CN202584162U - Image processing device - Google Patents
Image processing device Download PDFInfo
- Publication number
- CN202584162U CN202584162U CN 201220195299 CN201220195299U CN202584162U CN 202584162 U CN202584162 U CN 202584162U CN 201220195299 CN201220195299 CN 201220195299 CN 201220195299 U CN201220195299 U CN 201220195299U CN 202584162 U CN202584162 U CN 202584162U
- Authority
- CN
- China
- Prior art keywords
- image
- confidence
- degree
- unacceptable product
- state
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
The utility model discloses an image processing device. Through a confidence coefficient calculating unit, a reference image template prepared in advance is taken as a standard, the confidence coefficient for reflecting the high or low of the approximation degree of unqualified products or qualified products is prescribed, and according to the related confidence coefficient, the advantage and disadvantage determination on the state of an object is carried out. In this way, the accurate and fast determination can be carried out even on the object with an abstract fuzzy concept area, just like the processing of an object image.
Description
Technical field
The utility model relates to a kind of image processing equipment.
Background technology
At present; Through the outward appearance of reference object thing such as bottleneck, obtain photographic images, with the image that obtains be that the basis checks whether the sneaking into of foreign matter is arranged, the technology of defectives such as damaged, breach and dirt is very many; But these carry out 2 values with the view data of obtaining with certain threshold value; And test the black or the white area (area ratio) of 2 value data, and checked zero defect with this, this 2 simple value methods are difficult to improve the reliability of inspection sometimes; The image processing equipment that has can judge whether local shape evident characteristic specific part such as glass bottle opening is qualified, can check rapidly and exactly, but the object beyond the vial because there is not this tangible local feature, therefore can't be checked; Though the image processing equipment that has can be the basis with various parameters such as color that image information was showed, shape, structure, sizes; Automatically judge the attribute of object; But can't judge that to being presented at each state that cuts part this just has a strong impact on the accuracy of judged result.
Summary of the invention
In order to overcome above-mentioned defective, the utility model provides a kind of image processing equipment, and this image processing equipment can judge fast and accurately whether the state of object is good from the object image that photographs.
The utility model for the technical scheme that solves its technical matters and adopt is: one judges the image processing equipment of object state; Object to having certain form is taken; From the object images that photographs, judge the state of object, it comprises filming apparatus: the reference object thing also obtains object images; With reference to image storage unit: when judging the object state,, and store as the object of reference form parameter the related information parametersization of benchmark with reference to image based on object images; Extract the unit of object form parameter: from the object images that filming apparatus obtains, extract the object form parameter, as the comparison other of comparing with the object of reference form parameter; Confidence computation unit: degree of confidence is exactly to quantize the degree of approximation of object form parameter and object of reference form parameter; The state judging unit: the degree of confidence of calculating with confidence computation unit is a benchmark, judges the state of object.Here said form parameter is, parametrizations such as the shape that cuts in the image, color, texture, sizes, shows with the form of multi-C vector.
Through above-mentioned image processing equipment, the state judging unit is that the degree of confidence of calculating with confidence computation unit is benchmark, judges that whether good the state of object etc., therefore can judge the state of object fast and accurately.Because degree of confidence is to calculate from the relevant comparative result of form parameter; When judging that according to images such as photos the object state is whether good; Even if there is the fuzzy object of abstract concept; Also can compare fast and accurately, the higher degree of confidence of service precision is then judged the state of object.
Further improvement as the utility model; Also be equipped with the object of reference form parameter extraction unit that is used for extracting the object of reference form parameter; Obtain sample image through image processing equipment, after the related information parametersization of sample image, leave in reference in the image storage unit.
In this case,, utilize filming apparatus, can create the object of reference form parameter from the sample image that photographs at object of reference form parameter extraction unit.
As the further improvement of the utility model, confidence computation unit is calculated first degree of confidence and second degree of confidence respectively, and the basis of reference of two degree of confidence is different.The state judging unit according to first degree of confidence and second degree of confidence, is judged the state of object.Like this, judge from many aspects, can judge the state of object more accurately according to different benchmark.
