Search

CN-121999362-A - Rapid two-dimensional cfar detection method for sar imaging identification

CN121999362ACN 121999362 ACN121999362 ACN 121999362ACN-121999362-A

Abstract

The invention discloses a quick two-dimensional cfar detection method for sar imaging identification, relates to the field of imaging, and aims to solve the problems of high time complexity, insufficient detection precision, large calculation amount and long operation time of a traditional cfar algorithm. The method comprises the steps of dividing a sar two-dimensional amplitude image into four areas, calculating the area of a detection window annular area of a central area, calculating cfar a threshold value based on the area of the detection window annular area, performing sliding window processing on a two-dimensional integral image of the central area to obtain detection windows and values of each detection point, calculating cfar values of each detection point based on cfar the threshold value and the detection windows and values, comparing each detection point of the sar two-dimensional amplitude image of the central area with cfar values of each detection point to obtain a binary image of the central area, and performing cfar detection on a four-corner boundary area, a left-right boundary area and an upper-lower boundary area to obtain the binary image of the four-corner boundary area, the left-right boundary area and the upper-lower boundary area respectively. The invention improves the detection accuracy and shortens the calculation time of cfar.

Inventors

  • ZHANG YIMENG
  • CHANG BOBO
  • Weng xuan
  • CHEN JIAYING
  • LI YIHE
  • ZHENG ZHIDONG

Assignees

  • 北京遥感设备研究所

Dates

Publication Date
20260508
Application Date
20251230

Claims (10)

