CN104155636A - Optimization method based on constant false alarm target detection - Google Patents

Abstract

The present invention discloses an optimization method based on constant false alarm target detection. The method relates to the digital signal processing technology field, and comprises the steps of on the condition that the reference units are arranged at the right or the left of a detection unit, utilizing the pointers to select a first detection unit and a second detection unit to determine a first data accumulation sum and a second first data accumulation sum, calculating a first threshold value and a second threshold value, and determining whether the detection unit is a target by the threshold values; on the condition that the references unit are arranged at the left and the right of the detection unit, utilizing the pointers to select the first and second detection units to determine the first data accumulation sum and the second first data accumulation sum, calculating the threshold value and the second threshold value, and determining whether the detection unit is the target by the threshold values. The optimization method based on the constant false alarm target detection mainly utilizes the two pointers to select the units, so that such a parallel processing method enables the power consumption to be reduced, and the execution efficiency of a piece of software to be improved.

Description

A kind of optimization method based on CFAR target detection

Technical field

The invention belongs to digital signal processing technique field, relate to a kind of optimization method based on CFAR target detection.

Background technology

Early stage radar system is all video informations that obtain directly to be delivered to display show, and distinguishes clutter or target, realize target measuring ability by operator.In this case, the detectability of radar is determined by operator's supervision, be subject to the impact of the conditions such as operator's experience very large, and manual operation is restricting the ability that Radar Multi Target is caught, followed the tracks of.Along with digitized deep, the automatic detection of target becomes a kind of trend gradually.In radar automatic checkout system, need to provide a detection threshold, and utilize decision rule to make the judgement whether target exists according to threshold value.The simplest method is rule of thumb provide a fixing detection threshold, but this method to be poor for the adaptive faculty of the clutter background changing.Such as, in non-stationary clutter, if adopt fixed threshold to detect, clutter average power level increases several decibels and will make false-alarm probability sharply increase, to such an extent as to display picture is saturated or data processing equipment overload, even if at this moment signal to noise ratio (S/N ratio) is very large, can not make correct judgement.Therefore the CFAR problem in the automatic detection of radar is each radar system research and the unavoidable major issue of designer.CFAR (CFAR) treatment technology is to detecting strategy, detection threshold to be provided and to make clutter on the false-alarm probability of system, affect minimized signal processing technology with disturbing in radar automatic checkout system.Its uses adaptive threshold to replace fixed threshold, and adaptive threshold can be adjusted along with the size of ground unrest, clutter and the interference of tested measuring point.If ground unrest, clutter and interference are large, adaptive threshold will improve thereupon; If ground unrest, clutter and interference are little, adaptive threshold will lower thereupon, thereby guarantees that false-alarm probability is constant.The CFAR of radar signal (CFAR) is processed and in Radar Signal Processing, has been occupied indispensable critical role, is therefore widely used in recent years.

At present, about the Project Realization of CFAR, be mainly to have developed in the hardware platform of DSP, its weak point mainly contains following 2 points:

First, conventionally adopt serial processing, according to order from front to back, each range unit in the corresponding pulses having arranged in internal memory is detected, carrying out successively so in order detection can cause the processing time longer, if data volume is large, extremely can affect the real-time of whole system, can cause system to break down, not bring into play the high speed processing ability of dsp chip, waste the internal resource of chip.

Secondly, if common serial processing occurs that in implementation procedure result is incorrect, be difficult for finding out its wrong place, need single-step debug step by step to determine the place of makeing mistakes, such program is difficult for revising debugging, and portability is not high, and then can reduce the reliability of system.

Summary of the invention

For the deficiency of above-mentioned prior art, a kind of optimization method based on CFAR target detection has been proposed, to reduce power consumption, improve the execution efficiency of software.

For achieving the above object, the present invention is achieved by the following technical solutions.

An optimization method based on CFAR target detection, is characterized in that, comprises the following steps:

Step 1, radar receiving target echo data, carries out pulse compression to target echo data, then carries out moving-target detection, obtains the data W of N pulse after detecting through moving-target, W=[W ₁, W ₂..., W _i..., W _n];

Step 2, the data W of i pulse after detecting through moving-target _iin there is M range unit, M range unit sequentially arranged, and selects successively from left to right the first detecting unit X _ijand select successively from right to left the second detecting unit X simultaneously _ik;

Step 3, as described the first detecting unit X _ijthe right while there is q reference unit and p protected location, and described the second detecting unit X _ikthe left side while there is q reference unit and p protected location, j=1,2 ..., p+q, k=M, M-1 ..., M-(p+q-1), and j+k=M+1;

The q of a first detecting unit reference unit is sequentially rearranged to the first reference unit vector A, q the corresponding data of reference unit in the first reference unit vector A are sequentially rearranged to the first data vector C, the element addition calculation in the first data vector C is gone out to the first data accumulation and S _ij; The q of a second detecting unit reference unit is sequentially rearranged to the second reference unit vector B, q the corresponding data of reference unit in the second reference unit vector B are sequentially rearranged to the second data vector D, the element addition calculation in the second data vector D is gone out to the second data accumulation and S _ik; And according to false-alarm probability P _faset thresholding factor T;

By the first data accumulation and S _ijdivided by the right reference unit number q, obtain the first reference unit data average, thresholding factor T and the first reference unit data average are multiplied each other and obtain the first detection threshold R _j; By the second data accumulation and S _ikdivided by left side reference unit number q, obtain the second reference unit data average, thresholding factor T and the second reference unit data average are multiplied each other and obtain the second detection threshold R _k;

By the first detection threshold R _jwith the first detecting unit X _ijcorresponding data Y _ijrelatively, if Y _ij≤ R _jjudge the first detecting unit X _ijthere is not target, if Y _ij>R _jjudge the first detecting unit X _ijthere is target; By the second detection threshold R _kwith the second detecting unit X _ikcorresponding data Z _ikrelatively, if Z _ik≤ R _kjudge the second detecting unit X _ikthere is not target, if Z _ik>R _kjudge the second detecting unit X _ikthere is target;

