Classifications

• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
  • G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
  • G06F16/51—Indexing; Data structures therefor; Storage structures
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
  • G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
  • G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
  • G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
  • G06F16/5838—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using colour
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T7/00—Image analysis
  • G06T7/0002—Inspection of images, e.g. flaw detection
  • G06T7/0004—Industrial image inspection
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T7/00—Image analysis
  • G06T7/0002—Inspection of images, e.g. flaw detection
  • G06T7/0004—Industrial image inspection
  • G06T7/0008—Industrial image inspection checking presence/absence
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T7/00—Image analysis
  • G06T7/0002—Inspection of images, e.g. flaw detection
  • G06T7/0004—Industrial image inspection
  • G06T7/001—Industrial image inspection using an image reference approach
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T7/00—Image analysis
  • G06T7/40—Analysis of texture
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T7/00—Image analysis
  • G06T7/90—Determination of colour characteristics
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/10—Image acquisition modality
  • G06T2207/10024—Color image
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/20—Special algorithmic details
  • G06T2207/20021—Dividing image into blocks, subimages or windows
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/20—Special algorithmic details
  • G06T2207/20212—Image combination
  • G06T2207/20224—Image subtraction

Definitions

• the invention relates to the detection and referencing of streaks in images.
• a striated zone is such that it presents a pattern, for example a shape, a line or a curve, regularly repeated in a given direction
• Streak detection methods are known in images in which a spectral approach is used to detect streak frequencies.
• Fourier filtering is generally used.
• the Fourier transform of the image to be studied is carried out.
• the image obtained which represents the frequencies of the image studied in the Fourier space
• peaks are obtained which correspond to different frequencies of gray levels or colors of the image.
• This approach is expensive in computing time, especially on large images whose Fourier transform must be calculated. It is moreover rather delicate and imprecise, in particular because it is complicated to separate the frequencies corresponding to streaks frequencies corresponding to background noise. Indeed, the frequency peaks corresponding to streaks are rarely clearly defined in the image obtained by the Fourier transform.
• Another type of streak detection method in an image consists of comparing the studied image with so-called reference images comprising streaks, and calculating a correlation rate between the images.
• the major disadvantage of this type of method is that it requires a very large memory to hold the reference images and a very important computation time to compare the image portions between them and determine a correlation rate.
• An object of the invention is to provide a method that is less expensive in computing time and in memory, and is simpler, more accurate and more reliable than the aforementioned methods.
• determining at least one representation comprising a type of streaks to be referenced identifying at least one segment of pixels or voxels of the representation
• the or each recorded value is then used as a control to which one or more values measured on a studied image are compared, in order to determine whether the image studied comprises streaks or not.
• the segmental approach makes it possible to reference values which relate to a series of successive pixels, instead of studying the image pixel by pixel for example. This approach is particularly well suited when the segment is perpendicular to streaks, the values relating to differences between the gray levels or colors then highlighting the variation of pixel levels. This method is independent of the type of streak detection method used later. The same observations are worth considering voxels rather than pixels. The same goes for the rest of the time, each time we consider pixels.
• the digital representations on which are implemented a method according to the invention can be of three types:
• 3D - Representations
• each pixel of the image carries topographic information of depth of the streaks.
• the level of gray or color corresponds to this depth of the streaks.
• the level corresponds, for example, to the contrast between the background and the top of the streaks.
• the value or at least one of the values is chosen from the following group:
• the average serves as a control to determine if a segment of pixels of a studied image is in a potentially streaked zone or not. Indeed, the more streaks there are, the more gray level pixels are next to each other, and the higher the average difference in levels between adjacent pixels.
• the average period makes it possible to establish an average interval of distance between the vertices of the streaks or between the valleys of the inter-striations.
• the length makes it possible to compare segments large enough to contain relevant information about the streak area and small enough that the calculation of the other reference value (s) is not too time-consuming.
• the length of the segment is advantageously fixed at 3.5 mean periods.
• identifying a segment of pixels or voxels centered on the pixel or voxel considered determining at least one value relating to differences between gray levels or colors of pixels or voxels of the segment
• the segmental approach makes it possible to determine values that relate to a sequence of successive pixels, instead of studying the image pixel by pixel.
• This approach is therefore particularly well suited when the segment to be studied is perpendicular to streaks, the values relating to differences between the gray or color levels then highlighting the variation of pixel levels.
• the segment centered on this pixel comprises values relating to differences in levels of pixels or voxels already calculated previously. It is thus possible to calculate values relating to differences within several rapidly overlapping segments by factoring the calculations.
• the value or at least one of the difference values is an average of absolute values of differences in the levels within each pair of adjacent pixels or voxels of the segment.
