Weak Transient Signal Detection Based On Statistical Feature Correlation

Weak transient signal detection has been widely used in military reconnaissance,space surveillance and control,emergency rescue,biomedical and industrial measurement.However,the waveform and arrival time of transient signal are usually unknown,and in the most practical application,the signals are completely submerged in background signal and noise,which make it become great difficult to detect the target signal from the noise.To solve the problem,a weak transient signal detection method based on correlation of the statistical characteristics.In the proposed methods,the weak signals are detected by using the slight perturbation of statistical characteristics which is brought by the target signal.And in the detection,the great effort are made to correlate slight perturbation with the target signal.Then the methods are applied as detector in the detecting dim moving point target from high frame rate image sequence.The dim moving point target detection methods based on high frame image sequence detect the moving target by detecting the series signal of the moving target passing through a pixel in the image sequence,and high frame rate sampling can ensure that we can collect as much information as possible about the weak transient signal.The basic theory of weak transient signal detection based on statistical feature correlation and the basic theory of moving target detection based on high frame rate image sequence are firstly introduced in this paper.In the introduction of basic theory of weak transient signal detection based on statistical feature correlation,the common statistical features in weak signal detection and the commonly used methods of feature correlation in practical application are summarized,then the model of weak transient signal detection in the paper is introduced.In the introduction of the basic theory of moving target detection based on high frame rate image sequence,the detection model is firstly introduced,then the simulation model of moving point target in optical imaging system is introduced,which will be widely used in the detection,finally,the feasibility of dim moving target detection based on high time phase is theoretically analyzed and deduced.Aiming at the problem of low signal-to-noise ratio(SNR)of target signal in weak transient signal detection,and the target signal in weak moving target detection based on high frame rate image sequence is a kind of impact signal,a denoising algorithm for weak impact signal based on wavelet packet transform is proposed.In the proposed method,a new wavelet coefficients processing function is proposed.Through the function the wavelet coefficients related to noise are suppressed and the wavelet coefficients related to the target signal are enhanced at the same time,which is different from the traditional threshold functions of wavelet coefficients remove the noise by setting the wavelet coefficients whose absolute value are less than the threshold.For the target signal is completely submerged in noise and background signal in weak transient signal detection,a new weak transient signal detection method based on statistical feature correlation in kernel space is proposed.The weak transient signals linearly inseparable in low-dimensional space are mapped to high-dimensional feature space by kernel function,and weak transient signals are detected by feature correlation in high-dimensional feature space.The greatest advantage of using kernel function is that the correlation analysis of statistical eigenvectors in high-dimensional space can be completed when both feature space and mapping function are unknown.At the same time,by mapping transient signals into high-dimensional space,the problem of linear inseparability in low-dimensional space can be transformed into linear separability in high-dimensional space.In the weak transient detection method based on statistical feature correlation in the feature space of the kernel function,the feature in feature space of kernel function can not completely extract the characteristics of the target signal.A weak transient signal detection method based on the statistical feature space correlation is proposed.The new feature space is constructed on the feature space of the kernel function,the statistics sensitive to the target signal are selected to extract the feature of the target signal from different angles.Two feature extraction and(Principal Component Analysis)PCA are used to reduce the dimension of feature space to avoid “Curse of Dimensionality”.In this paper,the correlation between statistical features and target signals is acquired by machine learning method,and the data for training machine learning model is simulated by moving target simulation model.The experimental results show that this method has better detection performance.The detection methods proposed above in this paper is invalid when the background of the target signal changes.In this chapter we focused on detecting weak transient signal from changing background signal,and the background change is divided into two kinds: continuous background change and discontinuous background change.For discontinuous background change,according to the detection model of dim moving target detection in high temporal phase,a discontinuous change background signal removal algorithm under uniform motion of platform is proposed.The optimal discontinuous change background is fitted by gray time series signals of adjacent pixels.A weak transient signal detection method based on kernel function depth correlation is proposed for continuously changing background signals.The simulation experiment proves that the method proposed in this chapter has better detection results in both continuous change background and non-continuous change background.