Step 4, as described the first detecting unit X _ijthe right and the left side while simultaneously there is q reference unit and p protected location, and described the second detecting unit X _ikthe right and the left side while simultaneously there is q reference unit and p protected location; Wherein, j+k=M+1;

Q the reference unit on first detecting unit the right sequentially rearranged to the right the first reference unit vector A ₁, the q on a first detecting unit left side reference unit is sequentially rearranged to the left side the first reference unit vector A ₂, by the right the first reference unit vector A ₁the corresponding data of a middle q reference unit sequentially rearrange the right the first data vector C ₁, and by the left side the first reference unit vector A ₂the corresponding data of a middle q reference unit sequentially rearrange the left side the first data vector C ₂, by the right the first data vector C ₁with the left side the first data vector C ₂in element addition calculation go out the first data accumulation and S _ij;

Q the reference unit on second detecting unit the right sequentially rearranged to the right the second reference unit vector B ₁, the q on a second detecting unit left side reference unit is sequentially rearranged to the left side the second reference unit vector B ₂, by the right the second reference unit vector B ₁the corresponding data of a middle q reference unit sequentially rearrange the right the second data vector D ₁, and by the left side the second reference unit vector B ₂the corresponding data of a middle q reference unit sequentially rearrange the left side the second data vector D ₂, by the right the second data vector D ₁with the left side the second data vector D ₂in element addition calculation go out the second data accumulation and S _ik;

By the first data accumulation and S _ijdivided by the right and left reference unit number 2q, obtain the first reference unit data average, thresholding factor T and the first reference unit data average are multiplied each other and obtain the first detection threshold R _j; By the second data accumulation and S _ikdivided by the right and left reference unit number 2q, obtain the second reference unit data average, thresholding factor T and the second reference unit data average are multiplied each other and obtain the second detection threshold R _k;

By the first detection threshold R _jwith the first detecting unit X _ijcorresponding data Y _ijrelatively, if Y _ij≤ R _jjudge the first detecting unit X _ijthere is not target, if Y _ij>R _jjudge the first detecting unit X _ijthere is target; By the second detection threshold R _kwith the second detecting unit X _ikcorresponding data Z _ikrelatively, if Z _ik≤ R _kjudge the second detecting unit X _ikthere is not target, if Z _ik>R _kjudge the second detecting unit X _ikthere is target;

Step 5, obtains the testing result of each range unit in N pulse according to step 2,3,4.

The feature of technique scheme and further improvement are:

(1) step 3 comprises following sub-step:

As described the first detecting unit X _ijthe right while there is q reference unit and p protected location, and described the second detecting unit X _ikthe left side while there is q reference unit and p protected location:

3a) utilize the first pointer to select the first detecting unit X _ij, j=1,2 ..., p+q; Utilize the second pointer to select the second detecting unit X _ik, k=M, M-1 ..., M-(p+q-1);

Set the first detecting unit X _ijthe first reference unit vector be A=[X _{i, j+ (p+1)}, X _{i, j+ (p+2)}..., X _{i, j+ (p+q)}], the second detecting unit X _ikthe second reference unit vector be B=[X _{i, k-(p+q)}, X _{i, k-(p+q-1)}..., X _{i, k-(p+1)}];

(3b) set the first detecting unit X _ijcorresponding the first data vector C=[Y of the first reference unit vector A _{i, j+ (p+1)}, Y _{i, j+ (p+2)}..., Y _{i, j+ (p+q)}],

Set the second detecting unit X _ikcorresponding the second data vector D=[Z of the second reference unit vector B _{i, k-(p+q)}, Z _{i, k-(p+q-1)}..., Z _{i, k-(p+1)}];

Element in the first data vector C is added up, obtain the first data accumulation and S _ij=Y _{i, j+ (p+1)}+ Y _{i, j+ (p+2)}+ ...+Y _{i, j+ (p+q)}, the element in the second data vector D is added up, obtain the second data accumulation and S _ik=Z _{i, k-(p+q)}+ Z _{i, k-(p+q-1)}+ ...+Z _{i, k-(p+1)};

(3c) set thresholding factor T and false-alarm probability P _farelational expression and false-alarm probability P _faequal invariable false alerting Γ;

Pass through ask for thresholding factor T, wherein, exp represents to take the index that e is the end;

(3d) by following formula, obtain the first detection threshold R _j:

$R_j=\frac{S_{\mathrm{ij}}}{q}\×T;$

Wherein, S _ijrepresent the first data accumulation and, q represents reference unit number, T represents the thresholding factor;

By the first detection threshold R _jwith the first detecting unit X _ijcorresponding data Y _ijrelatively, if Y _ij≤ R _jjudge the first detecting unit X _ijthere is not target, if Y _ij>R _jjudge the first detecting unit X _ijthere is target;

(3e) by following formula, obtain the second detection threshold R _k:

$R_k=\frac{S_{\mathrm{ik}}}{q}\×T;$

Wherein, S _ikrepresent the second data accumulation and, q represents reference unit number, T represents the thresholding factor;

By the second detection threshold R _kwith the second detecting unit X _ikcorresponding data Z _ikrelatively, if Z _ik≤ R _kjudge the second detecting unit X _ikthere is not target, if Z _ik>R _kjudge the second detecting unit X _ikthere is target;

(3f) make j travel through from 1 to p+q, repeating step (3a), (3b), (3c), (3d); And with seasonal k, from M to M-(p+q-1), travel through, repeating step (3a), (3b), (3c), (3e), complete detection.

