High-Speed Imaging And Defect Detection For Wooden Material By Using Air-Coupled Ultrasonics

Air-coupled ultrasonics(ACU)removes the inherent limitation of contact ultrasonics by using air as the coupling medium and provides an efficient noncontact inspection resolution.The main advantage is that no couplant is needed like regular ultrasonic testing.The wooden samples would therefore not be contaminated during testing.Moreover,the absence of couplant allows for testing with high efficiency and high reproducibility.ACU also allows for continuous scanning with arbitrary transducer orientation and step size,so that measurements can be taken over the whole surface of the test sample rather than at specific points.Now the ACU techniques are inadequate for on-line testing for industrial applications.It is necessary to conduct further studies aiming to improve the ACU scanning imaging speed and improve the automation of wood defects detection.In this work,many algorithms,such as compressed sensing,pattern recognization,and digital image processing,were utilized to improve the ACU techniques.A three-layer model of ACU wave propagation in the wood nondestructive testing system was established.An ACU scanning imaging system for wooden material testing was developed.A fast ACU imaging method,wood defect recognization method and quantitative detection method were proposed.As a result,the higher accuracy and practicability of the ACU-based wood testing technique were achieved.The main research contents and conclusions are summarized as follows:(1)The propagation characteristics of ACU waves at the interface between wood and air was firstly discussed based on the boundary conditions of acoustic pressure and particle velocity.On the basis of this,a three-layer model was established to simulate the ACU wave propagation in the wood nondestructive testing system.Then the conclusion can be drawn that the transmission coefficient of ACU waves is not only related to the specimen thickness,but also to the acoustic impedance of the material to be measured.Wood is a typical heterogeneous material.Many factors,such as moisture content,sapwood,heartwood,earlywood,latewood,knot,and the early decay will cause variation in density of the material and acoustic velocity,resulting in changes in the propagation attenuation.Therefore,it is possible to evaluate properties of wooden materials with the propagation attenuation of ACU waves.(2)A scanning imaging system has been developed based on the principle of the ACU A-scanning and C-scanning.The ACU transducers work in transmitted-through mode and the center frequency of the ACU transducers is 125 KHz.The structure of the scanning imaging system was described,which comprises the pulser&receiver,the gantry scanner,and the computer software.The design methods of the mechanical system,circuit system,and control software are introduced in detail.With the developed system,A-scanning and C-scanning imaging can be implemented easily under the parameters the users setted.Experiments were conducted with the developed system,and wood defects,including knots,cracks,decay,insect damage,internal defects in sandwich board,and adhesive defects in multilayer boards,were detected.The effects of density change and internal crack on the propagation of the waves were discussed based on the analysis of the recorded signals and the effectiveness of the ACU was verified by comparing the imaging results and the photos of the samples.It can be concluded that the developed scanning imaging system can be used for the non-destructive testing of wood.(3)It has been proved that the compressed sensing theory can recover the original sparse signal by using a small amount of data.It can provide a solution for high-speed imaging method.However,it is difficult to apply the compressed sensing to the ACU scanning imaging directly because of limitation of the mechanical system.A random binary matrix was proposed to address the limitation to the application of CS caused by hardware system.Under the proposed matrix,the undersampled scanning can be easily implemented,while only minor modification was required to the existing imaging system.Discrete cosine transform was selected experimentally to represent the undersampled data sparsely.Finally,orthogonal matching pursuit was utilized to recovery the wood images with high accuracy.Experiments on three real air-coupled ultrasound images indicated the potential of the present method to accelerate air-coupled ultrasound imaging of wood.The same quality of ACU images can be obtained with scanning time cut in half.(4)Knot is a kind of typical wood defect and will have a great impact on the mechanical properties and the industrial processes of the wood product.Thus,it is one of the important topics in the field of nondestructive testing of wood to recognize the knots automatically.A method to recognize knots automatically was presented by fusing the gray feature and texture feature of the ACU images.At first,gray histogram of the ACU images of wood was computed.Then,seven statistics,including average gray value,variance,skewness,kurtosis,smoothness,entropy,and energy,was calculated as the gray feature of the knot defect.At the same time,the ACU image was scanned by the LBP operator with uniform pattern and the LBP-coded image was obtained.The normalized LBP histogram was subsequently extracted as the texture feature of the knot defect.Finally,the ultimate feature vector can be obtained by combining the gray feature and the texture feature.The support vector machine(SVM)was utilized to recognize the knot defect with the feature vector.The results showed that a higher recognization accuracy can be achieved by fusing the gray feature and the texture feature.When only the gray feature was used,the recognization accuracy was 71.06%.While only the texture feature was used,the recognization accuracy was 82.41%.The highest accuracy of 96.30%can be obtained with the fused feature vector.(5)It is a key step for wood inspection to detect the knot quantatitively and obtain the accurate ratio of the knot diameter and the wood diameter.A new method to detect knot quantatitively was proposed with the ACU images of wood and algorithms in the field of the digital image processing.Firstly,Otsu,morphological method,and LevelSet were compared in segmenting the knot from ACU images.The results showed that the knot profile obtained by LevelSet method was most close to the real knot.Additionally,the LevelSet method can provide smoother result compared with other two methods.Therefore the LevelSet was selected as the tool of knot segmentation.Then,a mathematical regression model between the knot size and the pixels of the ACU images was established with some round aluminum sheet with different diameters.At last,the width and length of the knot showed in the ACU image can be determined with this regression model.Experiments were conducted on twelve Chinese fir with fifteen knots.The results showed that the length and the width of the knot can be determined accurately with the proposed method.The absolute error is less than 2 mm.The method has great potential in the application.