As the further improvement of the utility model, said with reference to image storage unit, in the photographed image-related information of the photographed image-related information of representing the unacceptable product state and expression certified products state, can remember at least one information in the object of reference form parameter.Like this, with reference to image storage unit,, just can judge that object is unacceptable product state or certified products state according to canned data.
As the further improvement of the utility model, both comprised the image information that reflects the unacceptable product state with reference to image storage unit, comprise the image information of reflection certified products state again.Only from unacceptable product, sample the image information with reference to image template formation reflection unacceptable product state of the unacceptable product that sampling is obtained; Only from certified products, sample the image information with reference to image template formation reflection certified products state of the certified products that sampling is obtained.Like this under the situation, according to unacceptable product with reference to image template, can calculate degree of confidence about the unacceptable product state; According to certified products with reference to image template, can calculate degree of confidence about the certified products state.Calculate degree of confidence from unacceptable product state and two aspects of certified products state then, confidence level is just than higher like this.
As the further improvement of the utility model, when the state judging unit was the state judgement of benchmark judgement object with two threshold values, first threshold value relevant with degree of confidence was preferential judgment standard; When occurring than the little value of first threshold value, use first threshold value just can not judge, use second threshold value as benchmark.Like this, can carry out multistage judgement, like this, can guarantee reliability to a certain extent, can also reduce the probability that to judge.
As the further improvement of the utility model, with reference to image storage unit, the object of reference form parameter comprises in formal parameter, color parameter, structural parameters and the size parameter at least.In this case, in the profile of relevant image, color in structure or these information of size, for example according to the attribute and judgement target of object, is the basis with the obvious state that corresponding object was showed, and generates the object of reference form parameter.Even it is therefore judge the image that has abstract fuzzy concept zone as photograph image, also very reliable.
Further improvement as the utility model also has been equipped with unit.For be stored in reference to image storage unit with reference to image, learn with reference to the relevant information of image through upgrading.Like this,, not only can revise relevant information, can also improve the reliability of degree of confidence and judgement with reference to image through the study of unit.
Further improvement as the utility model; Comprise like this some with reference to image (as judge with a kind of benchmark of state by reference with reference to image) in the object of reference form parameter of relevant information; Learning a little unit can be from all object of reference form parameters average; Use divergence (refer to depart from, disperse degree), delete the parameter that those exceed predetermined value.In this case, as the not adopted information of benchmark, just can from relevant information, get rid of with reference to image.Just, can accumulate the object of reference form parameter that meets intention, improve the reliability of judging.
The beneficial effect of the utility model is: through above-mentioned image processing equipment; State judges that operation is that the degree of confidence of calculating with confidence computation unit is benchmark, judges that whether good the state of object etc., therefore can judge the state of object fast and accurately.Because degree of confidence is to calculate from the relevant comparative result of form parameter; When judging that according to images such as photos the object state is whether good; Even if there is the fuzzy object of abstract concept; Also can compare fast and accurately, the higher degree of confidence of service precision is then judged the state of object.
Description of drawings
Fig. 1 is the said image processing equipment constructing module of a utility model synoptic diagram;
Fig. 2 is the method flow diagram of the image processing equipment of use Fig. 1;
Fig. 3 A is for creating the process flow chart of unacceptable product with reference to image template;
Fig. 3 B is for creating the process flow chart of certified products with reference to image template;
Fig. 4 is for using the method flow diagram of amended example image treatment facility among Fig. 2;
Fig. 5 creates the process flow chart of unacceptable product with reference to image template for amended example image treatment facility among use Fig. 2;
Fig. 6 is the process flow chart that is used to judge that object is whether good;
The process flow chart of Fig. 7 for judging that amended example is whether good;
Fig. 8 is the process flow diagram of unacceptable product type judgment processing among Fig. 7;
Fig. 9 A-9B is the synoptic diagram that the image processing equipment among the 3rd embodiment comes auto divide image;
Figure 10 A is for creating the process flow chart of unacceptable product with reference to the split image template;
Figure 10 B is for creating the process flow chart of certified products with reference to the split image template;
Figure 11 is the method flow diagram of the image processing equipment of use the 3rd embodiment;
Figure 12 is the synoptic diagram of auto divide image;
Figure 13 is the method flow diagram of the image processing equipment of use the 4th embodiment;
Figure 14 is the method flow diagram of the image processing equipment of use the 5th embodiment.
