JP4846113B2 - Image processing method, image processing apparatus, and computer-readable storage medium - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image processing method, an image processing apparatus, and a computer-readable storage medium, and is particularly suitable for use in detecting surface defects based on an image signal obtained by photographing a surface of a steel plate.
[0002]
[Prior art]
Conventionally, using a fixed threshold as a reference for detecting wrinkles on a steel plate, only wrinkles having a size of a certain area or more are selected by a filter with respect to all of those recognized as wrinkle candidates. And after selecting, the characteristic amount was given, and based on the characteristic amount, the grape type and grade were given to each bag.
[0003]
[Problems to be solved by the invention]
However, in the conventional case, only those with a certain area or more were selected against those recognized as candidates, so there is a possibility of missing a defect that is a small but large collection of defects. was there.
[0004]
In addition, when the eyelid inspection apparatus recognizes a large eyelid by dividing it, each of the recognized eyelids becomes a small eyelid, so the score becomes small, and as a result, it may be erroneously determined as a harmless eyelid. there were.
Furthermore, since all feature values were calculated for those that were recognized as candidates for defects, this calculation would put a heavy load on the computer and could sacrifice other important processing. was there.
[0005]
For example, there is a case in which the above-described inconvenience is caused by recognizing the large ridge 50 shown in FIG. 5 by dividing it into five small ridges 51 to 55.
In view of the above-described problems, an object of the present invention is to make it possible to accurately perform wrinkle determination without imposing a great load on a computer.
[0006]
[Means for Solving the Problems]
The image processing method of the present invention is an image processing method for generating an image signal by imaging an object to be detected by an imaging unit, and performing image processing for detecting the surface defect of the object based on the image signal. In addition, a wrinkle ranking process for ranking each of a plurality of wrinkles present in one frame in the subject image generated by the imaging means, and a predetermined number from the top of each wrinkle ranked by the wrinkle ranking process Based on the upper cocoon sorting process for selecting several cocoons, the feature amount giving process for assigning a feature amount to each cocoon selected by the upper heel sorting process, and the feature amount given by the feature amount giving process A group selection process for selecting each basket into a plurality of groups, and a group for each of the groups selected by the group selection process to combine the respective font candidates into a single font candidate And if the processing for each flaw candidate integrated by the flaws group integrating process, and a determining flaw determination process of the type and grade of flaws, the flaws group integrating processing, flaw detection region for the flaw detection When an area consisting of one mesh is set, the number of wrinkles is detected for each of the set wrinkle detection areas, and the number of wrinkles in the detected wrinkle detection area is smaller than a predetermined wrinkle count threshold The wrinkles in the wrinkle detection area are excluded from the target of the wrinkle integration process, and the wrinkle detection area having a wrinkle that is not excluded by the wrinkle integration exclusion process and is the target of the wrinkle integration process is halved in the vertical direction and the left-right direction. The meshes are shifted to detect whether or not there is an integration target wrinkle that is equal to or greater than the predetermined wrinkle count threshold in each adjacent region. Above Integrating with the cocoon existing in the cocoon detection area having the cocoon to be processed is regarded as one cocoon, and the above adjacent area where the cocoon to be integrated is detected is used as a new cocoon to be subjected to the integration process. As a wrinkle detection area having, wrinkle integration processing is performed in which processing for integrating wrinkles is repeated. Another feature of the present invention is that the wrinkle ranking process is based on the wrinkle area, the absolute value of the difference between the maximum luminance value and the reference value, and the absolute value of the difference between the minimum luminance value and the reference value. It is characterized by giving a score.
[0007]
The image processing apparatus of the present invention is an image processing apparatus that generates an image signal by capturing an image of a subject to be detected by an imaging unit, and performs image processing for detecting a surface defect of the subject based on the image signal. The heel ranking means for ranking each of the plurality of wings present in one frame in the subject image generated by the imaging means, and a predetermined number from the top of each heel ranked by the heel ranking means Based on the upper cocoon sorting means for selecting several cocoons, the feature quantity giving means for giving the feature quantity to each cocoon sorted by the upper cocoon sorting means, and the feature quantity given by the feature quantity giving means A group selection means for selecting each bag into a plurality of groups, and a bag group that integrates each of the bag candidates for each group selected by the group selection unit to form one bag candidate And covering means, for each flaw candidate integrated by the flaw group integrating means, and a flaw determining means for determining the type and grade of flaws, the flaws group integrating means, flaw detection region for the flaw detection An area consisting of one mesh is set, and the number of wrinkles is detected for each of the set wrinkle detection areas. If the detected number of wrinkles is smaller than a predetermined wrinkle number threshold, the wrinkle detection area The wrinkle detection area having a wrinkle that is the target of the wrinkle integration process without being excluded by the wrinkle integration exclusion process is shifted by 1/2 mesh in the vertical direction and the left-right direction. It is detected whether or not there is an integration target wrinkle that is equal to or greater than the predetermined wrinkle count threshold in each adjacent region. Subject and A wrinkle detection region having a wrinkle that becomes a target of a new wrinkle integration process is considered as one wrinkle by integrating with a wrinkle that exists in a wrinkle detection region having a wrinkle It is characterized by comprising an integration means for performing a cocoon integration process that repeats the process of integrating the cocoons. Another feature of the present invention is that the heel ranking means is based on the heel area, the absolute value of the difference between the maximum luminance value and the reference value, and the absolute value of the difference between the minimum luminance value and the reference value. It is characterized by giving a score.