• the average of the absolute values of the differences makes it possible to determine whether the segment is in a strongly streaked zone or not. This value is simple and quick to determine, and we remove the pixels that are not part of a streaked area. For example, if the average difference is small, it means that the segment, and therefore the pixel located at its center, is in a rather homogeneous area in terms of gray levels or colors. On the other hand, if the average difference is high, it means that by traversing the segment of a pixel located at one end to the pixel located at the other end, the levels of gray or colors vary considerably. In this case, it is considered that the pixel located in the center of this segment may be included in a streak zone, and it is not eliminated.
• the value or at least one of the difference values is an average period of periods relating to the pixels or voxels of the segment.
• the average period represents the average distance between, within the segment, two identical changes of values of adjacent pixels. This corresponds to the average interval between two peaks of streaks of the segment, or between two inter-strikeths.
• the value or at least one of the values relating to the differences is a number of periods relating to pixels or voxels of the segment.
• the value or at least one of the values relating to the differences is one or more periods relating to pixels or voxels of the segment.
• each period for example an interval between two peaks of streaks, can be compared to a predetermined value.
• the automated means associate a binary value "0" or
• each pixel or voxel of the segment depending on its level, and the value or at least one of the difference values is relative to changes between values within adjacent pairs of pixels or voxels when the automated means traverse the segment in one direction, these changes being identical and preferably the first pixel of each pair comprising a value identical to that of a pixel or a voxel of the segment located at an end of the predetermined segment as a function of the direction.
• the segment is binarized so as to distinguish only two types of pixels: those located in a strip, those located between two stripes. All calculations of values relating to differences between gray levels or colors of pixels or voxels of the segment, and in particular the calculation of periods, are then simplified.
• an average value of the values of the pixels or voxels of the segment is determined
• each pixel or voxel of the segment is assigned a binary value as a function of the difference between its value and the average numerical value.
• the value of the pixel or voxel is greater than or equal to the average value, it is associated with the binary value "1", otherwise "0".
• the automated means associates a binary value "0" or "1" with each pixel or voxel of the segment as a function of its level, and the value or at least one of the values relating to the differences is relative to changes between values within pairs of adjacent pixels or voxels when the automated means traverses the segment in one direction, these changes being identical and preferably the first pixel of each pair comprising a preference value different from that of a pixel or a voxel of the segment located at one end of the predetermined segment depending on the direction.
• the secondary periods represent either an interval between two inter-striketholes, or an interval between two vertices of streaks, but not both, in the same way as the values defined above and complementary to them.
• the secondary periods concern the intervals between inter-striketholes, and vice versa.
• the segment is composed of at least three consecutive pixels or voxels of the representation.
• the dilation allows to obtain in the dilated image, in place of the striation zone of the base image, a smooth streak zone, the intervals between the streaks having been erased by the expansion.
• the erosion makes it possible to obtain in the eroded image, instead of the striation zone of the base image, a zone of intervals of smooth striations, the streaks being erased by erosion.
• the difference representation must comprise a perfectly homogeneous zone at the same place as the position of the streak zone in the basic representation.
• the difference representation includes an area corresponding to the streak area, whose gray or color levels are substantially constant, within a noise tolerance range.
• the difference representation comprises one or more zones whose gray or color levels are very different from those of the surrounding pixels.
• an aim is to provide a method for analyzing the conformity of tire grooves which is less expensive in terms of calculation time and faster to implement.
• the automated means determine one or more structuring elements of the expansion and erosion as a function of a dimension of the streaks, a gap between the striations and / or a striations orientation.
• the structuring elements are adapted to each type of streaks detected upstream. This makes it possible to carry out the most relevant expansion and erosion operations possible, by deleting the streaks and the intervals between the striations in the most correct manner, while preserving the other elements.
• the automated means determine at least two dilations of the basic representation with different respective structuring elements, to obtain the expanded representation.
• the automated means determines at least two erosions of the basic representation with respective different structuring elements, to obtain the eroded representation.
• the striations are separated into different types of streaks, and the dilation and / or erosion operations are repeated. for each type of streak, with structuring elements adapted to the different types of streaks of the zone.
• digital values of pixels or voxels of the difference representation are compared with at least one predetermined threshold.
• the streaks of the basic representation comprise a defect.
• all the values are within a predetermined interval relative to the predetermined threshold, it is considered that the striation zone of the basic representation does not include a defect, and that the tire therefore conforms.
• the threshold or at least one of the thresholds is a median of values of the pixels or voxels of the difference representation.
• the basic representation does not include any color other than black, white and gray levels.
• the basic representation can also include black, white and gray.
• automated means perform the following steps:
• the segmental approach makes it possible to determine values that relate to a sequence of successive pixels, instead of studying the image pixel by pixel.
• This approach is therefore particularly well suited when the segment to be studied is perpendicular to streaks, the values relating to differences between the gray or color levels then highlighting the variation of pixel levels. It is also not necessary to try to detect peaks of frequencies in an image transformed by complex calculations, for example by Fourier analysis, since one is directly interested in the image to be studied in order to detect the streaks.
