CN104156734A - Fully-autonomous on-line study method based on random fern classifier - Google Patents

Abstract

The invention discloses a fully-autonomous on-line study method based on a random fern classifier. With the method, on-line study targeted for a classifier of the target class can be carried out by selecting a target in a video frame for only once. The method comprises the following steps: carrying out affine transformation on the selected target to obtain an initial positive sample set, and extracting a small number of negative sample sets in a non- target area of the video to train an initial random fern classifier; then, carrying out target detection in the video frame by using the classifier; in the detection process, using a nearest neighbor classifier to collect on-line study new samples and automatically judging the class of the samples; and finally, applying the new samples to on-line training of the random fern classifier, updating random fern posterior probability, improving the precision of target detection of the classifier gradually, and realizing fully-autonomous on-line study of a target detection system.

Description

A kind of complete autonomous on-line study method based on random fern sorter

Technical field

The invention belongs to area of pattern recognition, be specifically related to a kind of complete autonomous on-line study method based on random fern sorter.

Background technology

On-line study belongs to the research category of incremental learning, in these class methods, sorter is only learned once each sample, rather than the study repeating, in on-line learning algorithm operational process, do not need so a large amount of storage spaces to store training sample, sample of the every acquisition of sorter, it is carried out to on-line study, by on-line study, make sorter in use still can, according to new samples self and improvement, further improve classifying quality.

Early stage on-line learning algorithm has Winnow algorithm, unified linear prediction algorithm etc., calendar year 2001 scholar Oza carries out combination by these algorithms and boosting algorithm, (this algorithm draws from " Online bagging and boosting " N.Oza and S.Russell to have proposed online boosting algorithm, In Proc.Artificial Intelligence and Statistics, 105-112, 2001), in the method for Oza, the corresponding Weak Classifier of each feature, and strong classifier is the weighted sum of the Weak Classifier of some, wherein Weak Classifier is all select from Weak Classifier set.During on-line study, each Weak Classifier in the renewal Weak Classifier set one by one of each training sample, the weight that comprises classification thresholds and this sorter of adjusting positive negative sample, the Weak Classifier weight making is more and more higher, and poor Weak Classifier weight is more and more lower, thereby each sample of on-line study just can be picked out a highest Weak Classifier of current weight and adds in strong classifier and make finally to train sorter out to have stronger classification capacity.

But each Weak Classifier will carry out on-line study to new samples in the Weak Classifier set of online boosting algorithm, when Weak Classifier number is more, on-line study speed will inevitably be slack-off.Grabner improves online boosting algorithm, make it also as Adaboost algorithm, can carry out feature selecting, and this feature selecting and more new capital of sorter is carried out online, (this algorithm draws from " On-line boosting and vision " H.Grabner and H.Bischof to be called online Adaboost, In Proc.CVPR, (1): 260-267,2006).But online Adaboost replaces general Weak Classifier to synthesize strong classifier with feature selecting operator, and the Weak Classifier number that feature selecting operator number and feature selecting operator are corresponding is all fixed, corresponding on-line study sorter structure is more rigid.When finding that its classification capacity cannot meet requiring of detection performance, even if going down, lasting on-line study also cannot improve accuracy of detection.

Ozuysal does not re-use Weak Classifier composition strong classifier and carries out sample classification, but randomly draw a plurality of features from sample characteristics set, form a random fern, by random fern statistics training sample posterior probability, distribute, by the posterior probability distribution of a plurality of random ferns, carry out sample classification again, (this algorithm draws from " Fast keypoint recognition using random ferns " In Pattern Analysis Machine Intelligence to be random fern classifier algorithm, IEEE Transaction on 32 (3), 448-461,2010).

Summary of the invention

The technical problem to be solved in the present invention is: a kind of complete autonomous on-line study method based on random fern sorter is provided, for the autonomous learning of sorter to improve classification performance.

