Defect detection system for lumber

• US 5,960,104 A
• Filed: 08/16/1996
• Issued: 09/28/1999
• Est. Priority Date: 08/16/1996
• Status: Expired due to Fees

• Current Assignee: The United States of America As Represented By The Secretary of Agriculture, Virginia Tech Intellectual Properties, Inc. (Virginia Polytechnic Institute and State University)
• Original Assignee: United States Department Of Agriculture, Virginia Polytechnic & State University
• Inventors: Conners, Richard W., Xiao, Xiangyu, Kline, David E., Araman, Phillip A., Drayer, Thomas H.
• Primary Examiner(s): Boudreau, Leo H.
• Assistant Examiner(s): Werner, Brian P.
• Application Number:
  US08/707,982
• Time in Patent Office: 1,138 Days
• Field of Search: 382/108, 382/110, 382/141, 382/154, 382/168, 382/171, 382/170, 382/173, 348/86, 348/88, 348/93, 348/125, 348/128, 356/371, 356/376, 356/381, 356/372, 356/406, 364/468.17, 364/469.01, 364/507, 250/559.25, 250/559.05, 250/559.08, 250/559.11, 209/517, 209/518, 144/391, 144/392, 144/398, 144/402, 144/403, 144/404, 144/408
• US Class Current: 382/141
• CPC Class Codes: G01N 21/8986 Wood
• CPC Class Codes: G01N 33/46 Wood
• CPC Class Codes: G06T 2207/30161 Wood; Lumber
• CPC Class Codes: G06T 7/001 using an image reference approach
• CPC Class Codes: G06V 10/28 Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns

Abstract

A machine vision system that can address a number of board, lineal, cant, and flitch inspection problems by incorporating all the sensors needed to address the surface feature detection problem, the three-dimensional shape detection problem, and the internal feature detection problem. To detect surface features, two color cameras are employed, one for imaging each of the major faces of a board, lineal, cant, or fitch. To address the three-dimensional shape detection problem, a high speed laser profiling device is employed. An x-ray scanning system is employed to detect internal features. The system is able to process material in a species-independent manner by using a histogram-based segmentation procedure for analyzing both the camera imagery and the x-ray imagery; and can detect small defects by removing the effects of large features from the histograms once they have been detected. The system also utilizes redundant information from the set of multiple sensors to improve system accuracy. The volume of data that must be analyzed due the use of three sets of sensors is reduced by ordering the way the data is analyzed. The laser profile data is processed first, followed by the x-ray data and the color imagery. Finally, the system reduces the required volume of data by incorporating a crack/check preserving filter. This filter is implemented in special purpose hardware, and filters the color imagery as it is collected.

Claims

1. Apparatus for inspecting lumber to determine the presence and location of defects to optimize cutting of the lumber, said apparatus comprising:

first and second color cameras for imaging first and second faces of the lumber and generating color image data thereof;

optical scanning means for imaging the whole width of the lumber for generating image data of a profile of the lumber;

electro-magnetic scanning means for generating image data of internal features of the lumber;

means for generating histograms corresponding to the color image data and the internal features image data;

dynamic selecting means for dynamically selecting histogram threshold levels for each piece of lumber relative to a peak in each of the histograms representing clear wood;

data ordering means for analyzing the profile image data from said optical scanning means first, for analyzing the internal features image data from said electro-magnetic scanning means second to locate large defect regions, for analyzing the internal features image data from said electro-magnetic scanning means again to locate small defect regions, and for analyzing the color image data from said first and second cameras last; and

data removing means for finding regions in the image data that are known to be defect areas based on the histogram threshold levels selected by said dynamic selecting means and removing them from all image data and their corresponding histograms once found.

2. The system of claim 1, further comprising filter means for filtering the color image data as it is collected to reduce the required volume of color image data while preserving suspect defect regions.

3. The apparatus of claim 1, wherein said electro-magnetic scanning means, said first and second color cameras, and said optical scanning means are arranged at different locations from each other, whereby said electro-magnetic scanning means, said first and second color cameras, and said optical scanning means start scanning the lumber at different times;

and

wherein said apparatus further comprises;

memory means for storing image data from said first and second color cameras, said optical scanning means, and said electro-magnetic scanning means;

pre-processing means for collecting image data from said first and second color cameras, said optical scanning means, and said electro-magnetic scanning means, synchronizing the image data as it is collected, and simultaneously transferring the data to said memory means; and

registering means for registering the image date collected by said pre-processing means and stored in said memory means.

4. The apparatus of claim 1, wherein said data removing means includes means for cross-referencing the profile image data and the color image data with the internal features image data to measure attributes of the lumber and means for identifying suspect defect areas found from the internal features image data using the attributes measured using said means for cross-referencing.

5. The apparatus of claim 4, wherein said data removing means uses fuzzy logic to assign a defect class based on the measured attributes.

6. The apparatus of claim 1, further comprising:

a computer memory; and

means for storing images into said computer memory such that said apparatus can image a piece of lumber and generate the image data therefor while simultaneously processing the image data generated by the imaging of a previous piece of lumber.