  1. 1. A method for rapid two-dimensional cfar detection of sar imaging identification, comprising: Converting a two-dimensional distance Doppler sar image into a sar two-dimensional amplitude image, and dividing the sar two-dimensional amplitude image into four areas, wherein the four areas are a central area, a four-corner boundary area, a left boundary area, a right boundary area, an upper boundary area and a lower boundary area; Calculating the area of the annular region of the detection window of the central region based on the Doppler dimension value range of the central region, and calculating cfar threshold values based on the area of the annular region of the detection window; The method comprises the steps of processing an integral graph of a sar two-dimensional amplitude image of a central area to obtain a two-dimensional integral graph, processing a sliding window of the two-dimensional integral graph to obtain a detection window sum value of each detection point, calculating cfar values of each detection point based on cfar threshold values and the detection window sum value, comparing each detection point of the sar two-dimensional amplitude image of the central area with cfar values of each detection point to obtain a binary graph of the central area; cfar detection is performed on the quadrangular boundary region, the left and right boundary regions and the upper and lower boundary regions to obtain binary maps of the quadrangular boundary region, the left and right boundary regions and the upper and lower boundary regions respectively.
  2. 2. The method for rapid two-dimensional cfar detection of sar imaging identification according to claim 1, wherein obtaining the doppler range of the central region based on the range of the central region comprises: Firstly, acquiring a distance direction value range of a central area: [(Pro_Dis_L+Dest_Dis_L-1),M-(Pro_Dis_L+Dest_Dis_L-1)]; wherein Pro_Dis_L is the half window length of the protection window on the distance dimension of the central area, M is the distance dimension length, and dest_Dis_L is the half window length of the distance dimension detection window of the central area; the Doppler value range of the central area is as follows: [(Pro_Dpl_L+Dest_Dpl_L-1),N-(Pro_Dpl_L+Dest_Dpl_L-1)]; where Pro_Dpl_L is the half window length of the Doppler guard window, N is the Doppler length, and dest_Dpl_L is the half window length of the Doppler detection window.
  3. 3. The sar imaging identification rapid two-dimensional cfar detection method of claim 2, wherein the method of calculating cfar threshold comprises: Firstly, calculating the area of the annular area of the detection window: cen_win_S=((Det_L+Pro_L)×2+1)×((Det_L+Pro_L)×2+1)-(Pro_L×2+1)×(Pro_L×2+1); Wherein cen_win_s is the annular area of the detection window, det_l is the length of the Doppler detection window, and Pro_l is the length of the Doppler protection window; then calculating cfar a threshold value based on the detection window annular region area; Wherein CFAR_PFA is CFAR constant false alarm factor, cen_win_S is the annular area of the detection window, cen_pfa is CFAR threshold.
  4. 4. The method for rapid two-dimensional cfar detection by sar imaging recognition according to claim 3, wherein the step of performing sliding window processing on the two-dimensional integral map to obtain a detection window sum value for each detection point comprises: sliding window processing is carried out on the two-dimensional integral graph, and the outer window sum of each detection point is calculated: sum1 i ＝rb i +lu i -ru i -lb i ; Wherein rb i is the value of the bottom right point of the i-th detection point outer window in the two-dimensional integral graph, lu i is the value of the top left point of the i-th detection point outer window in the two-dimensional integral graph, ru i is the value of the top right point of the i-th detection point outer window in the two-dimensional integral graph, lb i is the value of the bottom left point of the i-th detection point outer window in the two-dimensional integral graph, sum1 i is the sum of the i-th detection point outer windows; sliding window processing is carried out on the two-dimensional integral graph, so that an inner window sum of each detection point is obtained; the detection window sum value of each detection point is: sum i ＝sum1 i -sum2 i ; Where sum i is the detection window sum of the ith detection point, sum 1 is the outer window sum of the ith detection point, sum2 i is the inner window sum of the ith detection point.
  5. 5. The sar imaging identification rapid two-dimensional cfar detection method of claim 4, wherein calculating cfar values for each detection point based on cfar threshold and detection window sum values comprises: cfar_y i ＝sum i ×cen_pfa; Wherein sum i is the detection window and value of the ith detection point in the central area, cen_pfa is the cfar threshold value of the ith detection point in the central area, and cfar _y i is the cfar value of the ith detection point in the central area.
  6. 6. The sar imaging identification rapid two-dimensional cfar detection method of claim 5, wherein performing cfar detection on the four corner boundary region comprises: The size of the four corner boundary areas is as follows: cfar_L=((Det_L+Pro_L)*2+1)×((Det_L+Pro_L)*2+1); wherein Det_L is the length of the detection window, pro_L is the length of the protection window, cfar _L is the size of the four corner boundary region; Calculating cfar detection values of four corner points of the four-corner boundary area based on cfar values of each detection point of the central area; lu_th=sum [0 x*cen_pfa; ru_th=sum [N-2×(Det_l+Pro_l)-1] ×cen_pfa; lb_th=sum [(M-2×(Det_l+Pro_l-1))*N] ×cen_pfa; rb_th=sum [(M-2×(Det_l+Pro_l)-1)*N+N-2×(Det_l+Pro_l)-1] ×cen_pfa; Wherein lu_th is cfar detection value of upper left corner, ru_th is cfar detection value of upper right corner, lb_th is cfar detection value of lower left corner, rb_th is cfar detection value of lower right corner, M represents length of two-dimensional distance doppler sar image distance direction, N represents length of two-dimensional distance doppler sar image direction, det_l is length of detection window, pro_l is length of protection window, sum [0] is outer window sum of 0 th detection point, sum [N-2×(Det_l+Pro_l)-1] is outer window sum of (det_l+pro_l) -1 detection point, sum [(M-2×(Det_l+Pro_l-1))*N] is outer window sum of (M-2× (det_l+pro_l) -1) N detection point, sum [(M-2×(Det_l+Pro_l)-1)*N+N-2×(Det_l+Pro_l)-1] is outer window sum of (M-2× (det_l+pro_l) -1) n+n-2× (det_l+pro_l) -1 detection point.