Embodiment
Embodiment 1:
Below with reference to accompanying drawings, the image processing equipment relevant with first example of present embodiment described.
As shown in Figure 1, the image processing equipment relevant with the utility model comprise filming apparatus 1, with reference to image storage unit 2, confidence computation unit 3 and state judging unit 4.After the state judging unit is received input instruction,, do various corresponding processing according to these programs and data then from read the program and the data of appointment with reference to image storage unit.
Specifically; The state judging unit is in the view data of object (needing to judge the object of its state); And on the basis of the taken image information of filming apparatus; Whether well and in advance be stored in the object state and compare, judge and the output judged result thereby make with reference to the view data of image storage unit.For instance, the state judging unit can also be printed through output device judgment result displays on display device.
The solid-state image pickup device of being made up of CCD etc. is equipped with in filming apparatus 1 inside.The detected image of this solid-state image pickup device can be used as data image signal output.1 pair of object as the judgement object of filming apparatus is taken, and produces the object images that can judge its state through digital processing.
With reference to program storage area and data temporary storage area are arranged in the image storage unit.But program storage area is used for storing the various programs of a plurality of runs image processing equipment, and input instruction, input data, result etc. can be deposited in the data temporary storage area temporarily.When the approximation of the captured image of measuring and calculating filming apparatus,, various programs have been stored in the memory device for the ease of relatively.Such as, parametric program is the parametrizations such as shape that cut from image.Be provided with the confidence calculations program in the confidence computation unit,,, calculate degree of confidence high the quantizing of those approximation through the relevant view data of comparative parameter image.In addition, with reference to also deposit in the image storage unit typical image information with reference to various information such as image templates, these information are meant, in order to judge the object state, relatively the time, as the whether good state of benchmark.For example, the image that filming apparatus is captured is as extracting with reference to image, the view data after the parametric program parametrization, as preserving with reference to the object of reference form parameter of image template (typical show state whether preferable image).From read the object of reference form parameter with reference to image storage unit, comparison other thing form parameter and reference figure present situation parameter are calculated the higher index of approximation to confidence computation unit as required, and whether state judgment unit judges object is good.And before carrying out the operation that image processing equipment judges that the object state very denys, be prepared in advance with reference to image template.
After the various information dataizations that image is correlated with, be included in the above-mentioned form parameter.Form parameter is as image information, shows with the form of multi-C vector.For example meet correlation parameters such as color in the image, shape, texture, size, and serve as that the object state is judged on the basis with these characteristics.About color parameter, can be made into mean value, the median of RGB three looks, the histogram of color, the deviation that draws can be used as judgment standard on the basis of measuring approximate altitude.About form parameter,, can be used as judgment standard measuring on the basis of approximate altitude of circularity, complexity, kurtosis and moment that profile (edge) cuts figure.The circularity here is the index according to the aspect ratio decision that cuts figure; Whether complexity is for having the index of range of flexion decision of number or the lateral profile in cavity, cavity according to cutting figure.Kurtosis according to the distance of air line distance and outline line than, the central angle of outline line and the ratio decision of pixel count; Moment is according to the shape decision that cuts figure.About parametric texture, the pattern related data deviation that will obtain through the one dimension sciagraphy can be used as judgment standard on the basis of measuring approximate altitude.About the i.e. parameter of size of size,, can be used as judgment standard measuring on the basis of approximate altitude that profile (edge) cuts area of graph, girth and Fu Leite diameter (vertical or horizontal).
Based on the image processing method of utilization image processing equipment shown in Figure 1, will an example of the main operation of judging object be described below.Fig. 2 is a flow chart illustration, is the abstract specification to image processing method.Judge whether good creating at this as benchmark, be kept at, just the thing of preparing for judgement is described with reference in the image storage unit with reference to image template.About the set-up procedure with reference to image template, the back can be elaborated in Fig. 3 A the inside.
At first, filming apparatus operation photographing program, the image that extraction will be judged (S1).Secondly, filming apparatus positions (S2) to the judgement object images of taking.For example, suppose to judge that object images is the image with housing, so just can along the profile of housing as the benchmark of locating.Secondly, read parametric program, from judge object images, extract form parameter (S3), be saved in reference in the image storage unit with reference to image storage unit.Through above step, but judge that the object data turn to interpretable state.
