Target detection and identification method under complex background

Abstract

The invention relates to the technical field of optical target detection and identification, and particularly discloses a target detection and identification method under a complex background. The method comprises the steps of determining an application scene to obtain scene simulation parameters; constructing a three-dimensional simulation scene by utilizing 3D modeling software; utilizing a background model to simulate spectral features, scattering intensity and polarization features of a specific target and various complex backgrounds; scene simulation is carried out according to the scene simulation parameters, the three-dimensional simulation scene and the background model, and a simulation grayscale image is acquired; and carrying out gray scale comparison on the simulation gray scale image and an actually measured image of the application scene, and carrying out identification detection on a specific target which is not determined to exist in the application scene through a target identification algorithm to obtain an analogue simulation scene and a target identification result. According to the invention, a spectrum detection method is used to improve the target identification probability under a complex background.

Description

Technical Field

[0001] The present invention relates to the technical field of optical detection and identification of targets, and particularly to a method for detecting and identifying targets in a complex background. Background Art

[0002] With the development of technology, currently, target recognition technology has been widely applied in fields such as biomedicine, satellite remote sensing, robot vision, cargo detection, target tracking, autonomous vehicle navigation, banking, transportation, military, e-commerce, and multimedia network communication. The existing target recognition methods mainly use deep learning methods as the mainstream recognition methods, and such methods usually rely on a large number of data samples to support the recognition accuracy; moreover, the existing target recognition methods generally have problems of poor anti-interference ability and low recognition rate, especially in the process of recognizing targets of the same color in a complex background; Therefore, those skilled in the art urgently need to provide a brand-new method for target recognition in a complex background to overcome the defects existing in the above-mentioned prior art. Summary of the Invention

[0003] Therefore, the technical problem to be solved by the present invention is to overcome the defects existing in the prior art, so as to provide a method for detecting and identifying targets in a complex background. By combining the Bidirectional Reflectance Distribution Function, abbreviated as BRDF function, which is an important function used to describe the reflection characteristics of the object surface to incident light, the detection and identification of targets in a complex background are carried out.

[0004] A method for detecting and identifying targets in a complex background includes the following steps: Determine the application scenario and obtain the scenario simulation parameters; Use 3D modeling software to process the scenario simulation parameters to construct a three-dimensional simulation scenario; Use the pre-constructed background model to simulate the spectral characteristics, scattering intensity, and polarization characteristics of a specific target and various complex backgrounds; Perform scenario simulation according to the scenario simulation parameters, three-dimensional simulation scenario, and background model, and further obtain a simulated grayscale image; Compare the grayscale of the simulated grayscale image and the measured image of the application scenario, and identify and detect a specific target whose existence in the application scenario is uncertain through a target recognition algorithm to obtain a simulation scenario and a target recognition result.

[0005] Preferably, during the simulation process, it also includes performing simulation modeling on the application scenario according to the reflectivity of the specific target material itself and the spatial coordinates of the three-dimensional simulation scenario.

[0006] Preferably, obtaining the simulated grayscale image specifically includes the following steps: Introduce the solar incident direction for spatial calculation, and perform a full-space traversal calculation according to the BRDF function to obtain the full-space BRDF value of a specific target; Set the observation direction, calculate the BRDF value of each surface element of the specific target material at the corresponding angle, and obtain the simulated grayscale image of the simulated scene.

[0007] Preferably, the multiple complex backgrounds specifically include: woodland, desert, grassland, highway, and snowfield.

[0008] Preferably, the background model includes: a woodland pBRDF model, a sandy land pBRDF model, a grassland pBRDF model, a highway pBRDF model, and a snowfield BRDF model corresponding to the multiple complex backgrounds; Among them, the expression of the snowfield BRDF model is: ; The formula, represents the specular reflection component of the BRDF on the sample surface; is the normal distribution function of the small surface element on the sample surface, is the Fresnel approximation function, is the masking function; represents the BRDF value of the snowfield model; , , , b are undetermined parameters; Specifically: and respectively reflect the magnitudes of the specular reflection and diffuse reflection components, and are related to the roughness and reflectivity of the target surface; reflects the slope distribution of the target surface, and is related to the roughness and texture distribution of the target surface; b is a parameter reflecting the Fresnel reflection coefficient of the target surface and is related to the refractive index of the target; are the incident zenith angle, incident azimuth angle, scattering zenith angle, and scattering azimuth angle respectively; represents the angle between the light ray from the point on the aperture to the observation point and the aperture normal; represents the angle between the target surface normal and the microfacet normal.