The present invention solves the problems of the technologies described above taked technical scheme to be: a kind of complete autonomous on-line study method based on random fern sorter, is characterized in that: it comprises the steps:

1) prepare the sample set of initial training sorter:

For frame of video to be detected, at frame of video center, select a Target Photo, this Target Photo is carried out to picture that affined transformation obtains as positive sample; Using do not contain target background image region as negative sample; The so random positive sample that obtains some and negative sample are as the sample set of initial training sorter; Positive and negative samples is the image block that size is identical;

2) random fern sorter initial training:

Use the sample set of ready initial training sorter to carry out initial training to random fern sorter, add up the posterior probability of positive negative sample on each random fern and distribute;

3) the good random fern sorter of initial training is traveled through to frame of video to be detected as current goal detecting device and carry out target detection, obtain object module, and calculate the degree of confidence of each object module;

4) build positive negative sample template set:

Using following three kinds of samples as positive sample template, add positive sample template set M to ⁺, all the other add negative sample template set M to ^-:

A, step 1) in the positive sample that obtains;

B, to step 3) in the degree of confidence that obtains surpass the object module of degree of confidence preset value, adopt optical flow method to follow the tracks of and obtain tracking module its place frame of video, if tracking module and this object module have overlapping region, and coincidence factor surpasses default coincidence factor, think that this tracking module is real goal, as positive sample template, add M to ⁺in;

C, to step 3) in the degree of confidence that obtains surpass the object module of degree of confidence preset value, adopt optical flow method to follow the tracks of and obtain tracking module its place frame of video, if tracking module and this object module have overlapping region, and coincidence factor is over default coincidence factor, by conservative similarity S ^cjudge that can this tracking module add positive sample template set:

$S^c=\frac{S_{50\%}^{+}}{S_{50\%}^{+}+S^{-}}$

Wherein:

If S ^cbe greater than default conservative similarity threshold, this tracking module adds M as positive sample template ⁺, for the similarity of the first half template of sample to be sorted and current positive sample template set, S ⁺, S ^-be respectively the similarity of sample to be sorted and positive and negative samples template set, be the similarity of two picture frames, p ⁺, p ^-be respectively positive sample and negative sample, p is sample to be sorted, and in this step, sample to be sorted is tracking module;

Often add a positive sample template, get in same frame of video its around the image block of four formed objects determine whether negative sample, if add negative sample template set M as negative sample template ^-;

5) use nearest neighbor classifier, obtain the positive negative sample of on-line study:

Arranging of nearest neighbor classifier is as follows: for each sample p to be sorted, calculate respectively itself and the similarity S of positive negative sample template set ⁺(p, M ⁺) and S ^-(p, M ^-):

$S^{+}(p,M^{+})={\max }_{p_i^{+}\∈M^{+}}S(p,p_i^{+})$

$S^{-}(p,M^{-})={\max }_{p_i^{-}\∈M^{-}}S(p,p_i^{-})$

Can obtain similarity S accordingly ^r:

$S^r=\frac{S^{+}}{S^{+}+S^{-}}$

If similarity S ^rbe greater than threshold value θ _nN, judge that this sample to be sorted is real goal, as the positive sample of on-line study; Otherwise be false-alarm, as the negative sample of on-line study;

In this step, sample to be sorted is step 3) object module and the step 4 that obtain) the positive negative sample template set that obtains;

(6) the online training of random fern sorter:

The positive negative sample of on-line study use step 5) obtaining, carries out on-line study to random fern sorter, improves gradually its nicety of grading;

The random fern sorter of on-line study is carried out to target detection as the detection system of sustainable renewal.

Press such scheme, described step 2) concrete grammar as follows:

2.1) construct random fern:

To on the single sample in the sample set of initial training sorter, get at random s to unique point as one group of random fern, it is identical that each sample is got the position of unique point, every pair of unique point is carried out the comparison of pixel value, in every pair of unique point, greatly to get eigenwert be 1 to previous unique point pixel value, otherwise get eigenwert, be 0, the s that s obtains more afterwards to a unique point eigenwert forms the binary number of a s position according to random order, be the random fern numerical value that this organizes random fern, the sequence consensus of eigenwert in the random fern of each sample;

2.2) calculate the posterior probability of random fern numerical value in positive negative sample class:

In random fern, some obtains for positive sample, and other obtains for negative sample; The value kind of random fern numerical value has 2 ^sindividual;

