JPH063609B2 - Image processing method - Google Patents

Description

Description: FIELD OF THE INVENTION The present invention relates to an image processing system for advancing processing while sequentially cutting out a target image for each image area.

[Background of the Invention]

It is known that a target part is searched for in a picked-up image, its size, position, and the like are measured, and those data are used for controlling an industrial robot or the like that handles the part. The image includes target parts, other parts, minute parts in the background, and also noise, scratches, stains, and other elements. Therefore, when the background is complicated and there are many noises and the like, the number of apparent parts recognized in the image increases correspondingly, and it becomes necessary to set the capacity of the image processing storage means large in advance.

Further, since the processing time becomes long due to the apparent increase in the number of parts due to noise, it is important that the processing time is short in the graphic processing apparatus used for the above purpose.

As a conventional example for cutting out an image using a storage means for image processing and removing noise, there are methods as disclosed in JP-A-58-207181 and JP-A-49-149633. No consideration has been given to the lack of capacity and the inability to continue the cropping process for the next image area.

[Object of the Invention]

An object of the present invention is to effectively avoid the shortage of the capacity of the storage means that occurs during image processing and to shorten the processing time.

[Outline of Invention]

In the present invention, there are provided means for sequentially cutting out image areas to be processed and storage means for storing data relating to the cut out image areas, for example, analysis data. Further, it has means for judging the cut-out end of each part area included in the already cut-out image area. The parts area is the area on the screen occupied by one part that appears in the image. If the part area does not reach the image area to be subsequently cut out, the cutout for the part area is completed even if the image area is not completely cut out. On the other hand, it has a means for determining the necessity of data regarding the part area for which the cutting has been completed. For example, if the area of the component region is smaller than a predetermined reference value, the region is determined to be noise or the like and is determined as unnecessary data. The unnecessary data includes data indicating the existence of the parts area, for example, the parts area number. Further, the present invention further comprises means for erasing the unnecessary data judged by the means from the storage means. Therefore, even if the capacity of the storage means is insufficient due to the incomplete cutting of the image area, the unnecessary data can be deleted from it.
The cutting process of the next image area can be continued.

Example of Invention

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 shows an image processing circuit. This figure mainly shows the flow of data, and the CPU necessary for processing or its program storage means is not shown. An image pickup device 1 picks up an image of an object to be recognized. An image data input device 2 receives an analog signal transmitted by the image pickup device 1, converts the analog signal into digital multi-valued image data, and stores the multi-valued image memory shown in 3. Reference numeral 4 denotes a comparator having a function of binarizing the image data stored in the multi-valued image memory 3. The image data binarized by the comparator 4 is converted into change point data by the change point detection device shown in 5 and stored in the change point memory shown in 6. FIG. 2 shows a change point table formed on the change point memory 6. As shown in FIG. 3, this change point table shows that the bits of the image area (pixels in this case) are "1" to "0" or "0" to "0" when scanning the raster of the image from left to right. It is composed of a change value when changing to "1" and an address of the change point in the scanning direction. This is also a kind of analysis data regarding the image area. Reference numeral 7 denotes a register, which has a function of processing change point data on the change point memory 6. Reference numeral 8 indicates a component area data memory. This memory stores the component area feature relating to each component included in the image. That is, the change point data is analyzed via the register 7,
The component area feature generated by this is stored. 9
Indicates a component area number memory. The component area number memory 9 has information about which component area the change point on the raster belongs to, and is used for a component area numbering process described later.

Here, the terms used in this specification will be described. The parts included in the image are used in their ordinary meanings, such as A parts, B parts, and so on appearing in the one screen, but the distinction regarding the unity of the parts is not strict. This is because, for example, the respective figures showing one end face and the other end face of a single actual part may be grasped as two parts at least initially. Regarding areas, regions, and areas, what is cut out from the image in pixel units or raster units is an area for image processing.
The spread on the screen of the parts included in the image is the area,
A part of the area is a region.

