FIELD OF INVENTION

The present invention relates generally to the identification of wood species. More particularly, the invention relates to identify the wood species automatically without human intervention.

BACKGROUND OF THE INVENTION

More than 1 ,000 species of the wood are available in the Malaysian jungles. Since quality, uses and costs are dependent on the wood species, the identification of wood species is first and foremost step for the wood industry. To identify the wood based on anatomy is done by only few certified personal. The training for performing the identification of the wood is a tedious task.

The wood identification has been done based on the texture of wood species. In practice, the analysis of texture of the wood species requires an identification of texture attributes of the wood species which may be used for segmentation, discrimination, and recognition or shape computation. Analysis of texture attributes requires a vast knowledge and experience in wood recognition. Apart from the chance of human error, there is a great chance of biasness in identification of the wood species. To avoid the human error and bias, there is a need of automated the process for the recognition of the wood species.

Accordingly, the present invention proposes the automatic wood species recognition. SUMMARY OF THE INVENTION

In general, in one aspect, the invention features a method for recognizing of at least one wood species. In particularly, the method acquires an image of the at least one wood species for analyzing the image using an image acquisition module (IAM) (220). In addition, the method processes the image for enhancing quality of the acquired image using a pre processing module (PPM) (230). Additionally, the method extracts a plurality of features of the processed image for classifying at least one pattern using a feature extraction module (FEM) (240). Further, the method classifies the at least one partem for the recognizing the at least one wood species using a pattern classification module (PCM) (250).

In general, in another aspect, the invention features a system for recognition of at least one wood species. Moreover, an image acquisition module (IAM) (220) adapted to acquire an image of the at least one wood species for analyzing the image. In particular, a pre processing module (PPM) (230) adapted to process the acquired image for enhancing quality. Further, a feature extraction module (FEM) (240) adapted to extract a plurality of features of the processed image for classifying at least one pattern. Furthermore, a pattern classification module (PCM) (250) adapted to classify at least one pattern for recognizing the at least one wood species.

In general, in yet another aspect, the invention features a computer program, stored on a tangible storage medium, for recognition of at least one wood species. The program acquires an image of the at least one wood species for analyzing the image using an image acquisition module (IAM) (220). Particularly, the program features a process for enhancing quality of the acquired image using a pre processing module (PPM) (230). In particularity, the program extracts a plurality of features of the processed image for classifying at least one pattern using a feature extraction module (FEM) (240). In addition, the program classifies the at least one pattern for the recognizing the at least one wood species using a pattern classification module (PCM) (250). BRIEF DESCRIPTION OF THE DRAWINGS

These and other features, aspects, and advantages of the present invention will become better understood when the following detailed description is read with reference to the accompanying drawings in which like characters represent like parts throughout the drawings, wherein:

Fig. 1 is a flow chart illustrating automatic wood species recognition using automatic wood species recognition method and system (AWSRMS), in accordance with an aspect of the present technique;

Fig. 2 is a block diagram depicting a process of automatic wood species recognition using automatic wood species recognition method and system (AWSRMS), in accordance with an aspect of the present technique;

Fig. 3 is a block diagram depicting the image acquisition module (IAM), in accordance with an aspect of the present technique;

Fig. 4 is a block diagram depicting the feature extraction module (FEM), in accordance with an aspect of the present technique; and

Fig. 5 is a block diagram depicting the pattern classification module (PCM), in accordance with an aspect of the present technique.

DETAILED DESCRIPTION OF THE INVENTION

The present invention proposes automatic wood species recognition, which is one of the major needs for the wood industry. In particular, the invention is an integrated method and system for recognizing wood species. In general, recognition of the wood species may be done by analysis of texture attributes of the wood species. The analysis of texture requires the identification of those texture attributes which can be used for segmentation, discrimination, recognition, or shape computation. However present invention uses the computer vision which includes image analysis and statistical classification to explore quantitative features of the wood anatomy.

In many machine vision and image processing algorithms, simplified assumptions are made about the uniformity of intensities in local image regions. However, images of real objects often do not exhibit regions of uniform intensities. In the present invention, the macroscopic anatomy of the wood image is not uniform. It contains variations of intensities form which certain repeated patterns can be identified know as visual texture. The patterns are made up of several structural elements such as vessels or pores, wood parenchyma or soft tissue, rays and fibers.

Referring to Fig. 1 is a flow chart illustrating automatic wood species recogmtion using an automatic wood species recognition method and system (AWSRMS). As illustrated, methodology starts at step 100, wherein a wood sample (210) may be taken for the recognition. The wood sample (210) may be divided into cubes of approximately 1 inch by 1 inch in size for maintaining the consistency. At step 102, the image of the wood sample (210) may be acquired for analysis using an image acquisition module (220). At step 104, the acquired image may be processed to enhance the quality of the image using a pre processing module (230). Typically, the processed image may be up to the mark to qualify for analysis phase. At step 106, extraction of one or more features of the processed image may be done using a feature extraction module (FEM) (240). The feature may include the texture attributes of the macroscopic anatomy of the wood species. The texture attributes used for recognition are vessels or pores, wood parenchyma or soft tissue, rays and fibers and are not limited to other texture attributes known in the art. Further at step 108, the classification of the extracted features may be done using a pattern classification module (PCM) (250). More particularly, the mapping of the extracted features to a pre-built image database (IDB) (260) may be done. Finally at step 110, the wood species may be recognized based on the feature classified with the use of the pre-built image database (TDB) (260).