  7. 7. The sar imaging identification rapid two-dimensional cfar detection method of claim 6, wherein the cfar detection of the four corner boundary region further comprises: comparing each detection point of the upper left corner with a cfar detection value lu_th of the upper left corner respectively, if the detection point is larger than lu_th, setting the pixel value of the detection point to be 1, otherwise, setting the pixel value to be 0, and obtaining a binary image of the upper left corner; comparing each detection point of the upper right corner with a cfar detection value ru_th of the upper right corner respectively, if the detection point is larger than ru_th, setting the pixel value of the detection point to be 1, otherwise, setting the pixel value to be 0, and obtaining a binary image of the upper right corner; comparing each detection point of the lower left corner with a cfar detection value lb_th of the lower left corner respectively, if the detection point is larger than lb_th, setting the pixel value of the detection point to be 1, otherwise, setting the pixel value to be 0, and obtaining a binary image of the lower left corner; And comparing each detection point of the lower right corner with a cfar detection value rb_th of the lower right corner respectively, setting the pixel value of the detection point to be 1 if the detection point is larger than rb_th, otherwise, setting the pixel value to be 0, and obtaining a binary image of the lower right corner.
  8. 8. The sar imaging recognition rapid two-dimensional cfar detection method of claim 7, wherein cfar detecting left and right boundary areas comprises: The left boundary region detection window adopts a right half window cfar for detection, and the right boundary region detection window adopts a left half window cfar for detection; the size of a sliding window adopted for cfar detection of the left and right boundary areas is as follows: dpl_win_S=(Det_L+Pro_L+1)×((Det_L+Pro_L)×2+1); Wherein dpl _win_s is the size of a sliding window used for cfar detection of the left and right boundary regions, det_l is the length of the detection window, and pro_l is the length of the protection window; Comparing all detection points in the left and right boundary areas with the detection value cfar corresponding to the detection point, if the detection value is larger than the cfar detection value, setting the detection value to be 1, otherwise setting the detection value to be 0, and thus obtaining a binary image of the upper half area.
  9. 9. The method for rapid two-dimensional cfar detection of sar imaging recognition according to claim 8, wherein the step of cfar detection of the left boundary region specifically comprises: The outer window sum of the b-th detection point of the first row of the left boundary area is: huge_win_sum[b]=huge_win_sum[b-1]-huge_sum_col[b]+huge_sum_col[b+Pro_L+Det_L]; wherein, the huge_win_sum [ b ] is the outer window sum of the b-th sliding window block of the first row, the huge_win_sum [ b-1] is the outer window sum of the b-1-th sliding window block of the first row, the huge_sum_col [ b ] is the outer window sum of the detection points of the b-th column of the first row, and the huge_sum_col [ b+Pro_L+det_L ] is the outer window sum of the detection points of the b+pro_L+det_L column of the first row; the inner window sum of the b-th detection point of the first row of the left boundary area is: pro_win_sum[b]=pro_win_sum[b-1]+small_sum_col[b+Pro_L]-sumall_sum_col[b]; Wherein pro_win_sum [ b ] is the inner window sum of the b-th detection point of the first row, pro_win_sum [ b-1] is the inner window sum of the b-1-th detection point of the first row, small_sum_col [ b+Pro_L ] is the inner window sum of the b+Pro_L column of the first row, sumall _sum_col [ b ] is the inner window sum of the b-th column of the first row; the sum of detection windows corresponding to the b-th detection point of the first row in the left boundary area is as follows: det_win_sum[b]=huge_win_sum[b]-pro_win_sum[b]; The huge_win_sum [ b ] is the outer window sum of the b-th detection point of the first row, and the pro_win_sum [ b ] is the inner window sum of the b-th detection point of the first row; the outer window sums of detection points of the kth column of the q-th row except the first row of the left boundary region are: huge_sum_col[k]=huge_sum_col[k-1]-cfar_sb[(q-Pro_L-Dis_L-1)×N+k]+cfar_sb[(q+Pro_L+Dis_L)×N+k]; huge_sum_col [ k ] is the outer window sum of the kth row and the kth column except the first row, huge_sum_col [ k-1] is the outer window sum of the kth row and the kth-1 column except the first row, cfar _sb [ (q-Pro_L-Dis_L-1) x N+k ] is the pixel value of the (q-Pro_L-Dis_L-1) x N+k point on the sar image of the left boundary region, N represents the length of the two-dimensional distance Doppler sar image azimuth, cfar _sb (q+Pro_L+Dis_L) x N+k is the pixel value of the (q-Pro_L-Dis_L-1) x N+k point on the sar image of the left boundary region; The inspection point inner window sums of the kth column of the q-th row except the first row in the left boundary area are: small_sum_col[k]=small_sum_col[k-1]-cfar_sb[(q-Pro_L-1)×N+k]+cfar_sb[(q+Pro_L)×N+k]; Wherein small_sum_col [ k-1] is the inner window sum of the (q-Pro_L-1) th row and the (k-1) th column except the first row, small_sum_col [ k ] is the inner window sum of the (q-row