[0008]
The storage medium of the present invention is characterized by recording a program for causing a computer to execute any of the image processing methods described above.
Another feature of the present invention is that a program constituting each of the above-described means is recorded.
[0009]
DETAILED DESCRIPTION OF THE INVENTION
Next, embodiments of an image processing apparatus, an image processing method, and a computer-readable storage medium according to the present invention will be described with reference to the accompanying drawings.
FIG. 1 is a flowchart illustrating an example of main discrimination processing performed in the image processing apparatus in order to perform the eyelid inspection according to the present embodiment.
[0010]
As shown in the flowchart of FIG. 1, when the processing is started in the first step S11, in step S12, each of the plurality of eyelids present in one frame in the subject image generated by the imaging means is ranked. Perform the ranking process.
[0011]
In the present embodiment, the heel ranking process is scored based on the heel area, the absolute value of the difference between the maximum luminance value and the reference value, and the absolute value of the difference between the minimum luminance value and the reference value. To do.
[0012]
Next, in step S13, an upper eyelid sorting process is performed in which a predetermined number of eyelids are selected from the upper ranks in each eyelet ranked by the eyelid ranking process. In this higher-order cocoon sorting process, for example, the top 128 cocoons are sorted.
[0013]
Next, the process proceeds to step S14, and a feature amount providing process is performed for assigning a feature amount to each bag selected in step S13. Conventionally, feature amounts are assigned to all detected wrinkles, but in this embodiment, feature amounts are given only to selected wrinkles so as to reduce the load on the computer. Yes.
[0014]
Next, the process proceeds to step S15, and a group selection process is performed in which each bag is selected into a plurality of groups based on the feature value given by the feature value giving process in step S14. This group selection process is a process of grouping for each identical species.
[0015]
Next, the process proceeds to step S16, and a cocoon group integration process is performed by integrating the cocoons in a predetermined unit area for each group selected by the group selection process in step S15 to form one cocoon candidate. Thereby, large and small cocoons existing in the unit area are integrated as one cocoon candidate.
[0016]
Next, in step S17, a wrinkle determination process for determining the type and grade of the wrinkles is performed for each wrinkle candidate integrated by the wrinkle group integration process in step S16.
[0017]
Next, an example of the integration process for each basket type performed in step S16 will be described with reference to the flowchart of FIG.
After branching from the main routine in step S21, first, mesh ticks are set and input as a wrinkle detection region in which wrinkle detection is performed in step S22. As shown in step 1 of FIG. 4, this is a process of designating an area having (number of vertical 4 × horizontal 4) = 16 pixels as a unit area “1 mesh”.
[0018]
Next, the process proceeds to step S23, and a wrinkle number threshold value input process for inputting the number of wrinkles detected in the wrinkle detection region set by the unit area input process is performed. The number of wrinkles to be detected is, for example, about “3” per unit area. This is because, when detecting a wrinkle existing in “1 mesh” M1 indicated by hatching in step 1 of FIG. 4, if there are more than one wrinkles, “there is a wrinkle in one mesh”. ”Is a process of inputting a numerical value that decides to be determined.
[0019]
Next, proceeding to step S24, a wrinkle number detection process for detecting the number of wrinkles present in the wrinkle detection region set by the unit area input process is performed. An example is shown in which there are four wrinkles in the “1 mesh” M1 shown in step 1 of FIG.
[0020]
Next, proceeding to step S25, when the number of wrinkles in the wrinkle detection area detected by the wrinkle number detection process is smaller than the wrinkle number threshold value input by the wrinkle number threshold value input process,疵 Integrated exclusion process to exclude 疵 from the target of 疵 integration process.