• the segment centered on this pixel comprises values relating to differences in levels of pixels or voxels already calculated previously. It is thus possible to calculate values relating to differences within several rapidly overlapping segments by factoring the calculations.
• the or each recorded value is then used as a control to which one or more values measured on a studied image are compared, in order to determine whether the image studied comprises streaks or not.
• the segmental approach makes it possible to reference values which relate to a series of successive pixels, instead of studying the image pixel by pixel for example. This approach is particularly well suited when the segment is perpendicular to streaks, the values relating to differences between the gray levels or colors then highlighting the variation of pixel levels. This method is independent of the type of streak detection method used later.
• the device comprises a recording medium comprising a database of values relating to streaks.
• FIGS. 3 to 5 schematically illustrate respectively a digital image, a segment of this image, and the segment in binarized form
• FIG. 6 illustrates a method according to one embodiment of the invention
• FIGS. 7 to 11 schematically illustrate respectively an image, a segment of the image, the segment in binarized form, another segment of the image and this segment in binarized form;
• FIG. 12 illustrates a method according to another embodiment of the invention.
• FIGS. 13 to 16 schematically illustrate a digital image, the image in eroded form, the image in dilated form, and a difference image between the dilated and eroded images;
• FIGS. 17 and 18 respectively illustrate a digital image comprising a zone of streaks having a defect and the difference image resulting from this image according to one embodiment of the invention.
• FIG. 19 illustrates a device adapted to implement a method according to the invention.
• the tire control method aims at creating a tire image database to reference types of streaks and then to detect streaks similar to the types of streaks referenced in test images.
• the streak conformance check method aims to check whether tire striation areas have defects. I SEO process
• the method first consists of referencing streak types and then detecting streaks in images using the referenced streaks.
• Figures 1 and 2 illustrate different types of streaks in two-dimensional images 10 and 20. These types of striations differ from each other in the thickness of the striations, their orientation, their straightness, the interval between the striations, as well as the gray levels of the striations and the intervals of striations of each type.
• Figure 1 also proposes two zones 1 and 2 of different types of streaks. These are all types of streaks that we want to reference at first, and detect in a second time when we find these streaks in an image.
• a device 90 which notably comprises a processor 94, a memory 95 and is connected to a database 92.
• a computer program can request as input an image or a series of images including zones of streaks to be referenced, as well as an image or images to be studied.
• This program provides the user with data on each type of reference streaks, as well as the types of streaks determined and their location in the images to be studied.
• the same or a separate program allows for the application of a compliance check method described below.
• the input image may be provided automatically by the method itself when it has detected streaks in an image.
• the same program makes it possible to determine streaks in an image of a tire, and at the same time to determine whether or not these streaks have defects.
• this program can be made available on a telecommunications network, such as the web, or an internal network, so as to allow its download by a user.
• program or equivalent instructions may be recorded on a computer-readable storage medium 93 such as a hard disk, a USB key, a CD, or any other equivalent medium, which may include the database.
• a computer-readable storage medium 93 such as a hard disk, a USB key, a CD, or any other equivalent medium, which may include the database.
• reference images comprising streak zones such as images 10 and 20 of FIGS. 1 and 2 are selected to construct a reference base.
• This reference base may include any image comprising a striated zone, even not explicitly described in the present application.
• each pixel of the image 30, and a fortiori of the reference segment 4 includes a gray level value.
• a reference segment 4 of 21 pixels is selected.
• the pixel values are binarized according to the average of the previously calculated gray levels. Thus, if a gray value of one pixel equal to or exceeds the average of the gray levels of the reference segment, the corresponding pixel is assigned the value "0". If a gray value is below average, the corresponding pixel is assigned the value "1".
• a segment 50 is thus obtained, illustrated in FIG. 5. It is from this segment 50 that the calculations described below are carried out:
• a main period is the shortest distance, in number of pixels, between two changes between values within adjacent pairs of pixels when traversing the segment from left to right, these changes being identical and the first pixel of each pair comprising a value identical to that of the first pixel of the segment at the left end.
• the first pixel 14 located at the left end has the binary value "1”.
• the first change is therefore sought between a binary value pixel "1" and a binary value pixel "0". This change takes place between the pixels 15 and 16.
• the second identical change is then sought, that is to say between a pixel of binary value "1” and a pixel of binary value "0", by traversing the segment of left and right.
• the type of streaks of FIG. 3 was entered as a reference in the database.
• the three data entered for this type of streaks namely the reference average, the average reference period, and the reference length, must make it possible to detect this type of streaks in any image to be studied, if these streaks are present.
• each type of streak referenced can be compared to the values that will be determined during the detection.
• FIG. 6 which illustrates a method according to a preferred embodiment of the invention, the following steps are performed for a given type of streaks:
• step B If the average calculated on the study segment is in the range, then the segment is subjected to step B). If the result is not included in the interval, then another type of referenced streaks to which the study segment 62 is to be compared is compared, and step A) is repeated for the new referenced streak type considered. . This amounts to using a high threshold and a low threshold on both sides of the reference average and to compare the result with these thresholds.
• This criterion removes a large majority of bad pixels, leaving only the pixels on areas a minimum textured, but not necessarily still streaking.