[0009] Preferably, the expression of the sandy land pBRDF model is: ; ; Among them, is the specular reflection component of the sandy land, is the normal distribution function, θ represents the angle between the incident light and the reflected light; σ represents the surface roughness; is the masking factor, represents the relative azimuth angle; is the degree of polarization, represents the refractive index, is a 4×4 Fresnel reflection Mueller matrix; A is a fitting parameter, β represents the angle between the reflection direction and the normal direction, B is a fitting parameter, represents the reflected S wave, represents the reflected P wave, represents the BRDF value of the sand model.

[0010] Preferably, the pBRDF model expression for grassland is: ; Among them, represents the slope variance of the rough surface; is the specular reflection coefficient; is the reflectivity of the target surface; is the free scattering coefficient; represents the free fitting parameter, represents the six-parameter model; represents the depolarization matrix of volume scattering, represents the depolarization matrix of backscattering; represents the fitting factor.

[0011] Preferably, the pBRDF model expression for forest land is: = ； ; ; In the formula: represents the pBRDF model of forest land, represents the diffuse reflection component of the pBRDF model of forest land, is the surface autocorrelation length function, represents the BRDF value of the specular reflection component.

[0012] Preferably, the pBRDF model expression for highway is: .

[0013] The technical solution of the present invention has the following advantages: The method of using spectral detection is adopted to improve the recognition probability of targets in complex backgrounds, and at the same time reveal the light BRDF scattering situation of targets in all directions in space in complex backgrounds, and provide research ideas for target detection and recognition in complex backgrounds. Moreover, the method of the present invention simulates the spectral characteristics and scattering intensities of targets in five typical complex backgrounds, namely woodland, desert, grassland, highway and snowfield, and provides technical inspiration for other complex backgrounds with actual needs. Improve the recognition probability of targets in complex backgrounds, and at the same time reveal the light bidirectional reflection distribution of targets in all directions in space in complex backgrounds, and provide research ideas for target detection and recognition in complex backgrounds. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0015] Figure 1 It is the modeling diagram of the grassland background of the present invention; Figure 2 It is the BRDF polar coordinate diagram corresponding to the grassland background modeling diagram of the present invention; Figure 3 It is the modeling diagram of the woodland background of the present invention; Figure 4 It is the BRDF polar coordinate diagram corresponding to the woodland background modeling diagram of the present invention; Figure 5 It is the modeling diagram of the sandy land background of the present invention; Figure 6 It is the BRDF polar coordinate diagram corresponding to the sandy land background modeling diagram of the present invention; Figure 7 It is the modeling diagram of the snowfield background of the present invention; Figure 8 It is the BRDF polar coordinate diagram corresponding to the snowfield background modeling diagram of the present invention; Figure 9 It is the modeling diagram of the highway background of the present invention; Figure 10 It is the BRDF polar coordinate diagram corresponding to the highway background modeling diagram of the present invention; Figure 11 It is the schematic diagram of the modeling result of the jeep of the present invention; Figure 12 It is the schematic diagram of the modeling details of the jeep of the present invention; Figure 13 It is the simulation diagram of the target under the woodland background of the present invention; Figure 14Schematic diagram of the target simulation result before the camouflage of the present invention; Figure 15 Schematic diagram of the target simulation result after the camouflage of the present invention; Figure 16 Curve graph of the spectral reflectance of the target and grassland of the present invention; Figure 17 Schematic diagram of the target simulation result under the grassland background of the present invention; Figure 18 Flowchart of a method for detecting and identifying a target under a complex background of the present invention. Detailed implementation manners