[0021]
In the case of step 1 in FIG. 4, since there are four wrinkles in the “1 mesh” M1, it is not a target of the wrinkle integration exclusion process. Therefore, in this case, the process proceeds to step S26, and it is detected whether or not there is a wrinkle to be integrated in an area adjacent to the wrinkle detection area.
[0022]
In the present embodiment, the mesh (detection region) having the wrinkles to be integrated is shifted in the vertical and horizontal directions (1/2 mesh), and the number of wrinkles existing there is counted. If there are three or more group wrinkles, the process of integrating the areas and considering it as one wrinkle is executed.
[0023]
The example of FIG. 4 shows a case where Step 2 is shifted to the left side in FIG. 4 (1/2 mesh) to be mesh M2. There are five bags to be integrated, so the areas are integrated. Then, as shown in step 3, the integration process is further performed as M3 after shifting (1/2 mesh) to the left in FIG.
[0024]
Step 4 in FIG. 4 is an example in which M4 is shifted upward (1/2 mesh) from the region of Step 3, and Step 5 in FIG. 4 is downward from the region of Step 3 (1/2 This is an example of shifting to M5. As a result of performing the integration process as described above, in the present embodiment, as shown in step 6 of FIG. 4, a large bag “K1” can be detected.
[0025]
Next, in step S27 in FIG. 2, a process is performed in which the cocoons in the mesh that have cocoons greater than or equal to the cocoon number threshold and are not integrated in the process in step S26 are integrated and regarded as one cocoon. As a result of this processing, it can be detected as a small wrinkle indicated as “K2” in step 6 in FIG.
[0026]
Next, in step S28, a process is performed in which a bag that has not been subjected to the integration process described above is treated as an isolated bottle as it is. As a result of performing such integration processing, as shown in FIG. 3, it is possible to detect a large kite “K1” and a small kite “K2”.
[0027]
FIG. 6 is a diagram illustrating an internal configuration of a general personal user terminal device. In FIG. 6, reference numeral 1200 denotes a computer PC. The PC 1200 includes a CPU 1201 and executes network printing device control software stored in the ROM 1202 or the hard disk (HD) 1211 or supplied from the floppy disk drive (FD) 1212 and connected to the system bus 1204. Control overall.
[0028]
According to the program stored in the CPU 1201, the ROM 1202 or the hard disk (HD) 1211 of the PC 1200, the wrinkle ranking means, the higher-order wrinkle sorting means, the feature amount giving means, the group sorting means, the wrinkle group integration means, and the wrinkle determination means of this embodiment Each means includes an integration unit, a basket number threshold value input unit, a basket number detection unit, and a basket integration exclusion unit.
[0029]
A RAM 1203 functions as a main memory, work area, and the like for the CPU 1201. A keyboard controller (KBC) 1205 controls instruction input from a keyboard (KB) 1209, a pointing printing device (not shown), or the like.
[0030]
Reference numeral 1206 denotes a CRT controller (CRTC) which controls display on a CRT display (CRT) 1210. A disk controller (DKC) 1207 is a hard disk (boot program (start program: a program that starts execution (operation) of personal computer hardware and software)), a plurality of applications, editing files, user files, a network management program, and the like. HD) 1211 and floppy disk (FD) 1212 are controlled.
[0031]
Reference numeral 1208 denotes a network interface card (NIC) that exchanges data bidirectionally with a network printer, another network device, or another PC via the LAN 1220.
[0032]
【The invention's effect】
As described above, according to the present invention, according to the present invention, a plurality of wrinkles existing in one frame in a subject image are ranked, and a predetermined number of wrinkles are selected from the top of the ranked wrinkles. And assigning a feature amount to each of them, sorting each cocoon into a plurality of groups based on the feature amount, and integrating each cocoon candidate for each of the selected groups to form one cocoon candidate. On the other hand, since the type and grade of the cocoon are determined, the cocoon determination can be performed accurately.
Thereby, since the area is small, the inconvenience of being erroneously determined as a harmless flaw can be prevented, and the possibility of missing a flaw that is a serious defect can be reduced.
[0033]
Further, since the feature amount is given only to the selected candy candidates, the load on the computer can be reduced, and the inconvenience of sacrificing other important processing can be greatly reduced. .
[Brief description of the drawings]
FIG. 1 is a flowchart illustrating a procedure of main discrimination processing performed in a wrinkle inspection apparatus according to an embodiment of the present invention.
FIG. 2 is a flowchart illustrating an example of a procedure of integration processing performed for each type of straw.
FIG. 3 is a diagram illustrating an example of a result of performing a bag integration process.
FIG. 4 is a diagram for explaining an example of a procedure of cocoon integration processing.
FIG. 5 is a diagram illustrating an example of a case where it is determined as a harmless moth when it is recognized by being divided into small moths.