and the k-th column except the first row, cfar _sb is the length of the sar image N representing the azimuth direction of the two-dimensional range Doppler sar image, cfar _sb [ (q-Pro_L-1) x N+k ] is the (q-Pro_L-1) x N+k pixel point on the sar image of the left boundary region, cfar _sb [ (q+Pro_L) x N+k ] is the [ (q+Pro_L) x N+k ] pixel point on the sar image of the left boundary region; the detection point corresponding detection window sum value of the kth column of the qth row except the first row in the left boundary area is: det_win_sum[k]=huge_sum_col[k]-small_sum_col[k]; The half window detection area of the left boundary area is: S_det_dpl=((det_L+Pro_L)×2+1)×(det_L+Pro_L+1)-(Pro_L+1)×(Pro_L×2+1); The cfar detection values of the detection points in the left half area are: Wherein dpl _hf_ pfab is the cfar detection value of the b-th detection point in the left boundary region, cfar _sb is the detection point of the sar image of the left boundary region, and S_det_ dpl is the region area of the left boundary region; Comparing all detection points in the left boundary area with the detection value cfar corresponding to the detection point, if the detection value is larger than the detection value cfar, setting the detection value to be 1, otherwise setting the detection value to be 0, and obtaining a binary image of the upper half area.
  10. 10. The sar imaging identification rapid two-dimensional cfar detection method of claim 9, wherein the cfar detection of the upper and lower bounding regions comprises: The upper boundary region detection window adopts a lower half window cfar for detection, and the lower boundary region detection window adopts an upper half window cfar for detection; the sliding window of the upper and lower boundary areas has the following size: cfar_dis_win=(Det_L+Pro_L+1)×((Det_L+Pro_L)×2+1); Wherein det_l is the detection window length, pro_l is the protection window length; Adding the sums of each column in the first sliding window block of the first row to obtain the outer window sum of the first sliding window of the first row, sum_win_dis 1 ; Obtaining an outer window sum of a second sliding window based on an outer window sum of a first sliding window of the first row; The outer window sum of the m+1th sliding window of the first row is: sum_win_dis m+1 ＝sum_win_dis m +dis_sum_col[m+Det_L+Pro_L+1]-dis_sum_col[m-Det_L+Pro_L]; Wherein sum_win_dis m+1 is the outer window sum of the m+1th sliding window of the first row in the upper boundary region, sum_win_dis m is the outer window sum of the m-th sliding window of the first row in the upper boundary region, dis_sum_col [ m+det_l+pro_l+1] the outer window sum of the m+det_l+pro_l+1 column in the first row sliding window block of the upper boundary region, dis_sum_col [ m-det_l+pro_l ] is the outer window sum of the m-det_l+pro_l column in the first row sliding window block of the upper boundary region; The sum of the protection windows of the first row and the second row is obtained based on the sum of the protection windows of the first row and the first sliding window, the sum of the protection windows of the first row and the third sliding window is obtained based on the sum of the protection windows of the second sliding window of the first row, the sum of the (m+1) th protection windows of the first row is obtained based on the sum of the protection windows of the (m) th sliding window of the first row, and the sum of the protection windows corresponding to the (m) th detection point of the first row is sum_small_win_dis m ; the outer window sums of the detection points of the t column of the r row except the first row of the upper boundary region are: dis_sum_col[t]=dis_sum_col[t-1]-cfar_sm[(r-1)×N+t]+cfar_sm[(r+Det_L+Pro_L)*N+t]; Wherein dis_sum_col [ t ] is the outer window sum of the t-th column of the r-th row except the first row of the upper boundary region, dis_sum_col [ t-1] is the outer window sum of the t-1 th column of the r-th row except the first row of the upper boundary region, cfar _sm [ (r-1) ×n+t ] is the pixel value of the (r-1) ×n+t point on the sar image of the upper boundary region, cfar _sm [ (r+det_l+pro_l) ×n+t ] is the pixel value of the (r+det_l+pro_l) ×n+t point on the sar image of the upper boundary region; The inner window sums of the detection points of the nth row and the nth column except the first row of the upper boundary area are: dis_sum_small_col[t]=dis_sum_small_col[t-1]-cfar_sm[(r-1)×N+t]+cfar_sm[(r+Det_L+Pro_L)*N+t]; Wherein dis_sum_small_col [ t ] is the inner window sum of the r-th row and t-th column except for the first row, dis_sum_small_col [ t-1] is the inner window sum of the r-th row and t-1 column except for the first row, cfar _sm [ (r-1) x n+t ] is the pixel value of the (r-1) x n+t th point on the upper boundary region sar image, cfar _sm [ (r+det_l+pro_l) xn+t ] is the pixel value of the (r+det_l+pro_l) x n+t point on the upper boundary region sar image; The detection window and value sum_det_dis m of the detection points of the first row of the upper boundary region are: sum_det_dis m ＝sum_win_dis m -sum_small_win_dis m ; The detection window and value sum_det_dis t of the detection points except for the first row of the upper boundary region are: sum_det_dis t ＝dis_sum_small_col[t]-sum_det_dis t ; the upper boundary area half window detection area is: S_det_dis=((det_L+Pro_L)×2+1)×(det_L+Pro_L+1)-(Pro_L+1)×(Pro_L×2+1); The cfar detection value of the detection point of the upper boundary region is: Wherein cfar _sm is the pixel value of the mth point on the upper boundary region sar image, dis_hf_pfa m is the cfar detection value of the mth detection point of the upper boundary region, S_det_dis is the half window detection area of the upper boundary region; and for any detection point of the upper and lower boundary areas, if the amplitude value of the detection point is larger than the corresponding cfar detection value, setting the detection point as 1, otherwise, setting the detection point as 0.