[0016] Next, the technical solutions of the present invention will be described clearly and completely with reference to the accompanying drawings. Obviously, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0017] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. is based on the orientation or positional relationship shown in the accompanying drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation to the present invention. In addition, the terms "first", "second", "third" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0018] In the description of the present invention, it should be noted that unless otherwise clearly specified and limited, the terms "installed", "connected", "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0019] In addition, the technical features involved in different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0020] Embodiment 1: As Figure 18 This embodiment discloses a method for detecting and identifying a target under a complex background, including the following steps: Determine the application scenario and obtain the scene simulation parameters; Use 3D modeling software to process the scene simulation parameters and construct a three-dimensional simulation scene; Use the pre-constructed background model to simulate the spectral characteristics, scattering intensity, and polarization characteristics of specific targets and various complex backgrounds; Perform scene simulation based on the scene simulation parameters, three-dimensional simulation scene, and background model, and further obtain the simulated grayscale image; Compare the grayscale of the simulated grayscale image and the measured image of the application scene, and use the target recognition algorithm to identify and detect specific targets whose existence in the application scene is uncertain, and obtain the simulation scene and the target recognition result.

[0021] Specifically: During the simulation process, it also includes simulating and modeling the application scene according to the reflectivity of the specific target material itself and the spatial coordinates of the three-dimensional simulation scene.

[0022] In this embodiment, obtaining the simulated grayscale image specifically includes the following steps: Introduce the solar incident direction for spatial calculation, and perform full-space traversal calculation according to the BRDF function to obtain the full-space BRDF value of the specific target; Set the observation direction, calculate the BRDF value of each surface element of the specific target material at the corresponding angle, and obtain the simulated grayscale image of the simulation scene.

[0023] The pre-constructed background model includes: forest land pBRDF model, sandy land pBRDF model, grassland pBRDF model, highway pBRDF model, and snowfield BRDF model corresponding to various complex backgrounds; Snow albedo is an important parameter in the study of local or global energy budget balance and climate change. Remote sensing inversion provides a convenient means for obtaining snow albedo. The size of snow albedo mainly depends on the physical properties of the snow itself and the weather conditions. Existing remote sensing inversion of albedo is usually based on the bidirectional reflectance model BRDF, and the snowfield BRDF model is often obtained using the snow radiation transfer model.

[0024] Research shows that the weak absorption characteristics of snow are mainly affected by factors such as snow particle size, ice crystal shape, and pollutants. Among them, the particle size mainly affects the absorption in the near-infrared, and pollutants mainly affect the absorption in the visible light. The snowfield BRDF model algorithm used in this embodiment first inverts the snow albedo algorithm based on MODIS surface reflectance data, verifies the inversion results, and finally analyzes the reliability of MODIS snow albedo inversion.

[0025] Currently, computational simulation models are mainly used to analyze the BRDF reflection characteristics of snow. The computer simulation model is a practical processing algorithm, and the main representative model of this type is the kernel-driven model. The Monte Carlo method is the theoretical core of this model, which mainly uses a computer to simulate some given distribution functions and is a mathematical method with statistical characteristics. The kernel-driven model uses three kernels to describe three types of scattering. By weighting these three kernels, the expression of the kernel-driven BRDF model is obtained as follows: ; where are the incident zenith angle, incident azimuth angle, scattering zenith angle, and scattering azimuth angle, respectively; represents the band the corresponding BRDF value of the ground surface under different land cover at; is the geometric optics kernel; the volume scattering kernel is ; , , are the weights of isotropic scattering, geometric target scattering, and volume scattering, respectively; ; ; Among them, ； ; ; ; ; ; Since , , the three parameters are unknown, the MODIS satellite is used to obtain these three unknown parameters.

[0026] In the formula: is the scattering phase angle, represents the microstructural incident zenith angle, represents the microstructural reflection zenith angle, represents the incident angle of the micro-surface, D represents the geometric occlusion factor, b l represents the semi-major axis radius of the ellipsoid, l has no special definition and is only used to distinguish from b; r represents the semi-minor axis radius of the ellipsoid, t is the position of the intersection of the two ellipses, h represents the relative height, represents the relative azimuth angle, is the wavelength, represents the overlap function between the observed shadow and the illuminated shadow.

