Radar video detector and target tracker.

0001The invention relates generally to the art of processing radar video signals, and more particularly to simultaneously processing video signals from multiple target returns using adaptive threshold settings.

0002Detection systems for separating targets from noise have been proposed in which the detection threshold is either fixed or automatically adjusted in accordance with variations in background noise and interference. One such prior art system is described in US-A-4,005,416. The detection system proposed therein is automatically adjusted in accordance with variations in background noise and interference. The energy levels of video signals within each of three successive range cells are coupled to a comparator wherein the energy contained within a central range cell is compared with the energy contents within the two adjacent range cells. A target indication in a range cell is provided when the ratios of energy in that range cell to the energy in each of the two adjacent cells are both greater than a predetermined value. Noise and interference responses are minimised by summing the output signals of the comparator over a number of range sweeps. A target indication is provided when the sum achieves a specified number before a predetermined number of range sweeps have been completed. The existence of a target requires that both detection criteria be satisfied. This system requires continuous energy comparisons and redundant processing before an existing target is reported. Additionally, the system does not achieve noise reduction in the target area. Targets are reported when the energy in a range cell is greater than the energy in the two adjacent range cells. Thus, the detection threshold rides on the background noise. True noise reduction is not achieved. Further, this detection system does not provide for processing more than one target within a bearing sector.

0003In a prior art tracking system, disclosed in US-A-4-070,673, computer generated digital signals representative of leading and trailing edges of range and azimuth intervals defining a tracking window are compared with the instantaneous range and antenna position to provide pulses signifying the opening and closing of a tracking window. This control unit provides signals to a multiplicity of counting circuits wherein the number of pulse transmissions, range bins traversed, and radar target returns within the defined tracking windows are established. The pulse transmission and range bin counters are coupled to accumulators which are enabled by pulses from a comparator, provided when the radar video returns within the tracking window exceed a fixed signal threshold. When the accumulator associated with the pulse transmission counter is enabled, the transmission count, corresponding to the enabling radar return, is added to the total in the accumulator to achieve a weighted sum of angular returns. In a like manner, a radar return enables the accumulator associated with the range bin counter to add the range bin count, associated with the radar return, to the total in the range accumulator. These weighted sums and the radar target return count are utilised to determine the range and angle centroids of a target. Though this system provides accurate tracking of a detected target, it operates at a fixed detection threshold, requiring M of N processing to reduce the probability of false alarms. Further, the system does not have the capability of processing multiple targets on a given bearing.

0004Other prior art detection and tracking systems use either fixed thresholds for separating targets from noise or sequential methods which establish a predetermined detection or noise threshold and then increment or decre­ment the threshold in accordance with the observed target magnitudes, and noise and interference levels. Such systems provide less than optimum thresholds in a single scan, or require a multiplicity of scans to achieve an optimum threshold. Further, such prior art systems are incapable of tracking multiple targets on a bearing. Since data for only one target on a bearing can be processed on a single scan, multiple scans are required to update target information for multiple targets on a bearing. Moreover, the prior art systems are subject to switching from tracking one target to tracking a second target (target swapping) if the second target is in the vicinity of the first and its radar return exceeds that of the first.

0005The present invention, which is defined in the appended claims, processes radar video to reduce noise effects, provides variable thresholds in accordance with background noise and clutter, provides for multiple target tracking, and prevents target swapping. In a preferred embodiment, in order to enhance the radar video signals, the video of repetitive range sweeps is averaged over small sectors, to reduce the effort of noise and enhance signal amplitude. An adaptive detection threshold is provided by measuring the background noise in the vicinity of the target and computing a mean and mean deviation of the noise. This value is calculated for each target on every scan, so that the threshold computed is precise and adapts dynamically to changing conditions. Tracking multiple targets on a given bearing is accomplished by storing the video data so that the data may be subsequently processed as readily as targets on differing bearings. Finally, target swapping is reduced by a combination of adjusting the window size (a plurality of sectors) in accordance with the size of the target being tracked, and weighting the sectors so that the sectors at the edges of the window do not affect the calculations of target position as much as the central sectors.