The component area numbering algorithm will be described below. Comparison of change values and address data in the change point table of FIG. 2 is performed between two upper and lower rasters (areas) cut out from the image, and classified into the continuation, generation, and confluence patterns shown in FIG. . Each of these patterns will be described below. It should be noted that FIGS. 4 and 8 show two or more rasters in the upper and lower rows arranged in that order, and the numerals in the figures represent the component area numbers.

(1) Continuation pattern (Fig. 5) For example, the component area 1 of the upper raster in the adjacent rasters
When the area belonging to 0 and the area a of the lower raster are connected, the area number assigned to the part area 10, that is, 10 is accessed from the part area number memory 9 and the area a
The area number corresponding to is stored in the address on the memory. The parts area 10 is a "0" bit, that is, a background or missing area, but the same applies to the case where "1" bits are comrades.

(2) Occurrence pattern (FIG. 6) A new part area number such as 13, for example, is assigned to an area that is not connected to an area belonging to the part area 12 of the upper raster.

(3) Confluence pattern (Fig. 7) When an area belonging to the parts area 14 on the left side of the upper raster and an area belonging to the parts area 15 on the right side of the upper raster meet in the lower raster area b, the parts on the left side of the upper raster Area number 1
4 is attached to the area b of the lower raster, and the area number 15 attached to the parts area 15 on the right side of the upper raster is changed to the parts area number 14 on the left side of the upper raster.

As described above, in each of the continuous or merging patterns, the same number as the component area number corresponding to the upper raster is assigned to each area of the following rasters. For this reason, for example, if a part area number is assigned to the raster of the first row and it uses the propagating property for the following rasters one after another, the cutout for a certain part area is already completed during the cutout processing of the image area. Can be determined. The algorithm of this point will be described below. In the cut-out end portion of the part area 2 in FIG. 8, two separated areas belonging to the part area 1 having the same number as d among the two joining patterns c and d in the figure are cut out most recently. Image area (raster D)
It is judged that they will join together. In this case, it is not the component region 1 that cuts out, but the component region 2 that includes the region sandwiched by the two distant regions. Another confluence pattern C generated before the cutout of the raster D does not serve as a sign of the cutout of the component area 2. The merging pattern C is a merging with a component area that is considered to be a different component area at least at a stage before the merging occurs, and is not a merging of two regions belonging to the same component area. By detecting the cut-out end regarding the component area as described above, as shown in FIG. 9, the cut-out is performed even during the cut-out of the image area (the latest cut-out image area in FIG. 9 is the raster E). It is possible to distinguish between the finished component regions 20 to 28 and the component regions 30 to 34 being cut out.

As a special example in which it is difficult to determine the cutout end of the part area, there may be a case where there are part areas 41, 42, 43, 44 in contact with the outer frame of the imaged screen as shown in FIG. The reason why it is difficult to apply the above-described cutout end determination method to the component areas 41 to 44 in contact with the outer frame of the screen is that the propagating property of the component area number is cut off in the outer frame of the screen. Therefore, change point data between the screen outer frame 100 and the dotted line 200 (for example, 411) in the change point data of the screen imaged by the masking process as shown in FIG.
Is converted into image data (for example, 412) that does not contact the outer frame of the screen shown in FIG. 11 to cause the propagation of the component area numbers near the outer frame of the screen, the concept of the end determination is described. It can be applied as is.

FIG. 12 shows a memory table of the component area data memory 8 shown in FIG. The data related to each component area is stored in the order of the component area number, for example, 1 to 5. This area data memory 8 corresponds to a kind of storage means for storing data relating to the cut-out image area. Therefore, the total amount of data increases as the cutout of the image progresses. Further, the image (21 to 28, 30 to 3) of the image area cut out as shown in FIG.
If 4) is complicated, it will inevitably increase. If there are 30 parts based on the capacity prepared in advance as the part area data memory 8, the capacity will be insufficient when the 31st part area appears due to pattern generation. Under this situation, the data regarding the 31st parts area cannot be stored. The content of the memory table shown in FIG. 12 is returned, and the area relating to the component region here is a count of the number of pixel bits belonging to the region and is expressed in pixel bit units. The upper area number indicates the number of another component area that surrounds the component area. For example, in FIG. 9, the upper area number of the parts area 23 corresponds to the parts area number of the parts area 30.
The integration destination number in FIG. 12 is, for example, when a number other than zero is written therein, it means that the data relating to the parts area concerned is unnecessary, and the data of the parts area corresponding to the integration destination number is necessary. It indicates that they can be integrated. This integrated destination number often matches the upper area number,
It may be different. For example, looking at the part area 33 in the middle of cutting out in FIG. 9, the upper area is 30, but if the part areas 33 and 32 merge due to the subsequent cutting out, the part area 33 at the time of merging.
The number of the parts area 32 will be written as the integration destination of.