(2) step 4 comprises following sub-step:

As described the first detecting unit X _ijthe right and the left side while all there is q reference unit and p protected location, and described the second detecting unit X _ikthe right and the left side while all there is q reference unit and p protected location:

(4a) utilize the first pointer to select the first detecting unit X _ij;

Utilize the second pointer to select the second detecting unit X _ik;

Set the first detecting unit X _ijthe right the first reference unit vector is A ₁=[X _{i, j+ (p+1)}, X _{i, j+ (p+2)}..., X _{i, j+ (p+q)}], the first detecting unit X _ijthe left side the first reference unit vector is A ₂=[X _{i, j-(p+q)}, X _{i, j-(p+q-1)}..., X _{i, j-(p+1)}]; Set the second detecting unit X _ikthe right the second reference unit vector is B ₁=[X _{i, k+ (p+1)}, X _{i, k+ (p+2)}..., X _{i, k+ (p+q)}], the second detecting unit X _ikthe left side the second reference unit vector is B ₂=[X _{i, k-(p+q)}, X _{i, k-(p+q-1)}..., X _{i, k-(p+1)}];

(4b) set the first detecting unit X _ijthe right the first reference unit vector A ₁corresponding the right the first data vector:

C ₁＝[Y _i,j+(p+1),Y _i,j+(p+2),...,Y _i,j+(p+q)]；

Set the first detecting unit X _ijthe left side the first reference unit vector A ₂the corresponding left side the first data vector:

C ₂＝[Y _i,j-(p+q),Y _i,j-(p+q-1),...,Y _i,j-(p+1)]；

Set the second detecting unit X _ikthe right the second reference unit vector B ₁corresponding the right the second data vector:

D ₁＝[Z _i,k+(p+1),Z _i,k+(p+2),...,Z _i,k+(p+q)]；

Set the second detecting unit X _ikthe left side the second reference unit vector B ₂the corresponding left side the second data vector:

D ₂＝[Z _i,k-(p+q),Z _i,k-(p+q-1),...,Z _i,k-(p+1)]；

By the right the first data vector C ₁with the left side the first data vector C ₂in element add up, obtain the first data accumulation and S _ij=Y _{i, j-(p+q)}+ Y _{i, j-(p+q-1)}+ ...+Y _{i, j-(p+1)}+ Y _{i, j+ (p+1)}+ Y _{i, j+ (p+2)}+ ...+Y _{i, j+ (p+q)}; By the right the second data vector D ₁with the left side the second data vector D ₂in element add up, obtain the second data accumulation and S _ik=Z _{i, k-(p+q)}+ Z _{i, k-(p+q-1)}+ ...+Z _{i, k-(p+1)}+ Z _{i, k+ (p+1)}+ Z _{i, k+ (p+2)}+ ...+Z _{i, k+ (p+q)};

(4c) by following formula, obtain the first detection threshold R _j:

$R_j=\frac{S_{\mathrm{ij}}}{q}\×T;$

Wherein, S _ijrepresent the first data accumulation and, 2q represents the total number of the right and left reference unit, T represents the thresholding factor;

By the first detection threshold R _jwith the first detecting unit X _ijcorresponding data Y _ijcompare, if Y _ij≤ R _jjudge the first detecting unit X _ijthere is not target, if Y _ij>R _jjudge the first detecting unit X _ijthere is target;

(4d) by following formula, obtain the second detection threshold R _k:

$R_k=\frac{S_{\mathrm{ik}}}{2q}\×T;$

Wherein, S _ikrepresent the second data accumulation and, 2q represents the total number of the right and left reference unit, T represents the thresholding factor;

By the second detection threshold R _kwith the second detecting unit X _ikcorresponding data Z _ikcompare, if Z _ik≤ R _kjudge the second detecting unit X _ikthere is not target, if Z _ik>R _kjudge the second detecting unit X _ikthere is target;

(4e) make j from p+q+1 extremely travel through repeating step (4a), (4b), (4c); And with seasonal k from M-(p+q) extremely travel through, repeating step (4a), (4b), (4d), complete detection.

Compared with prior art, the present invention has outstanding substantive distinguishing features and significant progressive.The present invention compared with the conventional method, has the following advantages:

1) existing use DSP hardware platform completes the engineering implementation method to CFAR algorithm, the general serial processing that adopts, according to order from front to back, each range unit in corresponding pulses is detected, processing time is longer, and adopt in the present invention the method for parallel processing, utilize the integer arithmetic logic unit of DSP itself simultaneously, two operation blocks resources and stream line operation mode detect to centre from two ends the range unit in corresponding pulses simultaneously, so just can in the monocycle, complete the detection of two unit, having shortened the processing time has increased substantially work efficiency.Take that to process 6030 points be example, common serial processing method instruction cycles used is 97491, and method for parallel processing of the present invention instruction cycles used is 50916, situation the give an order cycle constant at shared DSP internal memory shortened 46575, and efficiency has improved approximately 47.77%.Due to the necessary real time execution of most DSP algorithm, it is particularly important therefore improving efficiency of algorithm;

2) existing use DSP hardware platform completes the engineering implementation method to CFAR algorithm, if there is the incorrect place that locates errors that is difficult for of result in implementation procedure, be difficult for revising debugging, the method that the present invention detects according to temporary location after the unit inspection of first both sides in algorithm implementation procedure is carried out the modularization of program, so can carry out wrong eliminating by module when occurring that result is incorrect, first location of mistake is being easy to modification debugging, is improving the reliability of system.

Accompanying drawing explanation

Below in conjunction with the drawings and specific embodiments, the present invention will be further described.

Fig. 1 is realization flow figure of the present invention;

Fig. 2 be in the present invention two edge element CFAR detection in a pulse realize schematic diagram;

Fig. 3 be in the present invention the temporary location CFAR detection in a pulse realize schematic diagram;

Fig. 4 adopts Matlab emulation data to be carried out to the result figure of CFAR processing in the present invention, wherein X-axis represents Doppler's channel number, and Y-axis represents that, apart from channel number, Z axis represents amplitude;

Fig. 5 adopts DSP emulation data to be carried out to the result figure of CFAR processing in the present invention, wherein X-axis represents Doppler's channel number, and Y-axis represents that, apart from channel number, Z axis represents amplitude.

Embodiment

With reference to Fig. 1, a kind of optimization method based on CFAR target detection of the present invention is described, the CFAR detection for realizing radar signal at monolithic DSP, comprises the following steps.