[0027] In this embodiment, a five-parameter semi-empirical model is specifically adopted for the snow BRDF model. Its formula is as follows: ; In the formula, it is composed of the specular reflection component term and the diffuse reflection component term of the BRDF on the sample surface; is the normal distribution function of the small facets on the sample surface, is the Fresnel approximation function, is the masking function; represents the BRDF value of the snow model; In this embodiment , , , b are actually four undetermined parameters, and are respectively used to reflect the magnitudes of the specular reflection and diffuse reflection components, and are related to the roughness and reflectivity of the target surface; reflects the slope distribution of the target surface and is related to the roughness and texture distribution of the target surface; b is a parameter reflecting the Fresnel reflection coefficient of the target surface and is related to the refractive index of the target; represents the angle between the light ray from a point on the aperture to the observation point and the aperture normal; represents the angle between the target surface normal and the microfacet normal.

[0028] Sandy land is land covered with sand on the surface and basically without vegetation. The surface of sandy land is flat and rough, and the sand grains have uniform and regular shapes. Therefore, in this embodiment, a polarization reflection model of a rough surface is considered when measuring sandy land.

[0029] The expression of the sandy land pBRDF model is: ; ; Among them, is the specular reflection component of the sandy land, is the normal distribution function, and θ represents the angle between the incident light and the reflected light; is the masking factor, is the degree of polarization, is a 4×4 Fresnel reflection Mueller matrix; A is a fitting parameter, β represents the angle between the reflection direction and the normal direction, B is a fitting parameter, represents the refractive index, represents the reflected S wave, represents the reflected P wave, represents the BRDF value of the sandy land model.

[0030] Each blade of grass in the grassland has the characteristics of irregular shape and mutual occlusion. At the same time, from the detection perspective, the observation distance is usually far, and the overall state is relatively uniform. Therefore, in this embodiment, the grassland pBRDF model uses a six-parameter pBRDF model.

[0031] The expression of the grassland pBRDF model is: ; Among them, represents the slope variance of the rough surface; is the specular reflection coefficient; is the reflectivity of the surface; is the free scattering coefficient; represents the free fitting parameter, represents the six-parameter model; represents the depolarization matrix of volume scattering, represents the depolarization matrix of backscattering, is the fitting factor.

[0032] The grassland pBRDF model is decomposed into three parts, and the first term: ; is the specular reflection term of the grassland background, representing the reflection amplitude of the whole grassland in the specular reflection direction of the incident light source.

[0033] The second term: ; is the volume scattering term of the grassland background, representing the reflection relationship between grass plants.

[0034] The third term: ; is the backscattering term, representing the hot spot effect of natural features, which cannot be ignored in this practical application.

[0035] In the model of the forest background, compared with the grassland background, the trees in the forest grow lushly, with irregular growth directions and shapes, and the branches occlude each other. They are natural ground objects growing. Compared with the grass clump distribution in the grassland, the tree distribution in the forest is relatively sparse, and the occlusion between branches and leaves is serious.

[0036] Compared with the model in the grassland background, for the polarization BRDF model in the forest background, the specular reflection term and the backscattering term in the model remain unchanged, while the volume scattering term changes.

[0037] The expression of the forest pBRDF model is: = ； ; ; Where: represents the forestland pBRDF model, represents the diffuse component of the forest floor pBRDF model, is the surface autocorrelation length function, Represents the BRDF value of the specular reflection component.

[0038] The road background has similar and different characteristics compared to the sand. The similarity is that cement and asphalt roads are smooth and the particles are highly similar, so their specular reflection and volume reflection terms remain unchanged on the basis of the sand background. The difference is that the road is a man-made object rather than a natural object, but because its hot spot effect is very weak, its backscattering can be ignored.

[0039] The highway pBRDF model expression is: .

[0040] Related results are shown below: Background scene modeling and BRDF simulation: The various complex backgrounds include: woodland, desert, grassland, road and snow.