Add up the positive number of samples of the value of every kind of random fern numerical value, thereby obtain random fern numerical value at positive sample class C ₁on posterior probability distribution P (F _l| C ₁); In like manner obtain random fern numerical value at negative sample class C ₀on posterior probability distribution P (F _l| C ₀); Combine all random ferns the sample set of initial training sorter is classified, be random fern sorter;

Described step 3) adopt above-mentioned random fern sorter in every frame video image, to carry out target detection:

Travel through every frame video image to be detected, in every frame video image, extract the image block of formed objects as sample to be tested, size and the step 1 of sample to be tested) in positive size equate, calculate the random fern numerical value of each sample to be tested, thereby obtain corresponding posterior probability, finally by random its classification of fern classifier calculated;

The image block that is positive sample for classification, is detected as target.

Press such scheme, described step 4) often add a positive sample template, when getting in same frame of video it image block of four formed objects determining whether negative sample, introduce Gaussian Background modeling around, if foreground pixel is less than foreground pixel threshold value in image block, judge that it is negative sample.

Press such scheme, described step 4) also comprise that template set subdues mechanism: the similarity of sample to be sorted and positive and negative template set equals in sample to be sorted and positive and negative template set the maximal value of similarity between single positive negative sample template; Each positive negative sample template of real-time statistics obtains this peaked number of times, if this peaked number of times that certain positive negative sample template obtains is less than maximal value number of times preset value, removes corresponding positive sample template or negative sample template.

Press such scheme, described step 6) on-line study of random fern sorter distributes and realizes by upgrading posterior probability.

Press such scheme, described step 6) concrete grammar is as follows:

6.1) using step 5) the positive negative sample that obtains is as on-line study sample; If an on-line study sample is (f _new, c _k), f wherein _newfor the binary number of random fern s position, c _kfor sample class, calculate the random fern numerical value of this on-line study sample;

6.2) to step 2.1) classification is c in sample set _ktotal sample number add 1, classification is c _kthe sample number identical with the random fern numerical value of this on-line study sample add 1; The sample number of other random fern numerical value is constant;

6.3), according to the sample number after upgrading, recalculate the posterior probability of random fern numerical value in this sample class and distribute;

6.4) often increase an on-line study sample newly, just repeat 6.1) to 6.3) posterior probability is distributed and upgraded once.

Beneficial effect of the present invention is:

1, only need select a target can carry out the sorter on-line study for this target class at frame of video center, that is: first to the target of frame choosing, adopt affined transformation to obtain initial positive sample set, in the nontarget area of video, extract a small amount of negative sample training and practice initial random fern sorter; Secondly, use this sorter in frame of video, to carry out target detection; In the process detecting, adopt nearest neighbor classifier to collect on-line study new samples, and automatic decision sample class; Finally, the online training by on-line study new samples for random fern sorter, upgrades random fern posterior probability, improves gradually the precision of random fern sorter target detection, the complete autonomous on-line study of realize target detection system.

2, this patent introducing template set is subdued mechanism, can avoid in template set, the shortcoming that the more system running speed that may cause of positive negative sample template declines.

Accompanying drawing explanation

Fig. 1 is the process flow diagram of the inventive method;

Fig. 2 is random fern sorter structural drawing in one embodiment of the invention;

Fig. 3 is the comparison diagram that performance is detected in the random fern sorter front and back of one embodiment of the invention on-line study, and wherein Fig. 3 (a)-3 (d) is the testing result before on-line study, and Fig. 3 (i)-3 (l) is the testing result after on-line study;

Fig. 4 is the sorter autonomous learning procedure chart under illumination condition at night;

Fig. 5 is the sorter autonomous learning procedure chart of pedestrian detection;

Fig. 6 is the ROC curve comparison diagram of one embodiment of the invention and other classical on-line study process.

Embodiment

Below in conjunction with instantiation and accompanying drawing, the present invention will be further described.