• step B) The segment is binarized in the same way as in step 4) of the referencing method, and the same calculations are performed as in steps 5) and 6).
• step 4) we obtain an average period of the main periods on the study segment 62, represented schematically by the binarized study segment 63 of FIG. 9.
• This average period is then compared with the reference period recorded for the type of streaks considered successfully in step A). In the same way as before, this comparison is carried out by a range of values centered on an "average reference period". If the average period of the segment 63 belongs to the range of values, then the pixel 61 and its binarized study segment 63 go to step C), otherwise another type of streaks is selected and the method is repeated. step A), for the new type of streaks considered.
• This criterion removes areas that do not look like the type of streaks searched, ie the types of streaks referenced.
• This criterion mainly removes some bad areas near the edges of the texts, but also the lateral borders of zones of striations that could have been recognized and localized. These areas can be found later by binary morphology steps such as dilations or erosions.
• step D If the study segment passes this step, proceed to step D). Otherwise, step A) is repeated with a new type of reference streaks.
• the pixel 61 on which the study segment 62 is centered is considered to belong to the type of streaks for which steps A) to D) were performed.
• step A) the process is repeated in step A) with a new type of reference streaks.
• step D If all types of reference streaks have been compared to the segment without the pixel successfully passing step D), it is considered that the pixel 61 of the study segment 62 does not belong to streak types that have been referenced, and we stop the process. It can be repeated in step A) with another pixel on which we will center another study segment.
• step D the process is restarted with another pixel, for the same type of referenced streak.
• the method stops when all the pixels of the image have been considered, that is to say when all these pixels are passed at least through step A of the method.
• only a certain portion of the image, or some pixels of the image, is selected, and the method is applied only to these pixels.
• a user may have visually located an area that might contain streaks in an image, and decide to apply the process only to that area of the image.
• step A certain differences between pixels are recorded. Indeed, if we first perform calculations for a given pixel, then calculations for a pixel located on the same pixel line of the image, it is possible that their study segments include identical pixels. It is then useful to reuse in the calculation already calculated results.
• step A) is independent of the other three steps because there is no need to binarize the segments to achieve it. This step is actually the simplest of all, that's why it's done first.
• step A instead of restarting the process in step A) with another type of streaks, the test of the same step is carried out with another type of streaks.
• a pixel can pass the test of step A) for a given type of streaks, pass step B) for another type of streaks given, and so on.
• the ranges of values that serve as a comparison are also stored in the database. They are not necessarily centered on the reference values such as the average reference period, the reference length or the reference average. They can understand these values without being centered on them. In this way, some variations from the reference values are tolerated in one direction but not in another.
• only one type of streaks or several types of particular streaks in the study segments are to be detected. This consists of comparing the data of the study segment with the referenced data concerning these types of striations and not with the other types of referenced streaks.
• a fault In a flat area, a fault is characterized by a higher or lower elevation than the average of the area.
• the principle of the method in the present mode consists of filter the streaks in two different ways to obtain two images of flat areas: an image representing an average of the streaks, and an image representing an average of the vertices of the streaks.
• an erosion of the image 70 is performed, so as to obtain an eroded image.
• the structuring element is chosen so that the streaks disappear in the eroded image.
• the structuring element one takes into account the interval between the striations, the orientation of the striations, and their size. Erosion takes place in gray level, one can also take into account the gray levels of the streaks and intervals.
• the eroded image 71 of Figure 14 thus represents an average of the bottoms of the streaks, in other words an average of the hollows between the streaks.
• An expansion of the image 70 is also carried out.
• the structural element of the expansion is chosen to expand the striations so that they fill the gaps between them, on the expanded image 72 of FIG. to choose the structuring element of the dilation are the same as those of erosion.
• the dilated image 72 thus represents an average of the vertices of the streaks.
• a difference is then made between the expanded image 72 and the eroded image 71 so as to obtain a difference image 73.
• This image here has a homogeneous content. There is therefore no defect in the streak area of the image 70.
• the method according to the invention makes it possible to automatically detect these defects, by comparing the value of the gray levels of the pixels with the median value of the pixels of the difference image. Thus, if the value of one of the pixels of the difference image is too far from the median value of the pixels of the image, it is considered that the pixel in question makes visible a defect in the streak zone of the image.
• the images include colors other than shades of gray.
• the above calculations, concerning the streak detection method, and the conformity control method can in particular be carried out on each type of color independently of each other, so as to detect and / or control, for example, streaks of red, of green, of blue.
• the calculations can also relate to values resulting from combinations between these color values.
• the images form non-two-dimensional but three-dimensional spaces including voxels.
• each voxel includes a luminance value. Previous calculations can therefore be further performed on the depth levels. Thus, even with identical or similar colors, it is possible to reference, determine, and / or control streaks which are distinguished from each other by their relief.
• the method for detecting the streak zones described in Part II and the streak compliance checking method described in Part III may be implemented independently of one another. In particular, it is possible to control the conformity of the streaks according to the method of Part III after detecting a zone of striations other than according to the method of Part II, and vice versa.