Step 1, radar receiving target echo data, carries out pulse compression to target echo data, then carries out moving-target detection, obtains the data W of N pulse after detecting through moving-target, W=[W ₁, W ₂..., W _i..., W _n].

In the present invention, it is that the data that paired pulses compresses after processing are carried out filtering processing that moving-target detects, and its detailed theory and process come from < < Radar Signal Processing > >.

Step 2, the data W of i pulse after detecting through moving-target _iin there is M range unit, M range unit sequentially arranged, and selects successively from left to right the first detecting unit X _ijand select successively from right to left the second detecting unit X simultaneously _ik.

Step 3, as described the first detecting unit X _ijthe right while there is q reference unit and p protected location, and described the second detecting unit X _ikthe left side while there is q reference unit and p protected location, j=1,2 ..., p+q, k=M, M-1 ..., M-(p+q-1), and j+k=M+1;

The q of a first detecting unit reference unit is sequentially rearranged to the first reference unit vector A, q the corresponding data of reference unit in the first reference unit vector A are sequentially rearranged to the first data vector C, the element addition calculation in the first data vector C is gone out to the first data accumulation and S _ij; The q of a second detecting unit reference unit is sequentially rearranged to the second reference unit vector B, q the corresponding data of reference unit in the second reference unit vector B are sequentially rearranged to the second data vector D, the element addition calculation in the second data vector D is gone out to the second data accumulation and S _ik; And according to false-alarm probability P _faset thresholding factor T;

By the first data accumulation and S _ijdivided by the right reference unit number q, obtain the first reference unit data average, thresholding factor T and the first reference unit data average are multiplied each other and obtain the first detection threshold R _j; By the second data accumulation and S _ikdivided by left side reference unit number q, obtain the second reference unit data average, thresholding factor T and the second reference unit data average are multiplied each other and obtain the second detection threshold R _k;

By the first detection threshold R _jwith the first detecting unit X _ijcorresponding data Y _ijrelatively, if Y _ij≤ R _jjudge the first detecting unit X _ijthere is not target, if Y _ij>R _jjudge the first detecting unit X _ijthere is target; By the second detection threshold R _kwith the second detecting unit X _ikcorresponding data Z _ikrelatively, if Z _ik≤ R _kjudge the second detecting unit X _ikthere is not target, if Z _ik>R _kjudge the second detecting unit X _ikthere is target.

As described the first detecting unit X _ijthe right while there is q reference unit and p protected location, and described the second detecting unit X _ikthe left side while there is q reference unit and p protected location:

3a) utilize the first pointer to select the first detecting unit X _ij, j=1,2 ..., p+q; Utilize the second pointer to select the second detecting unit X _ik, k=M, M-1 ..., M-(p+q-1); As shown in Figure 2.

From pointer above, select to find out, the first pointer starts chosen distance unit from pulse left end, and the second pointer starts chosen distance unit from pulse right-hand member.

In the present invention program, the first pointer of use is J pointer, and the second pointer is K pointer; Or the first pointer is K pointer, the second pointer is J pointer, does not limit the selection mode of pointer in the present invention, can select according to the practical application scene in DSP.

Set the first detecting unit X _ijthe first reference unit vector be A=[X _{i, j+ (p+1)}, X _{i, j+ (p+2)}..., X _{i, j+ (p+q)}], the second detecting unit X _ikthe second reference unit vector be B=[X _{i, k-(p+q)}, X _{i, k-(p+q}- ₁₎..., X _{i, k-(p+1)}];

(3b) set the first detecting unit X _ijcorresponding the first data vector C=[Y of the first reference unit vector A _{i, j+ (p+1)}, Y _{i, j+ (p+2)}..., Y _{i, j+ (p+q)}],

Set the second detecting unit X _ikcorresponding the second data vector D=[Z of the second reference unit vector B _{i, k-(p+q)}, Z _{i, k-(p+q-1)}..., Z _{i, k-(p+1)}];

Element in the first data vector C is added up, obtain the first data accumulation and S _ij=Y _{i, j+ (p+1)}+ Y _{i, j+ (p+2)}+ ...+Y _{i, j+ (p+q)}, the element in the second data vector D is added up, obtain the second data accumulation and S _ik=Z _{i, k-(p+q)}+ Z _{i, k-(p+q-1)}+ ...+Z _{i, k-(p+1)};

(3c) set thresholding factor T and false-alarm probability P _farelational expression and false-alarm probability P _faequal invariable false alerting Γ;

Pass through ask for thresholding factor T, wherein, exp represents to take the index that e is the end.

In the present invention, invariable false alerting Γ is according to the constant of actual items requirements set.

(3d) by following formula, obtain the first detection threshold R _j:

$R_j=\frac{S_{\mathrm{ij}}}{q}\&times;T;$

Wherein, S _ijrepresent the first data accumulation and, q represents reference unit number, T represents the thresholding factor;

By the first detection threshold R _jwith the first detecting unit X _ijcorresponding data Y _ijrelatively, if Y _ij≤ R _jjudge the first detecting unit X _ijthere is not target, if Y _ij>R _jjudge the first detecting unit X _ijthere is target;

(3e) by following formula, obtain the second detection threshold R _k:

$R_k=\frac{S_{\mathrm{ik}}}{q}\&times;T;$

Wherein, S _ikrepresent the second data accumulation and, q represents reference unit number, T represents the thresholding factor;

By the second detection threshold R _kwith the second detecting unit X _ikcorresponding data Z _ikrelatively, if Z _ik≤ R _kjudge the second detecting unit X _ikthere is not target, if Z _ik>R _kjudge the second detecting unit X _ikthere is target;

(3f) make j travel through from 1 to p+q, repeating step (3a), (3b), (3c), (3d); And with seasonal k, from M to M-(p+q-1), travel through, repeating step (3a), (3b), (3c), (3e), complete detection.

The CFAR detection essence that step 3 completes is: from pulse left end, start to detect to the left side of p+q unit, and start to detect to the right of the individual unit of M-(p+q-1) from right-hand member, namely realized both sides testing process.