0006Radar video signals from a target or clutter are averaged within small areas (cells) and stored in a memory for processing. These cells are elements in a matrix formed by n azimuth sectors and m range bins within present a target window, the window containing m X n cells. Stored values in the first and last rows of cells in the window are processed to establish the mean value and mean deviation value for each row. The mean and mean deviation values in the row have the smaller mean and mean deviation values are chosen for further processing. A linear combination of the chosen mean and mean deviation values is subtracted from each averaged signals to establish a new stored amplitude for each cell with significantly reduced background noise and clutter.

0007Each sector element is then compared with its neighbouring elements, and a detection is indicated in a cell when at least one of the two adjacent elements have positive resulting amplitudes.

0008The number of detections and their position within the window is then used to adjust the window size. A centre of gravity computation is then performed to centre the target within the adjusted window.

0009Target position is determined on successive antenna scans. From the difference in positions and the elapsed time interval between scans, an initial target speed and heading is computed. A recursive filter receives the computer target speed and heading as an initial value, and calculates a new estimated position. Each calculated position is compared with the estimated position from the previous scan to determine an estimated error, which is then used to establish the next estimated position. The error is quickly minimized and accurate tracking is accomplished.

0010A radar video detector and target tracker in accordance with the present invention will now be described in greater detail, by way of example, with reference to the accompanying drawings, in which:- <ul id="ul0001" list-style="none"><li>Figure 1 is a representation of a target window,</li><li>Figure 2 is a block diagram of the target detection and processing circuitry of the illustrated embodiment,</li><li>Figure 3 is a block diagram of the noise reduction unit of Figure 1, and</li><li>Figure 4 is a block diagram of the target size and position unit of Figure 1.</li></ul>

0011A target window within which a target is to be tracked is represented in Figure 1. An azimuth size for this window may comprise nine (9) azimuth sectors commencing at an azimuth A_s and termination at an azimuth A_t, each of which may be 0.35 degrees, while a range size, commencing at a range start R_s and terminating at a range stop R_t, may comprise sixteen (16) equal range bins. The number of azimuth sectors and range bins may be adjusted in accordance with a desired range size. The arrangement shown forms an array of 9 x 16 cells, each of which covers an area that is smaller than the anticipated physical size of the target so that the target will be detected in more than one cell. While an array of 9 columns x 16 rows with an azimuth sector of 0.35 degrees is shown, this is to be considered exemplary and not limiting. The actual window size is varied in accordance with the size of the target, as will be described subsequently.

0012An antenna scan rate for the system may be typically 22 rpm while the pulse repetition may be chosen to provide 640Hz, 1600Hz, or 3200Hz. Thus, for these values the number of pulses that may be emitted within each azimuth sector may vary between and one (1) and nine (9).

0013Referring now to Figure 2, with continued reference to Figure 1, analog video signals on line a 22 represent­ative of radar returns from a target are applied to a digitizer 28 of a tracking video processor 10. Radar trigger pulses on a line 24, antenna scan pulses on a line 26, representative of the azimuth rotation of an antenna not shown, and range start pulses on a line 36 provided by a tracking processor 12, are coupled to an averager 30 in the tracking video processor 10. The tracking processor 12 determines the size of the target and adjusts the window size in accordance therewith. A target position estimator 65, with the tracking processor 12, computes the range and bearing of the target and provides an estimated range and bearing at a desired future time.

0014The video data is encoded into digital words, at a sampling rate established by a sampling rate signal on a line 38 coupled from the tracking processor 12, and the digitized video data, corresponding to each cell position, is averaged over the number of consecutive radar pulses within the azimuth sector. The number of these pulses is a function of the repetition rate. Each trigger pulse denotes the start of a video data cycle. The averager 30 sums the signal amplitudes in digital from for each cell and divides the accumulated values by the number of trigger pulses within an azimuth sector on the line 24 to establish an average video value for each cell.