As described above, the data relating to the component areas 2 and 3 of FIG. 12 is unnecessary and can be erased. However, when the data is actually erased, the memory table of FIG. 12 is as shown in FIG. Is changed to. The data of the component area numbers 2 and 3 in FIG. 12 is completely erased after the processing of adding the area "10" and "17" to the area of the component area 1 corresponding to the integration destination number "1". It After the erasure, the parts area numbers 2 and 3 become blank, so the data of the parts areas 4, 5 ..., which are smaller than that, are packed toward the smaller numbers, and the area data table is rearranged as shown in FIG. As a result, a space for two component areas is generated in the component area data memory 8. Insufficient capacity of the component area data memory 8 is stored in advance as a reference value, for example, a number of 30 according to the size, and is judged by the occurrence of the 31st component area exceeding the number. However, under such a shortage condition, if the space for two component areas is created by erasing as described above, the 31st data can be stored at the component area number 29. Becomes In addition, the process of rearranging the area data table as shown in FIG. 13 is omitted, the 31st data is written in the column of the part area number 2 in FIG. 13, and the data before it is erased. It doesn't matter.

FIG. 14 shows a data table relating to the component area number memory of FIG. Here, the numbers of the component areas to which the area belongs are stored in order from the upper order of the raster to the lower order, and according to the order of the change points appearing in the same raster. In the upper left of FIG. 14, the order or rank is smaller. This component area number memory 9 is also a kind of storage means for storing data relating to the cut out image area. If unnecessary data is erased in a suitable manner as shown in FIGS. 12 and 13, for example, part number 2 in FIG. 14 is all blank, and the memory capacity is accordingly reduced. Will occur. As described above, the area of the component area is measured in the unit of pixel bit. This is, for example, when writing the component area number 4 in the memory table of FIG. 14, the pixel bit to which the area belongs in the raster is written. The number may be obtained from the address difference in FIG. 3, and the number may be added to the area “436” of the part region 4 in FIG. 3, for example.

Next, an algorithm for determining whether or not the data related to the part area that has already been cut out is necessary or unnecessary will be described. One effective means is to determine that the area is unnecessary when the area of the cut-out component area is less than a preset reference area. By the way,
If this reference area is set to 20 pixel bits, it is determined that the data concerning the component areas 2 and 3 in FIG. 12 is unnecessary, and the integration destination numbers 1 and 1 are recorded. Another means is a method of judging that the one having the smallest area out of the cut-out component regions is unnecessary. The above two means are based on the empirical rule that a component area having a very small area is an unintended component located far away from the screen background, or is likely to be noise, scratches, or stains. It is also possible to use the above two means together. For example, at first, the former method was adopted, and the capacity of the storage means became insufficient under the condition that the part area corresponding to the reference area or less was not found (including the case where the data of the corresponding part area was already deleted). In this case, the latter method is used.

The erasure of data related to the parts area has already been described. Although it has already been pointed out that the data relating to the area is added to that of the integration destination and then the data is erased, the timing of the erasure is advantageous when the capacity of the storage means becomes insufficient. This is to save the time required for the erase process and the subsequent memory table rearrangement process. Also, even if there is not enough space and unnecessary data remains because of this, it is more accurate to determine the necessity of data at a later stage of image processing, because the necessity determination will be more accurate. Is.