Landscapes

• Engineering & Computer Science (AREA)
• Theoretical Computer Science (AREA)
• Physics & Mathematics (AREA)
• General Physics & Mathematics (AREA)
• Computer Vision & Pattern Recognition (AREA)
• Quality & Reliability (AREA)
• Databases & Information Systems (AREA)
• Data Mining & Analysis (AREA)
• General Engineering & Computer Science (AREA)
• Library & Information Science (AREA)
• Software Systems (AREA)
• Image Analysis (AREA)
• Length Measuring Devices By Optical Means (AREA)
• Tires In General (AREA)
• Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
• Image Generation (AREA)

Applications Claiming Priority (3)

Publications (1)

Family

ID=55077542

Family Applications (1)

Country Status (5)

Families Citing this family (3)

Family Cites Families (12)

• 2015
  • 2015-12-16 CN CN201580069529.1A patent/CN107111873A/zh active Pending
  • 2015-12-16 US US15/534,177 patent/US20170330338A1/en not_active Abandoned
  • 2015-12-16 EP EP15821143.3A patent/EP3234915A2/de not_active Withdrawn
  • 2015-12-16 JP JP2017532109A patent/JP2018501480A/ja active Pending
  • 2015-12-16 WO PCT/FR2015/053538 patent/WO2016097598A2/fr active Application Filing

Non-Patent Citations (2)

Also Published As

Similar Documents

Legal Events