0015Averaged video data for each cell is stored in a video storage memory 32 comprising two memory units, wherein each storage element in each unit represents a cell of a window. While the first memory unit is accept­ing averaged data over a set of range starts in an azimuth size, the second memory unit transfers the averaged data stored therein, obtained over the previous azimuth size, to the tracking processor 12 for further processing and target tracking, yet to be described. As presently configured, the system stores and processes data for up to eight targets on a bearing and up to twenty targets for all bearings during an antenna scan.

0016Referring now to Figure 3, data transferred to the tracking processor 12 is stored a memory 41 in a noise reduction element 40. A mean and mean deviation unit 43 reads this data for each window and starts the noise reduction process by computing the mean and mean devia­tion of the values stored in the first and last rows (Y₁, Y₁₆) of the window. The mean is determined by summing the values in each cell and dividing the result by the number of cells. Mean deviation is then determined by subtracting the mean from the value in each cell in the row, summing the absolute values of the differences and dividing by the number of cells. The mean value plus a scale factor multiplied by the mean deviation in the row providing the smaller values of mean and mean deviation are then subtracted from the values in all the cells of the window. The scale factor multiplier is selected by the operator in accordance with his assessment of the extent of background clutter. This provides a clutter reduction that is based on the actual background clutter.

0017Further noise reduction is accomplished by a 2 out of 2 correlation processor 45, within the tracking processor 12, which checks the amplitude in the cells to the right and left of each cell having a positive amplitude after the initial noise reduction. If at least one of the cells abutting the central cell does not have a positive amplitude, the amplitude of the central cell is set to zero. Thus, there must be at least two adjacent cells with positive values to establish the presence of a target. Since each cell covers an area that is smaller than the anticipated target size, a target detection must appear in two or more abutting cells. Consequently, detections in isolated cells are due to noise or clutter. This correlation process removes all stored cell values in the window that do not correlate in azimuth with one of its neighbours. It should be recognised that a similar correlation may be performed along the range axis.

0018Referring again to Figure 2, following the correla­tion process, the window is searched by a target detection processor 44 to determine the existence of a target. A target detection is verified by the target detection processor 44 when the number of correlated returns in the window exceeds a pre-established value. If a target is detected in a cell, the cell data is coupled to target size and position processor 46 to establish the target size and position in that cell. Referring to Figure 4, target size and position processor 46 includes a correlated returns counter 47 which deter­mines the target size by counting the number of correlated returns in the window. Since each cell is of known range and azimuth dimensions this count provides the size of the target. After the target size has been determined, the size data is coupled to a window size processor 49, yet to be described.

0019With reference to Figure 4, the range and azimuth positions of a target in the window are determined by summing the amplitude in all cells of a range row and summing the amplitudes of all cells in an azimuth column, weighting the range row sums by the range row position and summing to obtain a range row weighted sum, weighting the azimuth column sums by the azimuth column position to obtain a column weighted sum, and dividing the range row and azimuth column weighted sums by the sum of the amplitudes in all the cells of the window. The range row weighted sum (my)_f, azimuth column weighted sum (mx)_f, and the sum of amplitudes n_f within the window are calculated by a range weighted amplitude processor 51, an azimuth weighted amplitude processor 53, and a window amplitude sum processor 55, respectively. The sum output of window amplitude sum processor 55 is coupled to dividers 57 and 59, while the sum outputs of range weighted amplitude processor 51 and azimuth weighted processor 53 are respectively coupled to dividers 57 and 59, wherefrom the target range (my_f/n_f and azimuth (mx)_f/n_f positions in the cell are respectively provided.