Pointer of available technology adopting takes out a range unit and carries out CFAR detection as detecting unit from i pulse, utilizes the method for this serial processing need to do 2 (p+q) inferior CFAR detection; The present invention is a parallel operation, adopts two pointers to take out two range units from the two ends of i pulse simultaneously and carries out CFAR detection as detecting unit, only need be p+q time CFAR detection.

Sum up step 3, when doing both sides unit inspection, for the detecting unit of left end in a pulse, there is reference unit in its right and the left side does not exist reference unit; For the detecting unit of right-hand member in a pulse, there is reference unit in its left side and the right does not exist reference unit, so the core point of step 3 reference unit that has been refinements has provided two these modules of edge element CFAR detection.

Step 4, as described the first detecting unit X _ijthe right and the left side while simultaneously there is q reference unit and p protected location, and described the second detecting unit X _ikthe right and the left side while simultaneously there is q reference unit and p protected location; Wherein, j+k=M+1;

Q the reference unit on first detecting unit the right sequentially rearranged to the right the first reference unit vector A ₁, the q on a first detecting unit left side reference unit is sequentially rearranged to the left side the first reference unit vector A ₂, by the right the first reference unit vector A ₁the corresponding data of a middle q reference unit sequentially rearrange the right the first data vector C ₁, and by the left side the first reference unit vector A ₂the corresponding data of a middle q reference unit sequentially rearrange the left side the first data vector C ₂, by the right the first data vector C ₁with the left side the first data vector C ₂in element addition calculation go out the first data accumulation and S _ij;

Q the reference unit on second detecting unit the right sequentially rearranged to the right the second reference unit vector B ₁, the q on a second detecting unit left side reference unit is sequentially rearranged to the left side the second reference unit vector B ₂, by the right the second reference unit vector B ₁the corresponding data of a middle q reference unit sequentially rearrange the right the second data vector D ₁, and by the left side the second reference unit vector B ₂the corresponding data of a middle q reference unit sequentially rearrange the left side the second data vector D ₂, by the right the second data vector D ₁with the left side the second data vector D ₂in element addition calculation go out the second data accumulation and S _ik;

By the first data accumulation and S _ijdivided by the right and left reference unit number 2q, obtain the first reference unit data average, thresholding factor T and the first reference unit data average are multiplied each other and obtain the first detection threshold R _j; By the second data accumulation and S _ikdivided by the right and left reference unit number 2q, obtain the second reference unit data average, thresholding factor T and the second reference unit data average are multiplied each other and obtain the second detection threshold R _k;

By the first detection threshold R _jwith the first detecting unit X _ijcorresponding data Y _ijrelatively, if Y _ij≤ R _jjudge the first detecting unit X _ijthere is not target, if Y _ij>R _jjudge the first detecting unit X _ijthere is target; By the second detection threshold R _kwith the second detecting unit X _ikcorresponding data Z _ikrelatively, if Z _ik≤ R _kjudge the second detecting unit X _ikthere is not target, if Z _ik>R _kjudge the second detecting unit X _ikthere is target.

As described the first detecting unit X _ijthe right and the left side while all there is q reference unit and p protected location, and described the second detecting unit X _ikthe right and the left side while all there is q reference unit and p protected location: as shown in Figure 3.

(4a) utilize the first pointer to select the first detecting unit X _ij;

Utilize the second pointer to select the second detecting unit X _ik;

Set the first detecting unit X _ijthe right the first reference unit vector is A ₁=[X _{i, j+ (p+1)}, X _{i, j+ (p+2)}..., X _{i, j+ (p+q)}], the first detecting unit X _ijthe left side the first reference unit vector is A ₂=[X _{i, j-(p+q)}, X _{i, j-(p+q-1)}..., X _{i, j-(p+1)}]; Set the second detecting unit X _ikthe right the second reference unit vector is B ₁=[X _{i, k+ (p+1)}, X _{i, k+ (p+2)}..., X _{i, k+ (p+q)}], the second detecting unit X _ikthe left side the second reference unit vector is B ₂=[X _{i, k-(p+q)}, X _{i, k-(p+q-1)}..., X _{i, k-(p+1)}];

(4b) set the first detecting unit X _ijthe right the first reference unit vector A ₁corresponding the right the first data vector:

C ₁＝[Y _i,j+(p+1),Y _i,j+(p+2),...,Y _i,j+(p+q)]；

Set the first detecting unit X _ijthe left side the first reference unit vector A ₂the corresponding left side the first data vector:

C ₂＝[Y _i,j-(p+q),Y _i,j-(p+q-1),...,Y _i,j-(p+1)]；

Set the second detecting unit X _ikthe right the second reference unit vector B ₁corresponding the right the second data vector:

D ₁＝[Z _i,k+(p+1),Z _i,k+(p+2),...,Z _i,k+(p+q)]；

Set the second detecting unit X _ikthe left side the second reference unit vector B ₂the corresponding left side the second data vector:

D ₂＝[Z _i,k-(p+q),Z _i,k-(p+q-1),...,Z _i,k-(p+1)]；

By the right the first data vector C ₁with the left side the first data vector C ₂in element add up, obtain the first data accumulation and S _ij=Y _{i, j-(p+q)}+ Y _{i, j-(p+q-1)}+ ...+Y _{i, j-(p+1)}+ Y _{i, j+ (p+1)}+ Y _{i, j+ (p+2)}+ ...+Y _{i, j+ (p+q)}; By the right the second data vector D ₁with the left side the second data vector D ₂in element add up, obtain the second data accumulation and S _ik=Z _{i, k-(p+q)}+ Z _{i, k-(p+q-1)}+ ...+Z _{i, k-(p+1)}+ Z _{i, k+ (p+1)}+ Z _{i, k+ (p+2)}+ ...+Z _{i, k+ (p+q)}.