The change point memory 6 in FIG. 1 is also a kind of storage means for storing data regarding the cut out image area. The image to be cut out is that of the multi-valued image memory 3. In this case, the data for each cut out image bit is binarized, the change point is detected, and stored in the change point memory 6. .
Even when the capacity of the change point memory 6 becomes insufficient, the subsequent image area cannot be cut out. In this case, the data stored in the cut-out change point memory 6 is analyzed in parallel with the above-described cut-out processing of the image area or when the change-point memory 6 becomes insufficient in capacity, and the parts area is analyzed. By generating the data in the data memory 8 or the component area number memory 9 and determining the unnecessary data, it is possible to delete the unnecessary data from the change point memory 6 in which the capacity is insufficient. For example, if the area of address 7 in FIG. 2 belongs to a parts area determined to be unnecessary, address 7
Then, the change values 1 and 0 for the next 8 can be erased, and if they are erased to eliminate the lack of capacity, the subsequent cutting can be resumed.

FIG. 15 is a CPU that handles data such as the register 7 in FIG.
2 is a flowchart showing the outline of the program.
This will be explained below, but it is assumed that the block numbers in the figure are enclosed by [] to indicate the corresponding block, and subsequently, the explanation about the block will be added.

[501] = Call the data regarding the raster of the first row from the change point memory 6 to the register 7 [502] = Part area number assigned to each area (part of the part area) in the raster of the first row Is registered in the component area number memory 9. Further, the area, the upper area number, and the integration destination number are stored in the parts area data memory 8. The upper area number is set to 0 in the first area, and the other area is set to the part area number in the left area. The integration destination numbers are all 0.

[503] = Call the register 7 for the data of the change point memory 6 relating to the raster that is the target of the cut-out corresponding to the second line.

[504] = Refer to the change point table for two rows of the register 7 to perform pattern classification processing.

[505] = Change the program jump destination according to the form of the pattern. [506] = Determine whether the component area data memory 8 or the component area number memory 9 is sufficient. If there is a shortage, the data regarding the newly generated component area cannot be stored.

[507] = Delete the data related to the component area to which the integration destination number (other than 0) in the component area data memory 8 is given. Prior to that, the data regarding the area is added to the component area data of the integration destination. Finally, the memory table of the component area data memory 8 is rearranged.

[508] = Register the newly generated component area number in the component area number memory 9.

[509] = Register the data on the newly generated component area in the component area data memory 8. In this case, the integration destination number is 0, and the upper area number is the number of the part area to which the area on the left side of the area belongs.

[511] = Move to the right of the reference position on the change table corresponding to the upper and lower two rasters in the register 7.

[512] = If the reference position has already reached the address of the final pixel bit of the pair of upper and lower rasters, the comparison of the upper and lower rasters is completed.

[513] = End when the lower one of the pair of rasters hits the final row.

[515] = Cut out the data related to the highest raster that has not been cut out yet.

[521] = Write a required component area number in the corresponding address of the component area number memory 9. Further, the area of the corresponding area is added to the required area in the component area data memory 8.

[531] = Determine whether or not they are merging in the same component area. This is done by comparing the part area numbers of the two areas to be merged.

[532] = Compare with a preset reference area to check whether the area of the part area that has been cut out and confirmed by merging of parts areas (regions) of the same number is small.

[533] = Regarding the cut-out completed part area, write the number of the upper part area that wraps it into the integration destination number. Further, regarding the merged area, area addition processing in the component area data memory 8 and writing processing to the component area number memory 9 are also performed.

[541] = In the case of merging with different numbers, regarding the component area to which the right one of the two merging areas belongs, that is, the component area to be integrated, the remaining left area belongs to the integration destination number. Write the number of the parts area. In this case, the integration destination number and the upper area number are different from each other. Although it is not mentioned in [507], it is preferable to perform integrated erasing of component areas having different numbers of the integration destination and the upper area with priority over that of the same number. This is because the erasure is formal and does not correspond to the actual erasure of the parts area.

[542] = This is also a process peculiar to merging different numbers. This corresponds to the rewriting process of the number of the part area to which the right one of the two joining areas belongs. Since one or more of the above numbers have already been entered in the component area number memory 9, all of them are rewritten to the integration destination number.