[0041] According to the material, characteristics, roughness, etc. of five backgrounds: grassland, forest, snow, desert, and road, simulation modeling is carried out. The background modeling and the corresponding BRDF polar coordinates are as follows Figures 1 - 10 As shown: In the simulation process, a variety of modeling models are actually applied, such as Figure 11 The jeep model is used as an example. The jeep model uses meshing to depict various components and details. Since there is only one target in the following example, the jeep is also a specific target in the following example. The total number of meshes in the simplified jeep model is 400,000. The meshing diagram of the jeep model is as follows: Figure 12 As shown: The target is coupled to the background and simulated with the BRDF: like Figure 13 This is a simulation image of a target against a forest background. Figure 14 This is a schematic diagram of the target simulation results before camouflage. Figure 15 This is a schematic diagram of the simulation results of the camouflaged target. Figure 16 Spectral reflectance curves of target and grass.

[0042] When the zenith angle and azimuth angle are set to 45° and 10° respectively, the simulation results of 500nm wavelength are as follows Figure 17 As shown; Figure 17 Schematic diagram of target simulation results against a grass background.

[0043] Target recognition algorithm: In the target recognition algorithm of this embodiment, three recognition algorithms are constructed according to different application scenarios, namely: 1. Target recognition algorithm based on Sobel edge detection: In the target recognition algorithm based on Sobel edge detection, the Sobel operator detects edges based on the phenomenon that the weighted difference in gray levels of the upper, lower, left, and right neighboring pixels of a pixel reaches an extreme value at the edge. This has a smoothing effect on noise and provides relatively accurate edge direction information for calculating the gradient of the image to identify edges and structures in the image; and the target recognition algorithm based on Sobel edge detection effectively extracts the edge information of the image through convolution operations in the spatial domain, especially suitable for detecting horizontal and vertical edges.

[0044] The image input to the Sobel operator is the grayscale image of the measured image, and the simulated grayscale image obtained in this embodiment is also a grayscale image. The target edge refers to the place where the pixel gray level changes most significantly in the image, and the Sobel operator for edge detection uses the sudden change in the gray level of the image edge to detect the edge.

[0045] The Sobel operator contains two groups of 3x3 filters, which are sensitive to edges in the horizontal and vertical directions respectively. By convolving the input image with these two operators respectively, the convolution results at each point in the x and y axis directions are obtained: the vertical gradient Gx and the horizontal gradient Gy. Further, a sum-of-squares operation is performed: if the gradient G of a certain point (x, y) is greater than a preset threshold, then this point (x, y) is considered an edge point. However, the edge detection calculation based on the Sobel operator has a single direction and is ineffective for complex texture situations. Directly using the threshold to judge edge points will cause misjudgment of more noise points. Therefore, in this embodiment, the interference noise points are removed through the four-time morphological erosion algorithm, and finally an external rectangle is generated to obtain the target recognition frame.

[0046] 2. Target recognition algorithm based on SIFT: Specifically, it adopts Scale-invariant feature transform, abbreviated as STFT, and includes the following steps: First, search and obtain the positions of the measured images at all scales: Identify potential scale- and rotation-invariant interest points through the Gaussian differential function, then determine the positions and scales at the candidate positions by fitting the refined simulation modeling results. Next, based on the image gradient, assign one or more directions to the key points. Finally, in the neighborhood around each key point, measure the local image gradient at the selected scale to describe the key points. Similar to the object recognition algorithm based on Sobel edge detection, it also identifies the specific object contour by obtaining the image gradient, and then reduces noise through morphological erosion to finally obtain the specific object recognition frame.

[0047] 3. Object recognition algorithm based on gray threshold segmentation: Build a simulation scenario, calculate the BRDF values of the specific object and the background in the corresponding scenario, generate threshold parameters based on the BRDF values of the specific object and the background and the observation direction, then filter the background in the real image through the threshold parameters, and finally denoise the remaining background to generate the object recognition frame.

[0048] Obviously, the above embodiments are merely examples given for clear illustration and not limitations on the implementation manners. For those of ordinary skill in the art, other different forms of changes or variations can be made based on the above description. It is not necessary and impossible to enumerate all the implementation manners here. And the obvious changes or variations derived therefrom are still within the protection scope of the present invention.