(4c) by following formula, obtain the first detection threshold R _j:

$R_j=\frac{S_{\mathrm{ij}}}{2q}\&times;T;$

Wherein, S _ijrepresent the first data accumulation and, 2q represents the total number of the right and left reference unit, T represents the thresholding factor;

By the first detection threshold R _jwith the first detecting unit X _ijcorresponding data Y _ijcompare, if Y _ij≤ R _jjudge the first detecting unit X _ijnot target, if Y _ij>R _jjudge the first detecting unit X _ijit is target;

(4d) by following formula, obtain the second detection threshold R _k:

$R_k=\frac{S_{\mathrm{ik}}}{2q}\&times;T;$

Wherein, S _ikrepresent the second data accumulation and, 2q represents the total number of the right and left reference unit, T represents the thresholding factor;

By the second detection threshold R _kwith the second detecting unit X _ikcorresponding data Z _ikcompare, if Z _ik≤ R _kjudge the second detecting unit X _iknot target, if Z _ik>R _kjudge the second detecting unit X _ikit is target;

(4e) make j from p+q+1 extremely travel through repeating step (4a), (4b), (4c); And with seasonal k from M-(p+q) extremely travel through, repeating step (4a), (4b), (4d), complete detection.

The CFAR detection essence that step 4 completes is: from p+q unit of pulse, start to finish to pulse temporary location, and finish since the individual unit of M-(p+q-1) to pulse temporary location, namely realized the testing process of temporary location.

Pointer of available technology adopting takes out a range unit and carries out CFAR detection as detecting unit from i pulse, utilizes the method for this serial processing need to be CFAR detection M-2p-2q time; The present invention is a parallel operation, adopts two pointers from i pulse, to take out two range units simultaneously and carries out CFAR detection as detecting unit, only need do inferior CFAR detection.

Sum up step 4, when doing temporary location detection, for the detecting unit of a pulse left end, all there is reference unit in its right and left; For the detecting unit of a pulse right-hand member, also all there is reference unit in its right and left, so the core point of step 4 reference unit that has been refinements has provided this module of temporary location CFAR detection.

It should be noted that, Yi Bian step 3 of the present invention realizes the situation that detecting unit only exists reference unit and protected location, the reference unit number of each detecting unit is q, and the number of protected location is p; Accordingly; in step 4, realize the situation that detecting unit both sides exist reference unit and protected location; the reference unit number of each detecting unit is the total number 2q of the right and left reference unit, and the number of protected location is the total number 2p of the right and left protected location.Known in those skilled in the art, when selecting reference unit and protected location, first determine detecting unit, at place, detecting unit adjacency unit, select protected location, in protected location outside, select reference unit.

Step 5, obtains the testing result of each range unit in N pulse according to step 2,3,4.

Set i and add 1, repeating step 2,3,4, until finish the range unit CFAR detection in all pulses, N indicating impulse sum during i=N.

After step 3 executes, done p+q detection, after step 4 executes, done inferior detection, need to do altogether so finish a detection in pulse inferior detection.

Below in conjunction with emulation experiment, effect of the present invention is described further.

1. simulated conditions:

The distance of target setting is 10km, and speed is 900m/s;

If transmit as linear FM signal, signal pulse width is 40us, and the pulse repetition time is 120.4us, and carrier frequency is 12GHz, and it is 16 points that accumulation is counted, and signal bandwidth is 4MHz, and sample frequency is 5MHz;

If the TS201 chip of the chip WeiADI company that DSP is selected;

2. emulation content:

Emulation 1, prior art utilizes Matlab to carry out CFAR processing to the data of finishing after coherent accumulation, and CFAR detection the results are shown in Figure 4, and the black blockage identifying from figure can be found out the target information detecting;

Emulation 2, adopt the present invention to carry out CFAR processing to the data of finishing after coherent accumulation, the data result that DSP is processed is derived, and utilizes these data to draw in Matlab the inside, see Fig. 5, the black blockage identifying from figure can be found out the target information detecting; The instruction cycles that the instruction cycles that algorithm realization of the present invention is used and prior art are used can be calculated by each self-corresponding instruction cycle initial value and stop value.

3. simulation analysis:

From Fig. 4, X-axis represents Doppler's channel number, Y-axis represents apart from channel number, Z axis represents amplitude, can find out and detect four targets, the specifying information of these four targets is as follows: target 1 the 11st Doppler's passage, the 329th apart from passage on, corresponding data are 2025345.4424; Target 2 the 11st Doppler's passage, the 330th apart from passage on, corresponding data are 1833841.5051; Target 3 the 12nd Doppler's passage, the 329th apart from passage on, corresponding data are 1789387.9752; Target 4 the 12nd Doppler's passage, the 330th apart from passage on, corresponding data are 1585127.3895.Wherein, above-mentioned corresponding data are the amplitude information of target.

From Fig. 5, X-axis represents Doppler's channel number, Y-axis represents apart from channel number, Z axis represents amplitude, also can find out and detect four targets, four targets in these four targets and Fig. 4 are compared, learn that the result that adopts the present invention data to be carried out to CFAR processing is consistent with the result that adopts Matlab to carry out CFAR processing to identical data, show the correctness that the present invention designs.

It is that detecting unit carries out CFAR detection that pointer of available technology adopting takes out a range unit from i pulse, utilizes the method for this serial processing need to be CFAR detection M time.In intercepting DSP, prior art is used single needle instruction cycle result, and in prior art emulation, the initial value of instruction cycle is 797, stop value is 98288, and can calculate the instruction cycles that prior art used is 97491.

The present invention is a parallel operation, adopts two pointers to take out two range unit X from the two ends of i pulse simultaneously _ij, X _ikas detecting unit, carry out CFAR detection, only need do inferior CFAR detection.In intercepting DSP, the present invention uses the instruction cycle result of two pointers, and in emulation of the present invention, the initial value of instruction cycle is 797, stop value is 51713, and can calculate the instruction cycles that the inventive method used is 50916.

Adopting prior art to do the instruction cycles that CFAR detection used is 97491, and adopting the present invention to do the instruction cycles that CFAR detection used is 50916, can calculate the present invention efficiency has been improved approximately according to these two data increased substantially work efficiency, for the real-time working of DSP has been made important contribution.