The invention discloses the nearest neighbor classifier training method in the complete autonomous on-line study process of based target detection system research, the method only need select a target can carry out the sorter on-line study for this target class at frame of video center.Step is: first to the target of frame choosing, adopt affined transformation to obtain initial positive sample set, extract a small amount of negative sample training practice initial random fern sorter in the nontarget area of video; Secondly, use this sorter in frame of video, to carry out target detection.In the process detecting, adopt nearest neighbor classifier to collect on-line study new samples, and automatic decision sample class; Finally, the online training by new samples for random fern sorter, upgrades random fern posterior probability, improves gradually the precision of sorter target detection, the complete autonomous on-line study of realize target detection system.

The invention provides a kind of complete autonomous on-line study method based on random fern sorter as shown in Figure 1, comprise the steps:

1) prepare the sample set of initial training sorter:

For frame of video to be detected, at the first frame center of video image, select a target, this Target Photo is carried out to picture that affined transformation obtains as positive sample; Using do not contain target background image region as negative sample; The so random positive sample that obtains some and negative sample are as the sample set of initial training sorter.

Sample in the sample set of described initial training sorter is exactly the image block of formed objects in the present embodiment, and stock size is 15 * 15 (pixels), if in image block, contain target to be detected this sample be positive sample, be not negative sample.

2) random fern sorter initial training:

Use the sample set of ready initial training sorter to carry out initial training to random fern sorter, add up the posterior probability of positive negative sample on each random fern and distribute, as shown in Figure 2.

Concrete grammar is as follows:

2.1) construct random fern:

To on the single sample in sample set, get at random s to unique point as one group of random fern (the present embodiment selects 5 pairs), it is identical that each sample is got the position of unique point, every pair of unique point is carried out the comparison of pixel value, in every pair of unique point, greatly to get eigenwert be 1 to previous unique point pixel value, otherwise get eigenwert, be 0, the s that s obtains more afterwards to a unique point eigenwert forms the binary number of a s position according to random order, be the random fern numerical value that this organizes random fern, the sequence consensus of eigenwert in the random fern of each sample;

2.2) calculate the posterior probability of random fern numerical value in positive negative sample class:

In random fern, some obtains for positive sample, and other obtains for negative sample; The random fern F of each sample _lthe feature comprising can form a decimal number by gang, and because this decimal number obtains by S position binary code, the value kind of therefore random fern numerical value has 2 ^sindividual, have 2 ^splant (to be 2 in the present embodiment ⁵plant possibility);

Add up the positive number of samples of the value of every kind of random fern numerical value, thereby obtain random fern numerical value at positive sample class C ₁on posterior probability distribution P (F _l| C ₁); In like manner obtain random fern numerical value at negative sample class C ₀on posterior probability distribution P (F _l| C ₀); Combine all random ferns the sample set of initial training sorter is classified, be random fern sorter.

3) the good random fern sorter of initial training is traveled through to frame of video to be detected as current goal detecting device and carry out target detection, obtain object module, and calculate the degree of confidence of each object module, be specially: travel through frame of video to be detected, in frame of video, extract the image block of formed objects as sample to be tested, size and the step 1 of sample to be tested) in positive size equate, calculate the random fern numerical value of each sample to be tested, thereby obtain corresponding posterior probability, finally by random its classification of fern classifier calculated;

The image block that is positive sample for classification, is detected as target, becomes object module.

4) build positive negative sample template set:

Using following three kinds of samples as positive sample template, add positive sample template set M to ⁺, all the other add negative sample template set M to ^-:

A, step 1) in the positive sample that obtains;

B, to step 3) in the degree of confidence that obtains surpass the object module of degree of confidence preset value (desirable 0.6), adopt optical flow method to follow the tracks of and obtain tracking module its place frame of video, if tracking module and this object module have overlapping region, and coincidence factor surpasses default coincidence factor (default coincidence factor gets 60% conventionally), think that this tracking module is real goal, as positive sample template, add M to ⁺in;

C, to step 3) in the degree of confidence that obtains surpass the object module of degree of confidence preset value (desirable 0.6), adopt optical flow method to follow the tracks of and obtain tracking module its place frame of video, if tracking module and this object module have overlapping region, and coincidence factor does not surpass default coincidence factor, by conservative similarity S ^cjudge that can this tracking module add positive sample template set:

$S^c=\frac{S_{50\%}^{+}}{S_{50\%}^{+}+S^{-}}$

Wherein:

If S ^cbe greater than default conservative similarity threshold (desirable 0.6), this tracking module adds M as positive sample template ⁺, for the similarity of the first half template of sample to be sorted and current positive sample template set, S ⁺, S ^-be respectively the similarity of sample to be sorted and positive and negative samples template set, be the similarity of two picture frames, p ⁺, p ^-be respectively positive sample and negative sample, p is sample to be sorted, and in this step, sample to be sorted is tracking module;

Often add a positive sample template, get in same frame of video its around the image block of four formed objects determine whether negative sample, if add negative sample template set M as negative sample template ^-.In when judgement, introduce Gaussian Background modeling, if foreground pixel is less than foreground pixel threshold value (desirable be less than 30%) in image block, judge that it is negative sample.

Step 4) also comprise that template set subdues mechanism: the similarity of sample to be sorted and positive and negative template set equals in sample to be sorted and positive and negative template set the maximal value of similarity between single positive negative sample template; Each positive negative sample template of real-time statistics obtains this peaked number of times, if this peaked number of times that certain positive negative sample template obtains is less than maximal value number of times preset value, removes corresponding positive sample template or negative sample template.

5) use nearest neighbor classifier, obtain the positive negative sample of on-line study:

Arranging of nearest neighbor classifier is as follows: for each sample p to be sorted, calculate respectively itself and the similarity S of positive negative sample template set ⁺(p, M ⁺) and S ^-(p, M ^-):

$S^{+}(p,M^{+})={\max }_{p_i^{+}\∈M^{+}}S(p,p_i^{+})$

$S^{-}(p,M^{-})={\max }_{p_i^{-}\∈M^{-}}S(p,p_i^{-})$

Can obtain similarity S accordingly ^r:

$S^r=\frac{S^{+}}{S^{+}+S^{-}}$

If similarity S ^rbe greater than threshold value θ _nN, judge that this sample to be sorted is real goal, as the positive sample of on-line study; Otherwise be false-alarm, as the negative sample of on-line study;

In this step, sample to be sorted is step 3) object module and the step 4 that obtain) the positive negative sample template set that obtains.

(6) the online training of random fern sorter:

The positive negative sample of on-line study use step 5) obtaining, carries out on-line study to random fern sorter, improves gradually its nicety of grading; The random fern sorter of on-line study is carried out to target detection as the detection system of sustainable renewal.

The on-line study of random fern sorter realizes by upgrading posterior probability distribution, and concrete grammar is as follows:

6.1) using step 5) the positive negative sample that obtains is as on-line study sample; If an on-line study sample is (f _new, c _k), f wherein _newbinary number (f in the present embodiment for random fern s position _newbe 00101, decimal number 5), c _kfor sample class, calculate the random fern numerical value of this on-line study sample;

6.2) as shown in Figure 2, to step 2.1) classification is c in sample set _ktotal sample number add 1, classification is c _kthe sample number identical with the random fern numerical value of this on-line study sample add 1; The sample number of other random fern numerical value is constant, and (in the present embodiment, classification is c _ktotal sample number M add 1, random fern F _lnumerical value be 5 sample number N adds 1, the sample number N of other numerical value _otherconstant);

6.3) according to the sample number after upgrading, recalculate the posterior probability of random fern numerical value in this sample class distribute (in the present embodiment, random fern F _lnumerical value be 5 posterior probability becomes the posterior probability values of other numerical value becomes );

6.4) often increase an on-line study sample newly, just repeat 6.1) to 6.3) posterior probability is distributed and upgraded once.

By testing at field of traffic, as shown in Figure 3 (in realistic objective testing process, we use several different scales in video image, to carry out target detection, the frames images that different scale is corresponding varies in size, therefore can detect is the image block that frame is selected different sizes), wherein Fig. 3 a-3d is the testing result (only by the testing result of initial training) before on-line study, Fig. 3 e-3h is the testing result after on-line study, from figure, can find that initial training sorter is lower to the effect of target detection, through after training to the effect height of target detection a lot.