Description

Rapid two-dimensional cfar detection method for sar imaging identification Technical Field The document relates to the field of imaging, in particular to a quick two-dimensional cfar detection method for sar imaging identification. Background Sar is an actively imaged sensor synthetic aperture radar. The radar has the characteristics of all weather, all-day time and strong penetrating power, and has wide application in civil and military fields. Under the condition of multiple targets, the radar system can measure the distance and the radial speed of the targets at the same time, and a two-dimensional distance-Doppler spectrum matrix is obtained after the received echo signals are subjected to correlation processing. During the detection phase, a constant false alarm (cfar) detector is typically used to find regions of interest of the object that may be present from a wide range of image regions. cfar is designed to provide a detection threshold value that can relatively avoid the influence of noise background clutter and interference variation, and the currently commonly used detection methods include ca-cfar, os-cfar, double parameters cfar, and the like, and the most commonly used detection method is double parameters cfar detection, but for a sar image with high resolution, the size of a rectangular ring window is often selected to be larger, the double parameters cfar detect each pixel point by using the sliding window, and the mean value and variance of the background are counted, so that larger calculation amount and longer calculation time are required. Therefore, a quick two-dimensional cfar detection method for sar imaging identification is provided. Disclosure of Invention The present disclosure provides a quick two-dimensional cfar detection method for sar imaging recognition, which is used for solving the problems of high time complexity and insufficient detection precision of the traditional cfar algorithm, for example, a two-dimensional cfar detection algorithm adopts a sliding window to detect each pixel point, and statistics is performed on the mean value and variance of the background, so that a relatively large calculation amount and a relatively long operation time are required, and the method comprises: Converting a two-dimensional distance Doppler sar image into a sar two-dimensional amplitude image, and dividing the sar two-dimensional amplitude image into four areas, wherein the four areas are a central area, a four-corner boundary area, a left boundary area, a right boundary area, an upper boundary area and a lower boundary area; Calculating the area of the annular region of the detection window of the central region, and calculating cfar a threshold value based on the area of the annular region of the detection window; The method comprises the steps of processing an integral graph of a sar two-dimensional amplitude image of a central area to obtain a two-dimensional integral graph, processing a sliding window of the two-dimensional integral graph to obtain a detection window sum value of each detection point, calculating cfar values of each detection point based on cfar threshold values and the detection window sum value, comparing each detection point of the sar two-dimensional amplitude image of the central area with cfar values of each detection point to obtain a binary graph of the central area; cfar detection is performed on the quadrangular boundary region, the left and right boundary regions and the upper and lower boundary regions to obtain binary maps of the quadrangular boundary region, the left and right boundary regions and the upper and lower boundary regions respectively. In a preferred embodiment, the obtaining the doppler range of the center region based on the range-wise range of the center region includes: the range of the distance direction value of the central area is as follows: [(Pro_Dis_L+Dest_Dis_L-1),M-(Pro_Dis_L+Dest_Dis_L-1)]; Wherein Pro_Dis_L is the half window length of the protection window on the distance dimension of the central area, M is the distance dimension length, Dest_dis_l is the half window length of the center region distance dimension detection window; the Doppler value range of the central area is as follows: [(Pro_Dpl_L+Dest_Dpl_L-1),N-(Pro_Dpl_L+Dest_Dpl_L-1)]; where Pro_Dpl_L is the half window length of the Doppler guard window, N is the Doppler length, and dest_Dpl_L is the half window length of the Doppler detection window. In a preferred embodiment, the method of calculating cfar the threshold comprises: Firstly, calculating the area of the annular area of the detection window: cen_win_S=((Det_L+Pro_L)×2+1)×((Det_L+Pro_L)×2+1)-(Pro_L×2+1)×(Pro_L×2+1); Wherein cen_win_s is the annular area of the detection window, det_l is the length of the Doppler detection window, and Pro_l is the length of the Doppler protection window; then calculating cfar a threshold value based on the detection window annular region area; Wherein CFAR_PFA is