Claims (3)

1. the optimization method based on CFAR target detection, is characterized in that, comprises the following steps:

Step 1, radar receiving target echo data, carries out pulse compression to target echo data, then carries out moving-target detection, obtains the data W of N pulse after detecting through moving-target, W=[W ₁, W ₂..., W _i..., W _n];

Step 2, the data W of i pulse after detecting through moving-target _iin there is M range unit, M range unit sequentially arranged, and selects successively from left to right the first detecting unit X _ijand select successively from right to left the second detecting unit X simultaneously _ik;

Step 3, as described the first detecting unit X _ijthe right while there is q reference unit and p protected location, and described the second detecting unit X _ikthe left side while there is q reference unit and p protected location, j=1,2 ..., p+q, k=M, M-1 ..., M-(p+q-1), and j+k=M+1;

The q of a first detecting unit reference unit is sequentially rearranged to the first reference unit vector A, q the corresponding data of reference unit in the first reference unit vector A are sequentially rearranged to the first data vector C, the element addition calculation in the first data vector C is gone out to the first data accumulation and S _ij; The q of a second detecting unit reference unit is sequentially rearranged to the second reference unit vector B, q the corresponding data of reference unit in the second reference unit vector B are sequentially rearranged to the second data vector D, the element addition calculation in the second data vector D is gone out to the second data accumulation and S _ik; And according to false-alarm probability P _faset thresholding factor T;

By the first data accumulation and S _ijdivided by the right reference unit number q, obtain the first reference unit data average, thresholding factor T and the first reference unit data average are multiplied each other and obtain the first detection threshold R _j; By the second data accumulation and S _ikdivided by left side reference unit number q, obtain the second reference unit data average, thresholding factor T and the second reference unit data average are multiplied each other and obtain the second detection threshold R _k;

By the first detection threshold R _jwith the first detecting unit X _ijcorresponding data Y _ijrelatively, if Y _ij≤ R _jjudge the first detecting unit X _ijthere is not target, if Y _ij>R _jjudge the first detecting unit X _ijthere is target; By the second detection threshold R _kwith the second detecting unit X _ikcorresponding data Z _ikrelatively, if Z _ik≤ R _kjudge the second detecting unit X _ikthere is not target, if Z _ik>R _kjudge the second detecting unit X _ikthere is target;

Step 4, as described the first detecting unit X _ijthe right and the left side while simultaneously there is q reference unit and p protected location, and described the second detecting unit X _ikthe right and the left side while simultaneously there is q reference unit and p protected location; Wherein, j+k=M+1;

Q the reference unit on first detecting unit the right sequentially rearranged to the right the first reference unit vector A ₁, the q on a first detecting unit left side reference unit is sequentially rearranged to the left side the first reference unit vector A ₂, by the right the first reference unit vector A ₁the corresponding data of a middle q reference unit sequentially rearrange the right the first data vector C ₁, and by the left side the first reference unit vector A ₂the corresponding data of a middle q reference unit sequentially rearrange the left side the first data vector C ₂, by the right the first data vector C ₁with the left side the first data vector C ₂in element addition calculation go out the first data accumulation and S _ij;

Q the reference unit on second detecting unit the right sequentially rearranged to the right the second reference unit vector B ₁, the q on a second detecting unit left side reference unit is sequentially rearranged to the left side the second reference unit vector B ₂, by the right the second reference unit vector B ₁the corresponding data of a middle q reference unit sequentially rearrange the right the second data vector D ₁, and by the left side the second reference unit vector B ₂the corresponding data of a middle q reference unit sequentially rearrange the left side the second data vector D ₂, by the right the second data vector D ₁with the left side the second data vector D ₂in element addition calculation go out the second data accumulation and S _ik;

By the first data accumulation and S _ijdivided by the right and left reference unit number 2q, obtain the first reference unit data average, thresholding factor T and the first reference unit data average are multiplied each other and obtain the first detection threshold R _j; By the second data accumulation and S _ikdivided by the right and left reference unit number 2q, obtain the second reference unit data average, thresholding factor T and the second reference unit data average are multiplied each other and obtain the second detection threshold R _k;

By the first detection threshold R _jwith the first detecting unit X _ijcorresponding data Y _ijrelatively, if Y _ij≤ R _jjudge the first detecting unit X _ijthere is not target, if Y _ij>R _jjudge the first detecting unit X _ijthere is target; By the second detection threshold R _kwith the second detecting unit X _ikcorresponding data Z _ikrelatively, if Z _ik≤ R _kjudge the second detecting unit X _ikthere is not target, if Z _ik>R _kjudge the second detecting unit X _ikthere is target;

Step 5, obtains the testing result of each range unit in N pulse according to step 2,3,4.

2. a kind of optimization method based on CFAR target detection according to claim 1, is characterized in that, step 3 comprises following sub-step:

As described the first detecting unit X _ijthe right while there is q reference unit and p protected location, and described the second detecting unit X _ikthe left side while there is q reference unit and p protected location:

3a) utilize the first pointer to select the first detecting unit X _ij, j=1,2 ..., p+q; Utilize the second pointer to select the second detecting unit X _ik, k=M, M-1 ..., M-(p+q-1);

Set the first detecting unit X _ijthe first reference unit vector be A=[X _{i, j+ (p+1)}, X _{i, j+ (p+2)}..., X _{i, j+ (p+q)}], the second detecting unit X _ikthe second reference unit vector be B=[X _{i, k-(p+q)}, X _{i, k-(p+q-1)}..., X _{i, k-(p+1)}];

(3b) set the first detecting unit X _ijcorresponding the first data vector C=[Y of the first reference unit vector A _{i, j+ (p+1)}, Y _{i, j+ (p+2)}..., Y _{i, j+ (p+q)}],

Set the second detecting unit X _ikcorresponding the second data vector D=[Z of the second reference unit vector B _{i, k-(p+q)}, Z _{i, k-(p+q-1)}..., Z _{i, k-(p+1)}];