0020All the sums n_f, (my)_f, and (mx)_f are accumulated by adding the value for a cell, range row, and azimuth column to the value accumulated from the addition of values from previous cells, range rows, and azimuth columns. Devoting the accumulated sums prior to the kth addition by the subscript k-1, the sums after the kth addition for the sum of the amplitudes in all the cells n, the azimuth column weighted sum mx, and the range row weighted sum my are: n_k = n_k-1 + a_k (my)_k = (my)_k-1 + J_k .a_Jk (mx)_k = (mx)_k-1 + I_ka_Ik where: a_k is the amplitude in the kth added cell J_k is the kth row position a_Jk is the sum of the amplitudes in row J_k for the kth sum I_k is the kth column position a_Ik is the sum of the amplitudes in column I_k for the kth sum.

0021The target position (my)_f/n_f, (mx)_f/n_f is coupled to a target centre of gravity processor 61 which offsets the range and azimuth positions by one half cell width in both dimension, to position the target centre of gravity at the centre of a cell, and adds the results to the range start R_s and the azimuth start A_s to obtain a range centre of gravity R_cg and an azimuth centre of gravity A_cg which are represented by: R_cg = R_s + [(my)_f/n_f + 0.5] A_cg = A_s + [(mx)_f/n_f + 0.5]

0022This target position within the window is coupled to the window size processor 49, to which the target size data is also coupled.

0023After the target size and centre of gravity in the window has been determined, the window size processor 49 adjusts the window so that the target covers between 25 and 75 percent of the window in the range dimension and positions the window along the range axis so that a predetermined number of empty range bins, which may be no less than 4, are between the target and both range start and range stop. The window size processor 49 also positions the window in azimuth to establish no less than 2 azimuth sectors on either side of the target.

0024Once the target is substantially centred, the window size processor 49 assigns a weight to each cell in the window that is to be applied to the amplitude of a detection in the cell on the subsequent sweep for that target. The assigned weighting is unity for cells in the target area and a factor less than unity for cells at and near the window borders. Since detections in the cells away from the target are given less weight than detec­tions in the target area, a second target entering the window appears less prominent than it actually is, consequently a significant reduction in the probability of target swapping is achieved.

0025With the determination of the target centre of gravity complete, target position data is coupled, via a line 63, from the target cg processor 61 to a speed and heading processor 52 wherein target speed and heading is computed by subtracting the previous computed position from the present computer position and dividing by the antenna scan time. Since the window starts must be established before the next update scan, the speed and heading processor 52 couples speed and heading represent­ative signals to a target position estimator 65. The target position estimator 65 comprises an alpha-beta filter and utilises these signals to predict the target location at the time for the next update scan and provides signals representative of the predicted location to the tracking processor 12.

0026Polar coordinate date of the present target position received by the speed and heading processor 52 is converted to rectangular coordinates and channelled for separate X and Y processing. The speed and heading processor 52 subtracts the previous measured target coordinate position from the present measured target coordinate position and divides by the antenna scan time to establish measured coordinate velocities vx and vy. The present measured coordinate positions, the present coordinate velocities and the previous predicted coordinate position and predicted coordinate velocities are utilised in the alpha-beta filter in the target position estimator 65 to establish a predicted target position and a predicted target velocity for the next antenna scan in the following manner: V_x2 = V_x1 + β (x-x₁)f_c/t_s x₂ = x₁ + α (x-x₁)f_c+(v_x1-v)t_s x = present position (x being representative of both x and y coordinates, since the same equations govern the predicted position and velocity for both). x₁ = past predicted position x₂ = present predicted position v_x1 = past predicted velocity v_x2 = present predicted velocity t_s = antenna scan time f_c = is a factor that is a function of the number of antenna scans since the last measurement (this factor permits the target position to be dead reckoned from the last measured velocity) β - 1/(f_scf_c)² f_sc = filter scan count α = √<o ostyle="single">2 β</o>

0027Should the predicted target position differ from the actual position on a number of successive update scans, the filter time constants are altered to achieve faster updates. When the actual target positions is once again within a predetermined tolerance of the predicted target position, the filter time constants are returned to their original values.

0028The target position estimator 65 provides signals representative of positions of tracked targets. These signals may be coupled to a display unit for display and to other systems for further processing.