7. A method for inspecting lumber to determine the presence and location of defects to optimize cutting of the lumber, said method comprising the steps of:

(a) obtaining color image data for first and second faces of a piece of lumber;

(b) obtaining profile image data for the whole width of the lumber;

(c) obtaining image data of internal features of the lumber;

(d) generating histograms corresponding to the color image data and the internal features data;

(e) analyzing the profile image data using a computer processor;

(f) dynamically selecting histogram threshold levels for the lumber relative to a peak representing clear wood in each of the histograms for the internal features image data;

(g) following said steps (e) and (f), analyzing the internal features image data using the computer processor to locate large defect regions based on the analysis of the profile image data from said step (e) and based on the histogram threshold levels selected in said step (f) and removing the large defect regions so located from all the image data and their corresponding histograms;

(h) following said step (g), dynamically selecting histogram threshold levels for the lumber relative to a peak representing clear wood in each of the histograms for the internal features image data as modified in said step (g);

(i) following said step (h) analyzing the internal features image data again using the computer processor to locate small defect regions including finding regions, in the internal features image data that are known to be defect areas based on the histogram threshold levels selected in said step (h) and once found, removing them from all the image data and their corresponding histograms as modified in said step (g);

(j) following said step (i), dynamically selecting histogram threshold levels for the lumber relative to a peak representing clear wood in each of the histograms for the color image data as modified in said step (i);

(k) following said step (j), analyzing the color image data using the computer processor, including finding regions in the color image data that are known to be defect areas based on the histogram threshold levels for the color image data selected in said step (j); and

(l) using the analysis of of said steps (e), (g), (i), and (k) to optimize cutting of the lumber.

8. The method of claim 7, wherein said step (e) includes identifying areas with insufficient thickness, and wherein said step (g) includes removing the areas of insufficient thickness identified in said step (f) from all of the histograms generated in said step (d).

9. The method of claim 8, wherein said steps (g) and (i) include identifying areas which are less dense than clear wood, and removing the less dense areas from all of the image data and their corresponding histograms.

10. The method of claim 9, further comprising the step of using a crack/check preserving filter to reduce the number of color pixels to be stored and analyzed, prior to said step (d).

11. The method of claim 10, wherein said step of using a crack/check preserving filter comprises:

creating disjoint N×

M sub-arrays of the color image data which in totality completely cover the color image, each point in the N×

M sub-arrays representing a color pixel;

finding the color pixel in each N×

M sub-array that has the darkest color; and

choosing the darkest color in each N×

M sub-array to represent the whole N×

M sub-array.

12. The method of claim 7, wherein said step (g) is performed using a histogram-based segmentation procedure including the steps of:

smoothing the histograms corresponding to the internal features data;

looking for inflection points in the smoothed histograms of the internal features image data;

using the inflection points to detect large features; and

removing the effects of the detected large features from the smoothed histograms to create modified histogram data; and

wherein said step (i) includes using the modified histogram data to detect smaller wood features.

13. The method of claim 7, wherein said steps (g) and (i) include cross-referencing the profile image data and the color image data with the internal features image data to measure attributes of the lumber and identifying suspect defect regions found from the internal features image data using the measured attributes.

14. The method of claim 7, wherein said steps (g), (i), and (k) are carried out using a segmentation procedure and a connected-component labeling algorithm.

15. The method of claim 7, wherein in said steps (g) and (i), said analyzing is performed using fuzzy concepts to define feature classes for the lumber.

16. Apparatus for inspecting a wood object to determine the presence and location of defects in the object, said apparatus comprising:

means for generating color image data representing surface features of an object;

means for generating profile image data representing the three-dimensional shape of the object;

means for generating x-ray image data representing internal features of the object;

first and second feature detecting computers to process the color, profile, and x-ray image data and locate defects, said first computer processing information obtained from a first surface of the object and said second computer processing information obtained from a second surface of the object, said first and second computers each including a memory;

means for registering the color image data, the profile image data, and the x-ray image data by establishing a common coordinate system in two-dimensional space and by ensuring that the spatial resolution of each set of image data is the same;

means for generating histograms corresponding to the color image data, the profile image data, and the x-ray image data;

means for analyzing the color image data, the profile image data, and the x-ray image data using a histogram-based segmentation procedure to produce a respective segmented profile image, segmented color image, and segmented x-ray image;

means for finding regions in each of the segmented profile, color, and x-ray images that are known to be defect areas by applying a connected component labeling algorithm to each of the segmented profile, color, and x-ray images and for removing the defect areas from each of the segmented profile, color, and x-ray images and its corresponding histogram, once found;

means for ordering the analysis of the segmented profile, camera, and x-ray image data to decrease the volume of data that must be analyzed by;

first processing the profile data to locate areas of insufficient thickness, removing areas of insufficient thickness from the x-ray image data and then analyzing the remaining x-ray image data to locate areas of less density, removing areas of insufficient thickness and less density from the color image data and then analyzing the remaining color image data to confirm defects having both surface and internal features;

image preprocessing means for preprocessing the image data from said means for generating color image data, said means for generating profile image data, and said means for generating x-ray image data, said preprocessing including synchronizing of the image data as it is collected; and

interface means for transferring image data from each of said means for generating color image data, said means for generating profile image data, and said means for generating x-ray image data into computer memory in a standard direct memory access method so that no CPU intervention is required and for filtering the color image data as it is collected using a crack-preserving filter to reduce the resolution of the volume of the color image data while preserving the level of detail needed to locate and identify small cracks.

17. The apparatus of claim 16, wherein the histogram-based segmentation procedure used by said analyzing means considers the largest peak in the histogram to be caused by the characteristics of clear wood, and looks for deviations from these characteristics by looking for inflection points in a smoothed histogram.

18. The apparatus of claim 16, further comprising an integrated evaluation and control computer for processing object feature locations, identities, and characteristics passed to it by said first and second computers for formulating a strategy for sawing the object based on the defects located by said first and second feature detecting computers.

19. The apparatus of claim 16, further comprising:

a computer memory; and

means for storing images into said computer memory such that said apparatus can generate the image data for one object while simultaneously processing the image data generated in connection with a previous object.