Referring to Fig. 2 is a bock diagram depicting the process of automatic wood species recognition using the automatic wood species recognition method and system (AWSRMS). The AWSRMS (200) further comprises the wood sample (210), the image acquisition module (IAM) (220), the pre processing module (PPM) (230), the feature extraction module (FEM) (240), the pattern classification module (PCM) (250) and the pre-built image database (IDB) (260). The image of the wood sample (210) of particular size and shape may be taken using the Image acquisition module (LAM) (220).

In one embodiment, the image acquisition module (IAM) (220) is made in such a way that the captured image quality should be good and consistent with the images in the pre-built image database (IDB) (260) which are used for training purpose. The quality of images varies due to several factors such as illumination, magnification or field of view, pattern and format. The image acquisition module (IAM) (220) is equipped in such a way that it maintains the same quality of image keeping these factors as constant and consistent as possible.

In another embodiment, the pre processing module (PPM) (230) may use the high-pass filtering to sharpen the image in order to give a clearer definition of the texture properties of the macroscopic wood anatomy. Further, the goal in image sharpening is to highlight the fine details of the image. The processed image may be.used as an input for feature extraction module (FEM) (240).

In another embodiment, the feature extraction module (FEM) (240) may extract information, features or properties from the texture of the processed image. Additionally, visual texture may contain variations of intensities form which certain repeated patterns can be identified know as visual texture. These variations in patterns are due to different physical surface properties and also exposure to environmental factors. Therefore, texture analysis method may be used to extract the distinct features of each wood. The features may be used as the input vector for the next module. In another embodiment, the pattern classification module (PCM) (250) may classify the wood species using the extracted features. In general, classification is a process that assigns input data into one or more of specified classes based on extraction of significant features or attributes and the processing or analysis of these attributes. The pattern classification module (PCM) (250) may use the pre-built image database (TDB) (260) to classify the attributes to recognize the wood species.

In yet another embodiment, the pre-built image database (IDB) (260) may be developed for the wood recognition using the known wood species. Firstly, samples of the similar sizes of the wood species may be taken. Thereafter, the images of the samples may be processed and the database may be trained and tested. Finally, the pre- built image database (IDB) (260) may be used to classify and recognize the wood species.

Referring to Fig. 3 is a bock diagram depicting the image acquisition module (IAM). The image acquisition module (IAM) (220) further comprises of an image capturing unit (222), an image capturing device (224), a source of light (226) and a connecting unit (228). As shown, the prepared wood sample (210) may be considered as the input for the image acquisition module (IAM) (220). The image acquisition module (IAM) (220) may be specially designed to make sure that the acquired image is consistent with the existing images in the database in terms of brightness, field of view, magnification, and format.

In one embodiment, the image capturing unit (222) may be equipped with a systematic focusing function, whereby the distance between the camera and the wood sample (210) is constant and need not be adjusted. In general, this may be done by calculating the expected field of view, at the shortest distance. Typically, the image capturing unit (222) is a tube based on the theoretical optimal object distance, and therefore the wood sample (210) may be just required to be laid against the image capturing unit (222).

In another embodiment, the image capturing device (224) is a special type of device which may be used for capturing the image. The image capturing device (222) is fixed in the image capturing unit (222) in such a way to maintain the distance with the wood sample (210). Further, the image capturing unit (222) may be connected to the software to acquire the image of particular specification.

In another embodiment, the source of light (226) may use the spatial resolution of the lighting to maintain the quality of the image to be more consistent. Specifically, the source of light (226) may need to provide constant optical power, enough optical power for dark samples and homogeneous lighting repartition on the wood sample (210). The source of light (226) may be an LED display that may be used to obtain the requisite quality of the image.

In yet another embodiment, the connecting unit (228) may act as a data transfer unit between the image acquisition module (IAM) (220) and software portion of the other modules. Thus the system becomes portable. Further the connecting unit (228) may facilitate the use of online and offline mode of the wood recognition. Specifically, offline version means that the images may be recognized using the image database (260) attached to the integrated system. For online version, the images may be captured automatically using the image capturing device (224) for online recognition with the up to date image database (260). The software for the wood recognition system may be developed using the programming language like C#. The connecting unit (228) may use the technology like USB (Universal Serial Bus), which is much more cost effective and can supply the power the power through USB (Universal Serial Bus) itself.

