Detailed description of the preferred embodiments
The method for identifying the low-slow small target and the sea surface small target based on the radar and the photoelectricity is implemented by the following specific steps of referring to the attached drawing 1.
(1) The radar guides the target of the photoelectricity.
The radar sends the detected track information of the target to be identified, including target batch number, direction, distance, elevation angle, course and navigational speed information to the photoelectricity, and the photoelectricity is guided to rapidly capture the target to be identified.
(2) And the radar and the photoelectricity carry out compound tracking on the target.
After the target to be identified is rapidly captured by photoelectricity, the target to be identified is tracked to form a photoelectricity target track, the photoelectricity target track including azimuth, elevation angle, course and speed information is sent to a radar, and the radar carries out track fusion on the target radar track and the photoelectricity track.
(3) And calling a radar to perform high repetition frequency detection on the target.
a) Radar target attitude resolution
Where θ is the radar target attitude angle, α is the radar target heading, β is the radar target azimuth;
b) resolving the radial movement speed of the target according to the resolved target attitude and the resolved target movement speed;
c) and selecting high repetition frequency detection parameters to perform high repetition frequency detection on the target according to the radial movement speed and the target distance of the target, so as to prevent the target from falling into a detection blind distance and blind speed.
(4) And calling photoelectricity to carry out infrared imaging on the target.
a) Photoelectrically capturing a target image, calculating a definition evaluation index of the target image, taking the sum of absolute values of 4 neighborhood gray differences as the evaluation index of image definition, and regarding an image with N multiplied by N pixels, f (x, y) represents the pixel value of a certain point in the image, and the definition evaluation index of the image is P1:
b) adjusting the focal length in one direction, then calculating a definition index P2 at the moment, if P1 is more than P2, adjusting the focal length in the opposite direction until the definition is poor, and at the moment, the state target image at the previous moment is clearest; if P1 is less than P2, the focal length is continuously adjusted towards the direction until the definition is poor, and the target image in the state at the previous moment is clearest;
c) capturing a clear target image, performing threshold transformation on the image, and performing threshold transformation on the image to obtain a binary image of the target; carrying out binarization on the image by setting a threshold value, and directly setting a gray value to be 0 for pixels with the gray value smaller than the threshold value; pixels with gray values greater than the threshold are set directly to 255. The threshold selection rule is as follows:
using f (i, j) {0 ≦ i < M,0 ≦ j < N } to represent the corresponding pixel value of the corresponding position of an M × N image, and the average value of all pixel sums of the whole image is:
let K be
The threshold R is:
(5) the method for extracting the target micro Doppler characteristics by the outlier skewness method comprises the following steps:
a) taking the target track distance as the center and taking 2 to the leftm1 distance unit, take 2 to the rightmA distance unit of 22mTaking the distance units as data samples M;
b) to 22mThe distance units are respectively subjected to FFT processing and modulus calculation;
c) to 2 after the module is solved2mA distanceThe units respectively accumulate, and a distance unit L corresponding to the accumulated maximum value is searched and used as a distance unit corresponding to a target;
d) taking out data corresponding to the distance unit L from the data sample M, carrying out FFT processing, carrying out modulus calculation, and taking a result after the modulus calculation as a target frequency spectrum K;
e) resolving a position corresponding to the maximum value in the frequency spectrum K, taking 3 units on the left and right of the position as the Doppler frequency spectrum of the target, and removing the Doppler frequency spectrum to obtain a target frequency spectrum K';
f) taking 5 distance units on the left and right of the distance unit L, respectively carrying out FFT processing and modulus calculation on data corresponding to the 5 distance units on the left and right, and obtaining a background frequency spectrum set { H }i},i=L+1,L-1,L+2,L-2,L+3,L-3,L+4,L-4,l+5,L-5;
g) Respectively calculating the skewness O 'of the target spectrum K' and the background spectrum set { HiBias set of { O }i}
Where O (X) is sample data skewness, X is target region sample data, μ is sample mean, σ is sample standard deviation, E (X- μ)3Is the 3 rd order center-to-center distance of the sample data.
h) Calculating the outlier skewness O' -max { OiIf O' -max { O }i}>And 0, the micro Doppler modulation characteristics exist in the target frequency spectrum, otherwise, the micro Doppler modulation characteristics are not detected in the target frequency spectrum.
(6) And extracting structural features of the target infrared image by a variable force field conversion method.
a) Calculating the magnitude and direction of variation resultant force suffered by pixel points in the image, wherein the position in the image is riThe pixel position of is rjThe variable gravitation F of the pixel pointsi(rj) The magnitude of the force of variation and riThe gray value of a point is proportional to the point rjAnd point riThe square of the distance between them is inversely proportional, the direction of the variation attraction, i.e. the direction of the line connecting two points, is represented by the following vector:
wherein I (r)
i) The representation position is a pixel point r
iIs determined by the gray-scale value of (a),
is a target attitude angle, r
i-r
jThe direction of the connecting line of (A) represents F
i(r
j) In the vector direction, | r
i-r
jAnd | represents the distance between two pixels. Pixel point r
jThe resultant of variation force of all pixels can be expressed as
Wherein N represents riNumber of pixels in a certain neighborhood, Fi(rj) In the direction of Fi(rj) The resultant force direction, the magnitude and direction of the resultant force of variation suffered by the pixel point are calculated as follows:
i. calculating the variation gravitation of the other point in the neighborhood of the point, and decomposing the variation gravitation along the horizontal axis and the vertical axis respectively according to the direction of the variation gravitation;
summing all other variant attractions to the point on the horizontal and vertical axes;
performing vector synthesis on the components obtained on the horizontal axis and the vertical axis to obtain the magnitude and the direction of the variation resultant force finally received by the point;
b) calculating the magnitude of the variation resultant force borne by each pixel point according to the magnitude normalization of the variation resultant force borne by all the pixel points in the image;
c) carrying out binarization processing on the normalized image to obtain an edge pixel point and a region image of a neighborhood thereof;
d) obtaining final coarse edge points according to the direction and the size of the variation resultant force, and extracting the infrared structural characteristics of the target;
e) and performing library comparison on the extracted target infrared structural features by a template matching method to obtain a target infrared structural feature comparison result.
(7) And constructing an object feature matrix.
Combining target radar motion characteristics, target photoelectric motion characteristics, target micro-Doppler modulation characteristics and target infrared image structure characteristics to construct a target characteristic matrix F epsilon R4×1
x1 represents the target attitude, x2 represents the target height calculated according to the radar track and the photoelectric track, x3 represents the target micro-Doppler modulation characteristic, and x4 represents the target infrared structural characteristic comparison result.
(8) And (4) combining a Sigmoid function to design a linear two-classifier.
The invention adopts a linear two-classifier as a classifier to judge the multi-target attribute of the radar, and the linear two-classifier consists of an input layer, an excitation function and an output node. The input layer is composed of 4 nodes, the input layer respectively corresponds to 4 characteristics of a target one-dimensional range profile energy gathering area characteristic matrix, a Sigmoid function is used as an excitation function, and the output node y is
Wherein F ∈ R4×1Is a characteristic matrix of a one-dimensional range profile energy gathering area of the target, W belongs to R4×1Weight matrix as input layer
(9) And (4) classifying and identifying the low-slow small target and the sea surface small target.
Classifying and identifying the low-slow small target and the sea surface small target according to the value of the output contact y in the step (8)