The state judging unit is according to comparing judgement from the determining program that reads with reference to image storage unit.At first, be stored in advance with reference to image storage unit as judgment standard with reference to image template in, confidence computation unit reads form parameter (S4a) from unacceptable product with reference to image template.Then, confidence computation unit from unacceptable product with reference to the form parameter of the judgement object that extracts among the form parameter (also being the object of reference form parameter) of image template and the S3 compare (S5a); According to the comparative result among the S5a, from judge object images, can calculate the degree of closeness of judging object images or object state and typical unacceptable product, also be degree of confidence, and this degree of confidence as first degree of confidence (S6a).Just be referred to as the unacceptable product degree of confidence to this first degree of confidence below.About the unacceptable product degree of confidence,, need consider and use various algorithm its situation about stipulating according to relevant form parameter such as the color of image, shape.Such as, consider the poor of form parameter from vectorial aspect, be the basis with the distance of the poor defined of each composition of vector, can determine the unacceptable product degree of confidence.Compare with the typical unacceptable product that compares among S5a, the S6a, calculating is tried to achieve; Quantize the height of the degree of approximation (such as the distance of above-mentioned vector) (hereinafter; Call the unacceptable product index to the data after quantizing), the unacceptable product index as the unacceptable product degree of confidence.Just, the unacceptable product index is high more, just is judged as unacceptable product to it.Such as when the parameter of the view data that obtains and the identical words of view data parameter of most typical unacceptable product, just be decided to be maximal value 100% to the unacceptable product index.
Here, for the unacceptable product exponential quantity of the top regulation of picture,, judge whether object is unacceptable product through the first threshold and second threshold value that provides.First threshold is set to than higher value, judges as unacceptable product, is to satisfy the value of confirming scope.The ratio first threshold of second threshold setting is little, through setting higher to a certain extent value, still just becomes second threshold value than those higher values as the possibility that above second threshold value is unacceptable product.The first threshold and second threshold value are judged as the unacceptable product about the known object thing, are appointed values in confirming category.The threshold value that can accurately judge under the situation of object position, had better be determined threshold value, such as can be through adjusting repeatedly.
About takeing a deviate with defective moral character certainty factor below the setting of first and second threshold value is the example that defined threshold is come on the basis.At first; From unacceptable product property certainty factor all objects from high to low, randomly draw sample (first sample); Calculate the standard deviation of sample on distributing, in this case, the value of the defective moral character certainty factor when being higher than regulation to deviate is defined as first threshold.Through stipulating above-mentioned first threshold, in the situation that surpasses first threshold, the numerical value that is this value unacceptable product as accurate judgement object is so used.In addition; For example calculate the standard deviation of sample (second sample) on distributing that the set of typical image inferior is promptly randomly drawed from the higher relatively object of unacceptable product property certainty factor; In this case; Though deviate is lower than the mean value of second sample, also have to a certain extent to be higher than mean value, be defined as second threshold value to the deviate that is higher than the above-mentioned first sample mean value.In the set of typical case's image inferior, because have defective moral character certainty factor to a certain degree, so use as judgement unacceptable product chance this value than higher numerical value.According to the distribution character of each sample, regulation first and second threshold value improves and judges to have the reliability of abstract fuzzy concept section object.In addition, through setting the first and second above-mentioned threshold value, when the adjustment first threshold, the object that can judge is limited in scope more accurately, is judging then.In addition on the one hand, when adjustment second threshold value, can judge more corresponding with it objects as much as possible.Through suitably adjusting first and second threshold value, can set according to the degree of accuracy of necessity.