Referring to Fig. 4 is the Feature Extraction Module (FEM). The FEM (240) comprises of a first set of techniques. The first set of techniques further comprises of various feature extraction methods such as a gray level co occurrence matrix (GLCM) (242), a basic gray level aura matrix (BGLAM) (244), a gabor filter (246) and a wavelet (248). It should be noted that the pre processed image which is the output of pre processing module (PPM) (230) may be used as input for feature extraction module (FEM) (240). Feature extraction is an important process to extract features or properties from the texture of pre processed image. In one embodiment, features may be extracted from the image to transform the image into a suitable representation, which is in the form of features vectors. Any of the above feature extraction method may be employed to transform an image into set of numerical values, called features. The extracted features may be used as the input for the pattern classification module (PCM) (250).

In another embodiment, visual texture of the any wood sample may contain variations of intensities, which form certain repeated patterns. These variations in patterns are due to different physical surface properties. In addition exposure to environmental factors may also generate variations in patterns. Therefore, texture analysis method may be used to extract the distinct features of each wood sample.

In another embodiment, a user may choose any of the feature extraction methods such as the gray level co occurrence matrix (GLCM) (242), the basic gray level aura matrix (BGLAM) (244), the gabor filter (246) and the wavelet (248) or the combination of the above.

In yet another embodiment, wood textures may be characterized by statistical means into first, second and higher-order statistic. Therefore, a texture analysis method may be used to extract the distinct features of each wood sample. The gray level co occurrence matrix (GLCM) (242) which have become one of the most well-known and widely used texture feature extraction method, may be employed for feature extraction,. In this approach, textural features of an image are based on the assumption that the texture information on an image is contained in the overall or average spatial relationship which the grey tones in the image has with one another. More specifically, this texture information is adequately specified by a set of grey tone spatial dependence matrices; that are computed for various angular relationships and distances between neighboring resolution cell pairs on the image. The features from these grey tone spatial dependence matrices. The GLCM approach can be described as follows. Consider {J(x, y),0≤x≤N - \,0≤y≤N - l} _such ^ _{¾ denQtes m} » - » ^ _G levels. The ^{G x G} grey level co-occurrence matrix Pd for a displacement vector d = (dx, dy) is defined as follows. The entry (ij) of Pd is the number of occurrences of the pair of grey levels i and j which are a distance d apart. Formally, it is given as in Equation (1) where: r, _s ), (t, v) e N x N, (t, _v ) = (r ₊ dx, _s + dy) _{) ani} \\ _is the cardinality of a set.

W J) = j}\

For each wood cube, the co-occurrence matrices are calculated from four directions, which are horizontal, vertical, diagonal 45° and diagonal 135°. A new matrix is formed as the average of these matrices that is used for extracting the features. In this way, the extracted features will be rotation invariant at least for 4 steps of rotation. The final cooccurrence matrix is normalized using Equation (2) to transform GLCM matrix into a close approximation of the probability table.

PC \ ^p ^ ^j) = o (2)

where Pd is GLCM matrices value of and N is range of i and j .

The total features extracted using the GLCM approach from each wood sample orientations are given as follows:

Angular Second Moment

Contrast (4) Correlation

∑∑ )P _V - μ _χ μ _γ

f ₌ -i_i

(5)

Where μχ and μ are mean value and σχ and ay are standard deviation.

Entropy

Inverse Difference Moment

1

2*9

(7)

For each orientation of an image, there are 5 features to be extracted which are angular second moment, contrast, correlation, entropy and inverse difference moment. This will give the total of 40 features to be extracted using this feature extractor for a single image since we are using 4 orientations.

Referring to Fig. 5 is pattern classification module (PCM). The pattern classification module PCM (250) comprises of a second set of techniques. The second set of techniques further comprises of an artificial neural network module (252) and a k- nearest neighbor module (254).

The pattern classification module (PCM) (250) may be employed to assigns input data into one or more of specified classes based on extraction of significant features or attributes. Further the output from feature extraction module (FEM) (240) will be trained using any or both of the artificial neural network module (252) and the k- nearest neighbor module (254) to find the best answer for the wood species. It should be noted that the extracted image which is the output of feature extraction module (FEM) (240) may be used as input for pattern classification module (PCM) (250). A classification problem may occur when an object needs to be assigned into a predefined cluster or class based on a number of observed related to that objects. Texture may also be characterized not only by grey value at a given pixel, but also by the grey value "pattern" in a neighbourhood surrounding pixel where the brightness level at a point depends on the brightness levels of neighbouring point.

The user can select artificial neural network module (252), the k-nearest neighbor module (254) or a combination of both to classifier the wood species. The pre-built image database (TDB) (260) of known wood species images may be employed for recognition of the wood species. The overall method is optimized in such a way to deliver accurate resulted in just a short time.

The foregoing description of the embodiments of the invention has been presented for the purposes of illustration and description. It is not intended to be exhaustive or to limit the invention to the precise form disclosed. Many modifications and variations are possible in light of the above teaching. It is intended that the scope of the invention be limited not by this detailed description, but rather by the claims appended hereto.