FIG. 6 is a block diagram illustrating an example of a computer system capable of executing the image processing method according to the embodiment.
[Explanation of symbols]
1200 Computer 1201 CPU
1202 ROM
1203 RAM
1204 System bus 1205 Keyboard controller (KBC)
1206 CRT Controller (CRTC)
1207 Disk controller (DKC)
1208 Network Interface Card (NIC)
1209 Keyboard 1210 CRT display (CRT)
1211 Hard disk (HD)
1212 Floppy disk (FD)
1220 LAN

Claims (6)

An image processing method for generating an image signal by imaging an object to be detected by an imaging unit, and performing image processing for detecting a surface defect of the subject based on the image signal, the image processing method generating the image signal A heel ranking process for ranking each of a plurality of cocoons present in one frame in the subject image, and a higher rank for selecting a predetermined number of folds from the higher ranks in each of the cocoons ranked by the above heel ranking process A selection process for selecting a plurality of baskets based on the feature amount given by the feature selection process, the feature addition process for assigning a feature amount to each selection made by the upper selection process, and the feature addition process Group selection processing, group selection processing for grouping the groups selected by the group selection process, integrating group selection candidates into a group selection candidate, and group selection processing For each defect candidate integrated by the integration processing, and a determining flaw determination process of the type and grade of flaws,
The wrinkle group integration process is a wrinkle detection area for performing wrinkle detection , and sets an area consisting of one mesh and detects the number of wrinkles for each of the set wrinkle detection areas. When the number of cocoons is smaller than a predetermined cocoon number threshold, the cocoons in the cocoon detection area are excluded from the cocoon integration processing target,
The wrinkle detection area having a wrinkle that is not subject to the wrinkle integration exclusion process is shifted by ½ mesh in the vertical direction and the left-right direction, and is adjacent to each of the adjacent areas by the above threshold value. Whether or not there is an integration target wrinkle, and if the detected wrinkles are integration target wrinkles, they are included in the wrinkle detection area having the wrinkles that are the targets of the above-described integration processing.疵 integration processing that repeats the process of integrating 疵 as a 疵 detection region having a 疵 that is the target of new 疵 integration processing And an image processing method.   The heel ranking processing is performed by assigning a score based on the heel area, the absolute value of the difference between the maximum luminance value and the reference value, and the absolute value of the difference between the minimum luminance value and the reference value. 2. The image processing method according to 1. An image processing apparatus that captures an image of a subject to be detected by an imaging unit, generates an image signal, and performs image processing for detecting a surface defect of the subject based on the image signal. A ranking means for ranking each of a plurality of wrinkles present in one frame in the subject image, and a higher rank for selecting a predetermined number of wrinkles from the higher ranks in each of the wrinkles ranked by the wrinkle ranking means A cocoon sorting unit, a feature amount assigning unit for assigning a feature amount to each cocoon selected by the upper heel sorter, and a selection of each cocoon into a plurality of groups based on the feature amount assigned by the feature amount assigning unit A group selection unit that performs selection, a group selection unit that integrates the individual font candidates for each group selected by the group selection unit, and sets the single group as a single candidate. For each defect candidate integrated by the integration means, and a flaw determining means for determining the type and grade of flaws,
The wrinkle group integration means is a wrinkle detection region for performing wrinkle detection, sets an area consisting of one mesh , detects the number of wrinkles for each of the set wrinkle detection areas, and the detected number of wrinkles is a predetermined value. If it is less than the threshold value for the number of wrinkles, the wrinkles in the wrinkle detection area are excluded from the target of the wrinkle integration process,
The wrinkle detection area having a wrinkle that is not subject to the wrinkle integration exclusion process is shifted by ½ mesh in the vertical direction and the left-right direction, and is adjacent to each of the adjacent areas by the above threshold value. Whether or not there is an integration target wrinkle, and if the detected wrinkles are integration target wrinkles, they are included in the wrinkle detection area having the wrinkles that are the targets of the above-described integration processing.疵 integration processing that repeats the process of integrating 疵 as a 疵 detection region having a 疵 that is the target of new 疵 integration processing An image processing apparatus comprising an integration unit for performing the processing.   The score ranking means performs scoring based on the score area, the absolute value of the difference between the maximum brightness value and the reference value, and the absolute value of the difference between the minimum brightness value and the reference value. The image processing apparatus according to 3.   3. A computer-readable storage medium having recorded thereon a program for causing a computer to execute the image processing method according to claim 1 or 2.   5. A computer-readable storage medium in which a program for causing a computer to function as each means of the image processing apparatus according to claim 3 is recorded.

Images