Claims

1. A method for target detection and recognition under complex backgrounds, characterized in that, Including the following steps: Determine the application scenario and obtain the scene simulation parameters; Use 3D modeling software to process the scene simulation parameters to construct a three-dimensional simulation scene; Use the pre-constructed background model to simulate the spectral characteristics, scattering intensity, and polarization characteristics of specific targets and various complex backgrounds; Perform scene simulation based on the scene simulation parameters, three-dimensional simulation scene, and background model, and further obtain the simulated grayscale image; Compare the grayscale of the simulated grayscale image and the measured image of the application scene, and use the target recognition algorithm to identify and detect specific targets whose existence in the application scene is uncertain, to obtain the simulation scene and the target recognition result; 2. The method for target detection and recognition in a complex background according to claim 1, wherein, During the simulation process, it also includes simulating and modeling the application scene according to the reflectivity of the specific target material itself and the spatial coordinates of the three-dimensional simulation scene; 3. A method for target detection and recognition in a complex background according to claim 1, characterized in that Obtain the simulated grayscale image, specifically including the following steps: Introduce the solar incident direction for spatial calculation, and calculate the BRDF value of the entire space of the specific target through full-space traversal calculation according to the BRDF function; Set the observation direction, calculate the BRDF value of each surface element of the specific target material at the corresponding angle, and obtain the simulated grayscale image of the simulation scene; 4. A method for target detection and recognition in a complex background according to claim 1, characterized in that, The various complex backgrounds specifically include: forest land, desert, grassland, highway, and snowfield; 5. A method for target detection and recognition in a complex background according to claim 4, characterized in that, The background model includes: forest land pBRDF model, sandy land pBRDF model, grassland pBRDF model, highway pBRDF model, and snowfield BRDF model corresponding to various complex backgrounds; Among them, the expression of the snowfield BRDF model is: ； Expression represents the specular reflection component of the BRDF of the sample surface; is the normal distribution function of the small facets on the sample surface, is the Fresnel approximation function, is the masking function; represents the BRDF value of the snow model; and and a, b are undetermined parameters; Specifically: and respectively reflect the magnitudes of the specular reflection and diffuse reflection components, and are related to the roughness and reflectivity of the target surface; reflects the slope distribution of the target surface and is related to the roughness and texture distribution of the target surface; b is a parameter reflecting the Fresnel reflection coefficient of the target surface and is related to the refractive index of the target; are the incident zenith angle, incident azimuth angle, scattering zenith angle, and scattering azimuth angle, respectively; represents the angle between the light ray from a point on the aperture to the observation point and the aperture normal; represents the angle between the target surface normal and the microfacet normal.

6. The method for target detection and recognition in a complex background according to claim 5, wherein The expression of the sandy land pBRDF model is: ； ； Among them, is the specular reflection component of the sand surface, is the normal distribution function, where θ represents the angle between the incident light and the reflected light; σ represents the surface roughness; is the masking factor, represents the relative azimuth angle; is the degree of polarization, represents the refractive index, is a 4×4 Fresnel reflection Mueller matrix; A is a fitting parameter, β represents the angle between the reflection direction and the normal direction, B is a fitting parameter, represents the reflected S wave, represents the reflected P wave, represents the BRDF value of the sand surface model.

7. A method for target detection and recognition in a complex background according to claim 6, characterized in that The expression of the grassland pBRDF model is: ； Among them, represents the slope variance of the rough surface; is the specular reflection coefficient; is the reflectivity of the target surface; is the free scattering coefficient; represents the free fitting parameter, represents the six-parameter model; represents the depolarization matrix of volume scattering, represents the depolarization matrix of backscattering; represents the fitting factor.

8. A method for target detection and recognition in a complex background according to claim 7, characterized in that, The expression of the forest land pBRDF model is: = ； ； ； In the formula: represents the forest land pBRDF model, represents the diffuse reflection component of the forest land pBRDF model, is the surface autocorrelation length function, represents the BRDF value of the specular reflection component.

9. A method for target detection and recognition in a complex background according to claim 8, characterized in that The expression of the highway pBRDF model is: 。

Images