control plots with known status categories in points;

get the evaluation of the state of the forests over the entire area of observation.The known method has the following disadvantages.1. The method cannot be directly used to identify vegetation types, because the received signal carries integral characteristics of the underlying surface from which it is impossible to distinguish component structure of the canopy.2. The coefficients of the spectral brightness of different vegetation types have the same type of dependency and differ from each other in units of percent, which does not provide the necessary accuracy and reliability of the identification. In principle, index of color in the vegetation period is difficult to distinguish young greens from green meadows.The problem solved by this invention is in the reception and transformation of the texture characteristics of images obtained by remote methods in the "roughness" of the canopy vegetation sufficient to identify its type, as well as the benefits witecka target information.This goal is achieved by introducing into the nearest analogue of the following technological operations:

provide high resolution imagery of the underlying surface containing control sites, and record the received image on the onboard tape recorder;

transmit communications over the air in the mission control Center received images and record them on cassettes;

break the video frame into a sequence of single plots and transform the image of each section in the digital function of the intensity I (x, y);

calculate the two-dimensional spatial spectrum from the intensity function and the physical dimensions of the elementary area direct Fourier transformation (software-based methods fast Fourier transform) in accordance with the formula

< / BR>

receive from the spatial spectrum G (f

_{x}f

_{y}) the energy spectrum of the image signal S(F);

calculate the inverse Fourier transform of the autocorrelation function B() of the image signal according to the formula

< / BR>

expect reference regression dependence between the parameters of the autocorrelation function B(=0) and /(B=0.5 B

_{max}) for different vegetation types of the control area is a major in the captured video frames over the entire area of observation.Comparative analysis of the proposed solutions with the closest analogue shows that the inventive method differs from the known to the introduction of new manufacturing operations, ensuring the achievement of the properties, laws which were not known and manifested in the claimed object for the first time. Indeed, the main breeding option closest analogue is the index of the color, whereas in the present method information parameter is the roughness of the vegetation cover. In turn, the "roughness" of the sensed area "extracted" from the texture image. This allows us to assert that the claimed method meets the criteria of the invention "novelty". The presence of such signs, as introduced operations associated with the functional transformation textures in the image, allowing you to retrieve information about the "roughness" of the sensed areas, the calculation of the regression function between the type of vegetation sample plots and autocorrelation parameters reflected from this type of vegetation signal, allows to make a conclusion on the conformity of the proposed technical solution the criterion of "Substantial differences".Technical essence isobariceski processing of space images of vegetation is the index of the color. Curves of spectral coefficients brightness all tree species have the same patterns and differ from each other in units of percent. This makes it difficult thematic image processing on the index color. In addition to physiological factors on the magnitude of the CNS influences and architecture of plant height, crown shape, the ratio of the tiers. Consequently, the "roughness" of the sensed surface can also serve as the breeding grounds of vegetation, i.e., the target information on the underlying surface contains not only the tone of the image, but also its texture. The sparsity or density of the canopy, the diameters of the crowns, their geometrical repeatability are involved in the formation of an image and its large discrete contrasting elements, which is reflected in the quality of the texture.Analysis of the known definitions of texture in the field of interpretation of the images showed that the generally accepted formal definition of texture as interpretive sign does not exist. Well-known definition of the textures are, as a rule, to its qualitative description: fine, coarse, smooth, linear, irregular, and so on, the Authors have developed a method of obtaining quantitative is sleduyushim the dimensionality reduction of its feature vector to identify up to two parameters: the width of the autocorrelation function and its maximum amplitude. Assessment methods "roughness" of the surface based on the calculation of the spatial spectrum of the corresponding area of the picture. Of mathematics known (see, for example, Piskunov N. With. Differential and integral calculus, so 2, M. : Nauka, 1964, S. 240 - 242) that any function can be decomposed into the Fourier integral. Moreover, the greater the rate of change of a function of the coordinate, the wider its range. Therefore, the roughness or irregularity of the canopy Dene the frequency-spectral image of vegetation. By definition, the spatial spectrum is calculated as a two-dimensional Fourier transformation from the intensity function I (x, y) image area:

< / BR>

However, in view of the envelope of the spatial spectrum is difficult to identify the type of vegetation because of its indentation, mnogoyadernosti and implicitly expressed peaks.Another characteristic of the speed fluctuations of some processes are their autocorrelation function. But directly a function of the image intensity autocorrelation function calculated could not be due to the loss of temporal coordinates at the image acquisition (instant snapshot).The authors have developed posledovatelnostyu autocorrelation of the image signal, including the following interim procedures:

calculation of two-dimensional spatial spectrum G (f

_{x}f

_{y});

enter the physical dimensions of the analyzed site;

rationing and obtaining a power spectrum of the image signal S (F);

the calculation of the autocorrelation function of B() by inverse Fourier transform from the energy spectrum by the formula

< / BR>

The estimated parameters to identify the type of vegetation are two values, the autocorrelation function is its maximum value at =0, B=0) and the width of the autocorrelation function at B = 0.5 B

_{max}. The strength of the relationship between vegetation type and estimated parameters is characterized by the interpretation of the regression function, calculated by reference to the shooting of control plots.The inventive method is implemented by the following operations:

select control sites with characteristic vegetation types and spend their thematic evaluations;

place video complex is next on automatic rotary platform orbital station;

monitor the specified areas monitoring input setpoints ballistic data in the control system of the Autonomous platform;

ASU and write the resulting image to the on-Board recorder;

passed in sessions radio at MCC acquired images and carry out a census on cassette;

consistently selected for analysis sections of the frame and transform their image in the digital function of the intensity I (x, y);

calculated on the basis of software-based fast Fourier transform complex Fourier spectrum G (f

_{x}f

_{y}) the function I (x, y) in accordance with the formula:

< / BR>

convert the two-dimensional Fourier spectrum G (f

_{x}f

_{x}in the energy spectrum S(F) signal, taking into account the physical dimensions of the entered image;

calculate the autocorrelation function B() of the image signal of the analyzed site inverse Fourier transformation from the energy spectrum S(F) by the formula

< / BR>

expect reference function regression between the parameters of the autocorrelation functions of control plots and type of vegetation on them;