Element in the first data vector C is added up, obtain the first data accumulation and S _ij=Y _{i, j+ (p+1)}+ Y _{i, j+ (p+2)}+ ...+Y _{i, j+ (p+q)}, the element in the second data vector D is added up, obtain the second data accumulation and S _ik=Z _{i, k-(p+q)}+ Z _{i, k-(p+q-1)}+ ...+Z _{i, k-(p+1)};

(3c) set thresholding factor T and false-alarm probability P _farelational expression and false-alarm probability P _faequal invariable false alerting Γ;

Pass through ask for thresholding factor T, wherein, exp represents to take the index that e is the end;

(3d) by following formula, obtain the first detection threshold R _j:

$R_j=\frac{S_{\mathrm{ij}}}{q}\&times;T;$

Wherein, S _ijrepresent the first data accumulation and, q represents reference unit number, T represents the thresholding factor;

By the first detection threshold R _jwith the first detecting unit X _ijcorresponding data Y _ijrelatively, if Y _ij≤ R _jjudge the first detecting unit X _ijthere is not target, if Y _ij>R _jjudge the first detecting unit X _ijthere is target;

(3e) by following formula, obtain the second detection threshold R _k:

$R_k=\frac{S_{\mathrm{ik}}}{q}\&times;T;$

Wherein, S _ikrepresent the second data accumulation and, q represents reference unit number, T represents the thresholding factor;

By the second detection threshold R _kwith the second detecting unit X _ikcorresponding data Z _ikrelatively, if Z _ik≤ R _kjudge the second detecting unit X _ikthere is not target, if Z _ik>R _kjudge the second detecting unit X _ikthere is target;

(3f) make j travel through from 1 to p+q, repeating step (3a), (3b), (3c), (3d); And with seasonal k, from M to M-(p+q-1), travel through, repeating step (3a), (3b), (3c), (3e), complete detection.

3. a kind of optimization method based on CFAR target detection according to claim 1, is characterized in that, step 4 comprises following sub-step:

As described the first detecting unit X _ijthe right and the left side while all there is q reference unit and p protected location, and described the second detecting unit X _ikthe right and the left side while all there is q reference unit and p protected location:

(4a) utilize the first pointer to select the first detecting unit X _ij;

Utilize the second pointer to select the second detecting unit X _ik;

Set the first detecting unit X _ijthe right the first reference unit vector is A ₁=[X _{i, j+ (p+1)}, X _{i, j+ (p+2)}..., X _{i, j+ (p+q)}], the first detecting unit X _ijthe left side the first reference unit vector is A ₂=[X _{i, j-(p+q)}, X _{i, j-(p+q-1)}..., X _{i, j-(p+1)}]; Set the second detecting unit X _ikthe right the second reference unit vector is B ₁=[X _{i, k+ (p+1)}, X _{i, k+ (p+2)}..., X _{i, k+ (p+q)}], the second detecting unit X _ikthe left side the second reference unit vector is B ₂=[X _{i, k-(p+q)}, X _{i, k-(p+q-1)}..., X _{i, k-(p+1)}];

(4b) set the first detecting unit X _ijthe right the first reference unit vector A ₁corresponding the right the first data vector:

C ₁＝[Y _i,j+(p+1),Y _i,j+(p+2),...,Y _i,j+(p+q)]；

Set the first detecting unit X _ijthe left side the first reference unit vector A ₂the corresponding left side the first data vector:

C ₂＝[Y _i,j-(p+q),Y _i,j-(p+q-1),...,Y _i,j-(p+1)]；

Set the second detecting unit X _ikthe right the second reference unit vector B ₁corresponding the right the second data vector:

D ₁＝[Z _i,k+(p+1),Z _i,k+(p+2),...,Z _i,k+(p+q)]；

Set the second detecting unit X _ikthe left side the second reference unit vector B ₂the corresponding left side the second data vector:

D ₂＝[Z _i,k-(p+q),Z _i,k-(p+q-1),...,Z _i,k-(p+1)]；

By the right the first data vector C ₁with the left side the first data vector C ₂in element add up, obtain the first data accumulation and S _ij=Y _{i, j-(p+q)}+ Y _{i, j-(p+q-1)}+ ...+Y _{i, j-(p+1)}+ Y _{i, j+ (p+1)}+ Y _{i, j+ (p+2)}+ ...+Y _{i, j+ (p+q)}; By the right the second data vector D ₁with the left side the second data vector D ₂in element add up, obtain the second data accumulation and S _ik=Z _{i, k-(p+q)}+ Z _{i, k-(p+q-1)}+ ...+Z _{i, k-(p+1)}+ Z _{i, k+ (p+1)}+ Z _{i, k+ (p+2)}+ ...+Z _{i, k+ (p+q)};

(4c) by following formula, obtain the first detection threshold R _j:

$R_j=\frac{S_{\mathrm{ij}}}{2q}\&times;T;$

Wherein, S _ijrepresent the first data accumulation and, 2q represents the total number of the right and left reference unit, T represents the thresholding factor;

By the first detection threshold R _jwith the first detecting unit X _ijcorresponding data Y _ijcompare, if Y _ij≤ R _jjudge the first detecting unit X _ijthere is not target, if Y _ij>R _jjudge the first detecting unit X _ijthere is target;

(4d) by following formula, obtain the second detection threshold R _k:

$R_k=\frac{S_{\mathrm{ik}}}{2q}\&times;T;$

Wherein, S _ikrepresent the second data accumulation and, 2q represents the total number of the right and left reference unit, T represents the thresholding factor;

By the second detection threshold R _kwith the second detecting unit X _ikcorresponding data Z _ikcompare, if Z _ik≤ R _kjudge the second detecting unit X _ikthere is not target, if Z _ik>R _kjudge the second detecting unit X _ikthere is target;

(4e) make j from p+q+1 extremely travel through repeating step (4a), (4b), (4c); And with seasonal k from M-(p+q) extremely travel through, repeating step (4a), (4b), (4d), complete detection.