Parallel with above-mentioned S4a-S6a, read form parameter (S4b) from certified products with reference to image template with reference to image storage unit.Then, confidence computation unit from certified products with reference to the form parameter of the judgement object that extracts among the form parameter (also being the object of reference form parameter) of image template and the S3 compare (S5b); Secondly, from judge object images, calculating the degree of closeness with certified products, also is degree of confidence, and this degree of confidence as first degree of confidence (S6b).Just be referred to as the certified products degree of confidence to this second degree of confidence below.Such as the parameter of working as the view data that obtains and the identical words of view data parameter of most typical certified products, the certified products index just is decided to be maximal value 100% so.For the value of regulation, just as first or second threshold value stipulated in the form parameter of above-mentioned unacceptable product with reference to image template such, certified products with reference to the form parameter of image template in also setting threshold, whether the threshold decision object is certified products thus then.But, exceed threshold value still for the possibility of certified products still than higher.
As stated, can judge judging object from first degree of confidence and two aspects of second degree of confidence here.Judge from many aspects according to different benchmark, can judge the state of object more accurately.Such as, use first degree of confidence to be judged as unacceptable product to an object, if use second degree of confidence not to be judged as certified products to it again, so just can conclude that this object is that the possibility of unacceptable product is very high.
If the unacceptable product degree of confidence that S6a calculates has exceeded first threshold; And the certified products degree of confidence that S6b calculates is judged as again and does not exceed threshold value (S7a:Yes), and the state judging unit just is judged as unacceptable product (S8) to this judgement object images (judgement object).
On the other hand, if in S7a, the unacceptable product degree of confidence is below the first threshold, and the certified products degree of confidence is threshold value above (S7a:No); The state judging unit will be made such judgement so: the certified products degree of confidence has exceeded threshold value, and whether the unacceptable product degree of confidence exceeds threshold value (S7b).If here the certified products degree of confidence exceeds threshold value, and the unacceptable product degree of confidence do not exceed the 1st threshold value (S7b:Yes), so just this judgement object images or judge that object is judged as certified products (S9).
On the other hand, if in S7b, the certified products degree of confidence is below the threshold value or the unacceptable product degree of confidence is first threshold above (S7b:No); The state judging unit will be made such judgement so: the unacceptable product degree of confidence has exceeded first threshold, and whether the certified products degree of confidence exceeds threshold value (S7c).Here; If the unacceptable product degree of confidence exceeds first threshold; And the certified products degree of confidence has also exceeded threshold value (S7c:Yes); Compared with the certified products degree of confidence as second degree of confidence, the unacceptable product degree of confidence of first degree of confidence will preferentially be used so, is judged as unacceptable product (S8) to this judgement object images or judgement object.
If in S7c, the certified products degree of confidence is below the threshold value or the unacceptable product degree of confidence is (S7c:No) below the first threshold; The state judging unit will be made such judgement so: whether the unacceptable product degree of confidence has exceeded second threshold value (S7d).,, preferentially use the unacceptable product degree of confidence here, be judged as unacceptable product (S8) to this judgement object images or judgement object if the unacceptable product degree of confidence has exceeded second threshold value (S7d:Yes).
The state judging unit judges that at S7d the unacceptable product certainty factor is lower than second threshold value (S7d:No), is the basis with the statistics ratio, carries out very not judging (SQ).Add up the certified products known and the ratio of unacceptable product in this so-called statistics ratio in the object for judging in inspection.Be judged as at S7d under the situation of No, judge that object images is that certified products certainty factor and the unacceptable product certainty factor of object is not high,, we can say that it very is not very difficult judging so be impossible judge through the method before the above-mentioned S7d.In SQ, the object that is in this situation is used random number etc., through deciding certified products or unacceptable product with the corresponding ratio of statistics ratio.For example, during object, the ratio of understanding unacceptable product and certified products roughly is q in inspection: (1-q), the state judging unit is certified products at SQ with the probabilistic determination object of q * 100%, is unacceptable product with the probabilistic determination object of (1-q) * 100%.
After the instruction of acceptance from the state judging unit, the judged result of above-mentioned S8 or S9 can show on a display device.At this moment, also can show numerical results such as unacceptable product certainty factor and certified products certainty factor together.
To use Fig. 3 A and Fig. 3 B below, explain how to create as judgment standard with reference to image template.Flow chart illustration 3A is to the explanation of unacceptable product with reference to image template establishment operation, and flow chart illustration 3B is the explanation of certified products being created operation with reference to image template.That is to say that the operation shown in Fig. 3 A is the form parameter of reading in the S4a of Fig. 2 providing.Operation shown in Fig. 3 B is the form parameter of reading in the S4b of Fig. 2 providing.