synthesized from consistently analyzed plots mosaic pattern of vegetation on the captured video frames over the entire area of observation.An example of a specific implementation of the method.Evaluation metric and informational characteristics of a frequency-spectral images of the orbital station "Mir". Implementation of the proposed method is reflected in the Methodology remote solution forestry and mesoeconomics tasks videopokeronline means of the orbital complex "Mir", NPO Energia, M., 1994, 31 C. the Inventive method can be implemented on the basis of the device on the circuit of Fig.1. Fig. 2-5 also explain the proposed method.On a stand-alone rotary platform 1, controlled from mission control Center and from the vehicle control system 2 of the orbital station "Mir", set video complex is next 3 high spatial resolution type "Astra". Under the program or in the direct control exercised capture and record images of objects 4 underlying surface, tracked videocompressor, on-Board recorder 5 type "Betta Kam". The accumulated information of the video frames is reset in sessions radio 6 type "Daisy" at MCC, where recorded on tape device 7.Ground set of hardware and software provides a PC/386/387 with a set of peripheral elements in the structure of the device televote 8 images in the computer, the floppy set 9 with reference videosnimci key areas, a set of diskettes for 10 write FAI is ing frame is entered in the random access memory of the Central processor (11) PC with visualization on the monitor 12 type Super VGA. A set of specialized processing programs recorded in the ROM 13. The resulting image processing target information in the form of graphs of the estimated functions is displayed on the printer 14.The procedure of identifying the type of vegetation on the operations of the proposed method by means of Fig.1 we illustrate on the example of individual video frame shown in Fig. 2. In Fig. 2 shows a section of the underlying surface, enters the field of view of the onboard videocomplete "Astra". For the analysis of vegetation type field of the frame is successively broken down into elementary parts (dash-dotted line in Fig. 2) defined by the requirements of identification accuracy and spatial resolution capabilities. A single element is chosen from the condition of commensurability with the dimensions of the control area (100 x 100 m). Each elementary area (single element) of the frame telewoda 8 is converted into a file of digital information intensity function I (x, y) of the image area in the format 515 x 512 elements. The operation of converting image files in digital information intensity function is implemented by means of televote (see, for example, the System televote Panasonik. Instruction iprocessor 11. From each array of the digital information file is calculated two-dimensional spatial spectrum using a set of specialized programs contained in the ROM 13. About the feasibility of this operation, the fast Fourier transform, see , for example, Marple, S. A. (ml), translation from English. Digital spectral analysis. M.: Mir, 1990, S. 77 - 79, the FFT algorithm.For structures that do not have a predominant periodicities, the so-called pseudostructure in the Fourier spectrum will contain many components. Usually two-dimensional Fourier spectrum of the pseudo-random structures looks like a vaguely elliptical shape. Dispersion (roughness) of the structure is reflected in the magnitude of the Fourier spectrum. The next step of the analysis is the assessment of the contribution of spatial structures of different size and type of envelope. By integrating the two-dimensional Fourier spectrum of the circular segments derive a one-dimensional function of the spatial spectrum S(1/). About the realizability of this, see, for example, a System of digital image processing. Videolab, MSU, 1990, version 2.1; 2.2 S. 63 - 65, technical report.In Fig.3 shows a frequency-spectral images of individual sections of two types of vegetation: (a) coniferous forest, with a span of CZK 5 - 7 the Ministers", , Moscow received satellite imagery. Canopy homogeneous coniferous forest has a large "roughness" and a darker tone in the picture, but the larch woodland contains a wider spatial range of wavelengths and has a lighter tone on the image, which determines the small interval of the correlation of the image signal.Qualitatively functions S(1/) characterize the "roughness" of a vegetative canopy, similar to the wavy sea surface, containing the entire range of wavelengths from the fine ripples to heightened excitement. Between the wavelength () of any vibration and its frequency (f) there is a clear dependence

< / BR>

where

v is the velocity of propagation of physical waves.Because shooting is done from orbit, and the objects of the underlying surface is stationary, the velocity of propagation of spatial waves, you can take the relative velocity of motion of objects and the meter. Thus, the spatial spectrum and a frequency spectrum associated with the dependence of G(f) = v/S(1/) or for a given physical phenomena through a certain constant factor of proportionality. By definition, between a range of frequencies G (f) and the power spectrum S(F) there is Saarsalu 7.30)

< / BR>

i.e., the energy spectrum is the square of the amplitude spectrum, averaged over the observation interval of a single implementation.The calculation of the autocorrelation function B() by their power spectra S(F) is the inverse Fourier transform on the well-known relations (see , for example, Driveways A. M. fundamentals of statistical radio engineering. M: Svyazist, 1969, S. 94; formula 7.35)

< / BR>

In this example, the specific implementation of the energy spectra S(F) two types of vegetation approximated functions

< / BR>

and was calculated analytically. The calculated values of the autocorrelation function presents graphs of Fig.4. This operation can be automated software-based calculation methods, specialized programs recorded in the ROM 13. About the realizability of the software calculations of the autocorrelation function, see, for example, Marple, S. A. Digital spectral analysis and its application. Translation from English. M.: Mir, 1990, S. 254.From the presented dependencies Fig. 4 it follows that the autocorrelation function of vegetation types differ significantly as the amplitude (power variable components), and the correlation interval (speed can be reliably decoded.Calibration device for obtaining the interpretive characteristics by typing in the PC and processing of images from floppies 9 key areas. The result of this procedure will receive a set of implementations of the autocorrelation functions corresponding to the types of vegetation in key areas. Peak values of amplitudes of B(=0) and the intervals of the correlation /B = 0.5 B