The description of the flow chart of FIG. 15 is finished above, but supplementary description will be made. [501], [503], and [515] are for sequentially cutting out image areas to be processed, and [505]
[531] corresponds to a means for determining the cut-out end of each component area included in the already cut-out image area, and [531]
[2] is a means for determining the necessity of data relating to the part area for which cutting has been completed, and [507] corresponds to a means for erasing unnecessary data from the storage means.

〔The invention's effect〕

The invention is to eliminate necessary unnecessary data from the storage means for storing the data relating to the cut-out image area in the process of cutting out the image area, thereby eliminating the lack of capacity of the storage means. When the raster scan of the screen is processed up to the last line, the whole process is completed. According to this, the capacity of the storage means prepared in advance may be small, which is useful for reducing the processing cost. In addition, there is an effect that the image cutting processing time can be significantly shortened.

[Brief description of drawings]

All figures relate to the invention. FIG. 1 is a circuit for image processing, FIG. 2 is a memory table of a change point memory,
FIG. 3 is an explanatory view of change point data, FIG. 4 is an explanatory view of a pattern of a parts area, FIG. 5 is an explanatory view of a continuation pattern, FIG. 6 is an explanatory view of a generation pattern, and FIG. 7 is a confluence pattern. Explanatory drawing, FIG. 8 is a more detailed explanatory view of the confluence pattern, FIG. 9 is a plan view showing an image in the process of being cut out, FIG. 10 is a plan view showing an image before mask processing, and FIG. 11 is a view after mask processing. FIG. 12 is a plan view showing an image, FIG. 12 is a memory table of a component area data memory before erasing unnecessary data, FIG. 13 is a memory table after erasing unnecessary data, FIG. 14 is a memory table of a component area number memory,
FIG. 15 is a flow chart showing the main part of the processing method. 501, 503, 515 ... Cutting-out means, 6, 8, 9 ... Storage means, 505, 531 ... Means for judging completion of cutting-out of parts area, 532 ... Means for judging necessity of data, 50
7 ... Erasing means

Front Page Continuation (72) Inventor Yasushi Shimura 7-1-1 Higashi Narashino, Narashino, Chiba Prefecture Hitachi Keiyo Engineering Co., Ltd. (72) Inventor Seiro Tsuji 7-1-1 Higashi Narashino, Narashino, Chiba Hitachi Narashino Factory (56) References JP-A-58-207181 (JP, A) JP-A-57-13581 (JP, A) JP-A-53-118945 (JP, A) Practical application Sho-49-149633 (JP , U)

Claims (5)

[Claims] 1. A means for sequentially cutting out image areas to be processed, a storage means for storing data relating to the cut out image areas, and a space between two adjacent upper and lower image areas based on the storage means. A pattern classifying unit that correlates image areas and classifies the pattern into continuous patterns, occurrence patterns, and confluence patterns; and a unit that stores the image areas in the storage unit with component area numbers according to the pattern classification. When the image area included in the image area most recently cut out has a confluence pattern of the image areas with the same number, the part area end determination means for determining that the cutting out of the part area sandwiched by the image areas with the same number is completed And a component that determines the necessity of data relating to the component region that has been cut out at a predetermined timing before the capacity of the storage means becomes insufficient. And a data erasing means for erasing the data determined to be unnecessary by the area necessity determination means and the component area necessity determination means from the storage means at a predetermined timing before the capacity of the storage means becomes insufficient. An image processing method characterized by being provided. 2. The component area necessity determining means for determining that the data regarding the component area is unnecessary when the area of the cut-out component area is less than a preset reference area. Image processing method. 3. A component area necessity determining means for determining, when the capacity of the storage means is insufficient, data relating to a component area having the smallest area out of the component areas which have already been cut out, as unnecessary. The image processing method according to item 1. 4. When the capacity of the storage means is insufficient, it is determined whether or not data regarding the part area for which cutting has been completed is necessary, and the data determined to be unnecessary is erased from the storage means. The image processing method according to claim 1. 5. Data for erasing the data relating to the unnecessary component area from the storage means after waiting for the data relating to the area of the unnecessary component area to be added to the area of the upper component area surrounding the component area. The image processing method according to claim 1, further comprising an erasing unit.