Fig. 4 is the sorter autonomous learning procedure chart under illumination condition at night, and the incipient stage that wherein Fig. 4 (a)-4 (d) is video, visible undetected more, owing to entirely independently training online, positive sample is less to be caused for this.Along with increasing of online training sample, verification and measurement ratio increases, and false-alarm also progressively increases, as shown in Fig. 4 (e)-4 (h).After the further on-line study of sorter, the posterior probability of its each random fern tends towards stability, and the vehicle target detecting is also tending towards accurately, as shown in Fig. 4 (i)-4 (l).

Fig. 5 is the sorter autonomous learning procedure chart of pedestrian detection, wherein Fig. 5 (a)-5 (d) is the detection case at complete autonomous on-line study initial stage, target detection situation after Fig. 5 (e)-5 (h) 200 frames that have been system autonomous learning, can find that from figure complete autonomous on-line study method can improve target detection performance gradually.

Fig. 6 is the ROC curve comparison diagram of one embodiment of the invention and other classical on-line study process, can find that complete autonomous on-line study method has good detection effect from figure.

Claims (6)

1. the complete autonomous on-line study method based on random fern sorter, is characterized in that: it comprises the steps:

1) prepare the sample set of initial training sorter:

For frame of video to be detected, at frame of video center, select a Target Photo, this Target Photo is carried out to picture that affined transformation obtains as positive sample; Using do not contain target background image region as negative sample; The so random positive sample that obtains some and negative sample are as the sample set of initial training sorter; Positive and negative samples is the image block that size is identical;

2) random fern sorter initial training:

Use the sample set of ready initial training sorter to carry out initial training to random fern sorter, add up the posterior probability of positive negative sample on each random fern and distribute;

3) the good random fern sorter of initial training is traveled through to frame of video to be detected as current goal detecting device and carry out target detection, obtain object module, and calculate the degree of confidence of each object module;

4) build positive negative sample template set:

Using following three kinds of samples as positive sample template, add positive sample template set M to ⁺, all the other add negative sample template set M to ^-:

A, step 1) in the positive sample that obtains;

B, to step 3) in the degree of confidence that obtains surpass the object module of degree of confidence preset value, adopt optical flow method to follow the tracks of and obtain tracking module its place frame of video, if tracking module and this object module have overlapping region, and coincidence factor surpasses default coincidence factor, think that this tracking module is real goal, as positive sample template, add M to ⁺in;

C, to step 3) in the degree of confidence that obtains surpass 0.6 object module, adopt optical flow method to follow the tracks of and obtain tracking module its place frame of video, if tracking module and this object module have overlapping region, and coincidence factor is over default coincidence factor, by conservative similarity S ^cjudge that can this tracking module add positive sample template set:

$S^c=\frac{S_{50\%}^{+}}{S_{50\%}^{+}+S^{-}}$

Wherein:

If S ^cbe greater than default conservative similarity threshold, this tracking module adds M as positive sample template ⁺, for the similarity of the first half template of sample to be sorted and current positive sample template set, S ⁺, S ^-be respectively the similarity of sample to be sorted and positive and negative samples template set, be the similarity of two picture frames, p ⁺, p ^-be respectively positive sample and negative sample, p is sample to be sorted, and in this step, sample to be sorted is tracking module;

Often add a positive sample template, get in same frame of video its around the image block of four formed objects determine whether negative sample, if add negative sample template set M as negative sample template ^-;

5) use nearest neighbor classifier, obtain the positive negative sample of on-line study:

Arranging of nearest neighbor classifier is as follows: for each sample p to be sorted, calculate respectively itself and the similarity S of positive negative sample template set ⁺(p, M ⁺) and S ^-(p, M ^-):

$S^{+}(p,M^{+})={\max }_{p_i^{+}\∈M^{+}}S(p,p_i^{+})$

$S^{-}(p,M^{-})={\max }_{p_i^{-}\∈M^{-}}S(p,p_i^{-})$

Can obtain similarity S accordingly ^r:

$S^r=\frac{S^{+}}{S^{+}+S^{-}}$

If similarity S ^rbe greater than threshold value θ _nN, judge that this sample to be sorted is real goal, as the positive sample of on-line study; Otherwise be false-alarm, as the negative sample of on-line study;

In this step, sample to be sorted is step 3) object module and the step 4 that obtain) the positive negative sample template set that obtains;

(6) the online training of random fern sorter:

The positive negative sample of on-line study use step 5) obtaining, carries out on-line study to random fern sorter, improves gradually its nicety of grading;

The random fern sorter of on-line study is carried out to target detection as the detection system of sustainable renewal.