_{max}calculate the regression function between the parameters of the autocorrelation functions for different types of vegetation in key areas. Approximation of nonlinear regression functions is carried out by known methods (see , for example, G. Korn, T. Korn. Handbook of mathematics for scientists and engineers. Translation from English. M.: Nauka, 1970, S. 495, p. 18.4.6, regression).Received by the reference imagery interpretation response (regression function) is shown in Fig. 5. With interpretive feature, they identify the type of vegetation on a single plot, the sequence of which synthesize mosaic of vegetation type of the entire scene.The positive effect of the proposed method is based on establishing a quantitative relationship between charactere compared with the prototype of the two parameters, namely, the amplitude of the autocorrelation function associated with the tone (color) image, and the correlation interval associated with the "roughness" of the canopy vegetation. A method for the identification of vegetation types by defining its status and thematic evaluation on sample plots, remote sensing of the underlying surface containing the control sites, with the orbit of the spacecraft means mounted on the stand-alone platform, monitor the specified areas monitoring input setpoints ballistic data in the control system of the Autonomous platform, characterized in that conduct remote, frame-by-frame video objects, transmit the received images and register them on a ground-based, split video frame in a sequence of single plots and transform the image of each section in the digital function of the intensity, calculate the two-dimensional spatial spectrum of the intensity function and the physical dimensions of the elementary section, receive from the spatial spectrum the energy spectrum of the image signal, calculates the inverse Fourier transform of the autocorrelation function of the image signal, rashipov vegetation control plots, synthesized from the consecutive analysis of single plots mosaic pattern of vegetation on the captured video frames over the entire area of observation.

**Same patents:**

Device for digital filtering // 2055396

The invention relates to computer technology, in particular to the processing of information, and can be used in devices receiving the information transmitted by the communication channel, in which there is a possibility of distortion, i.e

The invention relates to computer technology and can be used in specialized high-performance computing systems for solving least squares, solving systems of linear algebraic equations for the case when the number of equations is greater than the number of unknown or no confidence in a good condition matrix

Device for interpolation // 2042979

The invention relates to computer technology and can be used to perform functional transformations in a variety of digital information processing systems, such as onboard aircraft systems

Digital filter // 2029362

The invention relates to digital signal processing and can be used to filter the valid signals in radio devices for various purposes, in particular in radar

Multichannel digital filter // 2029361

The invention relates to computing, and in particular to means of digital signal processing in real time, and can be used in radar, navigation and communication for construction of high-speed digital processors

Linear interpolator // 2028665

The invention relates to the processing and transmission of signals, where the linear interpolation of signals at high speed, when working in real time, such as telemetry, automation, communications, computing, etc

The device for calculating the convolution // 2112274

The invention relates to the field of computer engineering and can be used in specialized computer systems to calculate swerski

Matrix setprocessor // 2079879

The invention relates to computing and is designed to build on the basis of specialized computers

The invention relates to computer technology and can be used in specialized computer systems to calculate the two-dimensional convolution

Device for multiplying two matrices // 2049352

The invention relates to computer technology and can be used in specialized computing systems, in particular, in digital signal processing for multiplying two matrices

The invention relates to computer technology and can be used in the automated command and control system for managing the movement of different types of vehicles on the road network with different cross-sections

Vector accelerator // 2042980

The invention relates to computer technology and can be used in the production of vector-based devices, providing high performance vector operations

The invention relates to computer technology and can be used for performance analysis of Queuing systems

A device for processing matrices // 2037200

The invention relates to computing and microelectronics, and can be used in specialized pre-processing of multidimensional signals

Device for the treatment of n x n matrices // 2037199

The invention relates to computer technology and can be used in specialized high-performance computing machines and devices, signal processing for handling n n-matrices

Digital correlator // 2067316

The invention relates to the field of radar and is intended to implement the operations of convolution of two signals

The invention relates to computer technology and can be used for the calculation of the correlation function of the input signal data reference function (PF) in the satellites of the type "DIAMOND" (Russia), ERS-1 (Europe), JERS-1 (Japan), RADARSAT (Canada), SIR-A (USA), for aircraft type AWACS and JSTARS, and in medicine in tomography and ultrasonic sensing and Geology

Sign correlates // 2055395

Digital correlator // 2051413

The invention relates to devices of automatics and computer engineering and can be used in correlation extreme systems such as television-correlation