At first, in Fig. 3 A,, prepared the sample that a plurality of (or) typically are evident as unacceptable product in order to create unacceptable product with reference to image template.Then, therefrom read parametric program, from reference to extracting form parameter (S103a) the image with reference to image storage unit.That is to say the function that object of reference form parameter extracting unit is arranged with reference to image storage unit.At last, for those comprise the form parameter information that extracted with reference to image, with reference to image storage unit these relevant informations with reference to image are preserved in the specific region above that (S104a).Through aforesaid operations, just created well with reference to image template.Because will prepare a plurality of typical samples usually, need repeat the operation of above-mentioned S101a-S104a, each sample is carried out parametrization.
In addition, shown in Fig. 3 B, certified products are with reference to the establishment of image template, are the same with above-mentioned unacceptable product with reference to the establishment of image template, and are just passable through the typical sample that human eye just can determine.Through the operation of above-mentioned S101b-S104b, can create with reference to image template.
Top sample is to be equivalent to the out and out things that reality is judged object, for image sample might as well.In this case, image sample can need not obtain image by filming apparatus through handling with reference to image storage unit as digital image.Sample becomes by parameterized data, directly this data storage is being got final product with reference to image storage unit in this case.
In this preoperative debug phase, the wrong situation of judged result has a lot, need repeat above-mentioned study in this case, up to can appropriately judging exactly.
The image processing equipment relevant with present embodiment; Confidence computation unit is benchmark with what provide in advance with reference to image template; Determine the degree of confidence of reflection and unacceptable product or certified products degree of approximation height, then according to this degree of confidence, whether the state of state judgment unit judges object is good.Therefore, even, also can prepare to judge rapidly as the process object object image to the object in abstract fuzzy concept zone is arranged.
In above-mentioned, even do not carry out based on the statistics shown in the SQ also more passable than the good not judgement of row.For example at S7d, be lower than second threshold value (S7d:No) when the unacceptable product degree of confidence is judged as, the state judging unit is judging object images or judging that object is judged as certified products and also is fine.
There are a lot of objects to proceed to SQ and can't judge that but the judgement of a lot of mistakes is also arranged, when the low situation of similar this judgement degree of accuracy continues, can solve through adjusting each threshold value.But hope to carry out the adjustment of threshold value before in the stage of promptly trying out in the previous stage of runs image processing equipment.In this operation; Under the situation that arrives S7d even SQ, for example stored the information of the form parameter that arrives these steps, can remove to observe the relevant image of these form parameters with human eye; Confirm whether be certified products; Be sample with this sampled images then, forcibly create again shown in Fig. 3 A or Fig. 3 B with reference to image template, also be fine.
In addition, if it is fewer to arrive the situation of SQ, do not make a decision passable at SQ yet.In this case,, consider not use equipment to judge automatically, judge and be to use such as the such method of human eye for the image that arrives SQ.
To the image processing equipment relevant with modified example be described with reference to process flow diagram Fig. 4 below.Because the formation of the image processing equipment in this modified example is identical with Fig. 1, just diagram and explanation have not been narrated here.In an above-mentioned example, with reference to image template be preparing in advance or the thing of prior learning before the runs image processing equipment.In this modified example, such as in operating process,, upgrade with reference to image template through suitably study, can realize improving, revise, developing the target of judgment standard.
Fig. 4 makes an amendment the wherein part of Fig. 2 to obtain; It has reflected through carrying out and same operation shown in Figure 2; Be judged as unacceptable product or after S9 is judged as certified products, upgrade operation (also can be described as the study operation) (S10a, S10b) at S8 with reference to image.Specifically, if be judged as unacceptable product at S8, CPU10 at first confirms to judge the deterministic process of the form parameter of object; If the situation that is judged as Yes, is judged as unacceptable product is arranged in S7, the information of this form parameter is added (S10a) with being used as new unacceptable product with reference to the part of image template so.Wherein, in step S7a, be judged as the object of Yes, just as described, surpass the object of threshold value above first threshold and certified products degree of confidence for the unacceptable product degree of confidence.Add as unacceptable product such image with reference to image template, can improve the reliability of unacceptable product, in other words, just can improve the reliability of the unacceptable product degree of confidence that S6a calculates with reference to image template.