2. the complete autonomous on-line study method based on random fern sorter according to claim 1, is characterized in that: concrete grammar described step 2) is as follows:

2.1) construct random fern:

To on the single sample in the sample set of initial training sorter, get at random s to unique point as one group of random fern, it is identical that each sample is got the position of unique point, every pair of unique point is carried out the comparison of pixel value, in every pair of unique point, greatly to get eigenwert be 1 to previous unique point pixel value, otherwise get eigenwert, be 0, the s that s obtains more afterwards to a unique point eigenwert forms the binary number of a s position according to random order, be the random fern numerical value that this organizes random fern, the sequence consensus of eigenwert in the random fern of each sample;

2.2) calculate the posterior probability of random fern numerical value in positive negative sample class:

In random fern, some obtains for positive sample, and other obtains for negative sample; The value kind of random fern numerical value has 2 ^sindividual;

Add up the positive number of samples of the value of every kind of random fern numerical value, thereby obtain random fern numerical value at positive sample class C ₁on posterior probability distribution P (F _l| C ₁); In like manner obtain random fern numerical value at negative sample class C ₀on posterior probability distribution P (F _l| C ₀); Combine all random ferns the sample set of initial training sorter is classified, be random fern sorter;

Described step 3) adopt above-mentioned random fern sorter in every frame video image, to carry out target detection:

Travel through every frame video image to be detected, in every frame video image, extract the image block of formed objects as sample to be tested, size and the step 1 of sample to be tested) in positive size equate, calculate the random fern numerical value of each sample to be tested, thereby obtain corresponding posterior probability, finally by random its classification of fern classifier calculated;

The image block that is positive sample for classification, is detected as target.

3. the complete autonomous on-line study method based on random fern sorter according to claim 1, it is characterized in that: described step 4) often add a positive sample template, when getting in same frame of video it around the image block of four formed objects determining whether negative sample, introduce Gaussian Background modeling, if foreground pixel is less than foreground pixel threshold value in image block, judge that it is negative sample.

4. according to the complete autonomous on-line study method based on random fern sorter described in claim 1 or 3, it is characterized in that: described step 4) also comprise that template set subdues mechanism: the similarity of sample to be sorted and positive and negative template set equals in sample to be sorted and positive and negative template set the maximal value of similarity between single positive negative sample template; Each positive negative sample template of real-time statistics obtains this peaked number of times, if this peaked number of times that certain positive negative sample template obtains is less than maximal value number of times preset value, removes corresponding positive sample template or negative sample template.

5. the complete autonomous on-line study method based on random fern sorter according to claim 2, is characterized in that: described step 6) on-line study of random fern sorter is distributed and realized by renewal posterior probability.

6. the complete autonomous on-line study method based on random fern sorter according to claim 5, is characterized in that: described step 6) concrete grammar is as follows:

6.1) using step 5) the positive negative sample that obtains is as on-line study sample; If an on-line study sample is (f _new, c _k), f wherein _newfor the binary number of random fern s position, c _kfor sample class, calculate the random fern numerical value of this on-line study sample;

6.2) to step 2.1) classification is c in sample set _ktotal sample number add 1, classification is c _kthe sample number identical with the random fern numerical value of this on-line study sample add 1; The sample number of other random fern numerical value is constant;

6.3), according to the sample number after upgrading, recalculate the posterior probability of random fern numerical value in this sample class and distribute;

6.4) often increase an on-line study sample newly, just repeat 6.1) to 6.3) posterior probability is distributed and upgraded once.