Likewise, if be judged as goodly at S9, the state judging unit at first confirms to judge the deterministic process of the form parameter of object; If the situation that is judged as Yes, is judged as certified products is arranged in S7, the information of this form parameter is added (S10b) with being used as new certified products with reference to the part of image template so.Through improving the reliability of the certified products degree of confidence that S6a calculates like this.
As above, in this modified example,,, bringing into play the effect of unit with reference to image storage unit through upgrading with reference to photographed image-related information in order to improve the reliability of unacceptable product degree of confidence and certified products degree of confidence.
The learning manipulation of relevant S10a and S10b, be included in also except the situation of in this operation, at every turn carrying out that this operation indirect property carries out etc. the operation carried out of various time.When adding with reference to the part of image template, for example carry out the grade rank through unacceptable product degree of confidence relatively, get rid of inferior grade, in reference to image template, carry out interpolation, change, the deletion of data, unified these data of preserving.
Embodiment 2:
To the image processing equipment relevant with the 2nd embodiment of the utility model be described through Fig. 5 etc. below.
Fig. 5 is the explanation of using as judgment standard in this example creating with reference to image template.Shown in diagram, improved multiple unacceptable product relevant with reference to image template.
For example, it is generally acknowledged that certified products have typical pattern, that is to say the data of collecting average through preparing a plurality of or single sample, can generate can accurately judge the certified products approximation with reference to image template.But according to the differences such as attribute of object, the reason that unacceptable product produces has a plurality of.All be very different on each reason as shape that unacceptable product showed etc.As the reason that produces unacceptable product, the sneaking into of dislocation, foreign matter for example arranged, break, a variety of causes such as breach and stain.According to the difference of these reasons, show the shape on the object, color, the difference of size etc. also has a great difference, and the characteristic that form parameter showed also can be variant.Although be such situation, the object of reference form parameter of concluding about these typical unacceptable products of multiple type generates a template, and parameter is too in confusion so, just may not judge rightly and unacceptable product between approximation.To this, as shown in Figure 5, present embodiment is that type is categorized as a plurality of different samples respectively with the reason that each unacceptable product produces, with these samples of being classified generate singly relevant unacceptable product with reference to image template, these as the benchmark of judging.For example, collect the such unacceptable product sample of breaking of relevant object,, generate a template (S201a-S204a) as Class1; Collect the such unacceptable product sample of breach of relevant object,, generate a template (S201b-S204b) as type 2.Carry out same operation below, produce many groups with reference to image template.
Under the situation about being prepared with reference to image template of Fig. 6 for the relevant multiple unacceptable product of explanation, judge the process flow diagram of an example of object method.At this, about a kind of being ready to of unacceptable product with reference to image template and relevant (L-1) middle unacceptable product with reference to image.The dissimilar L kind that has inferior takes place, and types typical inferior of preparing with reference to image template in the middle of these have (L-1) to plant, do not belong to these to the type that poor quality takes place in the middle of any one type as last L kind.
At first, confidence computation unit is the basis with relevant (L-1) type unacceptable product with reference to image template, calculates unacceptable product certainty factor (S26a).Confidence computation unit is the basis with relevant certified products with reference to image template, calculates certified products certainty factor (S26b) synchronously with S26a.
Secondly, if the state judging unit is made as judging: the certified products certainty factor that all types of unacceptable product certainty factors that S26a calculates are all calculated greater than first threshold and S26b surpass threshold value (S27a:Yes), can be judging that object images judges promptly that object is judged as and be unacceptable product (S28).
If the state judging unit is made as judging: among the S27a, all types of unacceptable product certainty factors all less than first threshold or certified products certainty factor greater than threshold value (S27a:No); The certified products certainty factor greater than threshold value and unacceptable product certainty factor whole step less than first threshold (S27b:Yes); Can judge promptly that object is judged as certified products (S29) to the judgement object images.
If the state judging unit is made as judging: among the S27b, the unacceptable product certainty factor all greater than first threshold or certified products certainty factor less than threshold value (S27b:No); The certified products certainty factor greater than threshold value and unacceptable product certainty factor whole step greater than first threshold (S27c:Yes); This moment, the preferential unacceptable product degree of confidence of using was judged, judged promptly that object is judged as to the judgement object images and was unacceptable product (S28).
If the state judging unit is made as judging: among the S27c, the unacceptable product certainty factor all less than first threshold or certified products certainty factor less than threshold value (S27c:No); Unacceptable product certainty factor whole step is greater than second threshold value (S27d:Yes); This moment, the preferential unacceptable product degree of confidence of using was judged, judged promptly that object is judged as to the judgement object images and was unacceptable product (S28).
If the state judging unit is made as judging: among the S27d, the unacceptable product certainty factor carries out state according to the statistics ratio and judges (SQ) all less than second threshold value (S27d:No), and judgement is certified products or unacceptable product (S28,29).
Fig. 7 be to present embodiment in describing of managing mutually of modified example with reference to image template.In example shown in Figure 7,, further the type of unacceptable product is carried out classification processing (SR) about the judgement object images that S28 among Fig. 6 is judged as unacceptable product.
Below, an example of the unacceptable product classification processing through SR in Fig. 8 key diagram 7.At first, whether the state judging unit carries out having in all types of the unacceptable product certainty factor to surpass the situation (SR1) of first threshold to the object images of judging that is judged as unacceptable product at S28.Judge that at SR1 any one type surpasses first threshold, (SR1:YES), the corresponding object of this judgement object images is judged as such unacceptable product (SJ1).A plurality of types of judging object images surpass first threshold, in this case, can judge with the corresponding all types of this object all be unacceptable product.For example; Class1 is represented the unacceptable product certainty factor that relevant object breaks; The unacceptable product certainty factor of the relevant object breach of type 2 expressions; If certain judges that the Class1 and 2 both sides of object images surpass first threshold this situation, so with this judgement object images corresponding judgment object as having taken place to break and the object of two kinds of poor quality property of breach is handled.
On the one hand, judge at SR1 whether the unacceptable product certainty factor that all types of unacceptable product certainty factors do not have to surpass the 1st threshold value (SR1:NO), judge Class1-type (L-1) surpasses the 2nd threshold value (SR2).At SR2, judge whether the unacceptable product certainty factor of Class1-type (1-1) surpasses second threshold value (SR2:YES), judge and this judgement object images corresponding judgment object Class1-type (L-1) in the unacceptable product of respective type (SJ2) with it.Relevant a plurality of type surpasses under the situation of second threshold value, is judged as corresponding with it all types of unacceptable products.
On the other hand,, judge that the unacceptable product certainty factor of Class1-type (L-1) is lower than second threshold value (SR2:NO), with this judgement object images corresponding judgment object unacceptable product (SJ3) that is type L at SR2.That is to say, do not belong to (L-1) that prepared with reference to image template and plant any one in the typical unacceptable product, is L kind unacceptable product.
Top classification processing the unacceptable product of SR.Be judged as judgement object images above first threshold at this at SR1 and add (SK) with reference to image template as relevant unacceptable product.That is to say the study operation of adding SK, improve the reliability of the unacceptable product certainty factor of every kind of type inferior with this.
Present embodiment also can be as the object image is carried out rapidly and accurately for the judgement of the object with abstract fuzzy concept zone.
Particularly, present embodiment also is suitable for for the judgement object that exists a plurality of unacceptable products to produce type.
Also have, if it is many that the situation of L unacceptable product takes place, can it add to newtype with reference in the image template.
In above-mentioned S27a etc., judge that except carrying out state mean value that can also a plurality of types of foundation judged according to the threshold value of the unacceptable product degree of confidence of each (L-1) type.Such as; Being assumed to A1-AK to the unacceptable product degree of confidence of K type, is all types of unacceptable product generation rates that frequency is decided to be α 1-α K, and weighting unacceptable product degree of confidence is set at ∑ α i * Ai; Just can carry out state and judge according to this weighting unacceptable product degree of confidence this moment.Specifically, also can set the relevant threshold value of weighting unacceptable product degree of confidence, carry out alternate process according to the statistics ratio of SQ, if the value of ∑ α i * Ai then is judged as unacceptable product greater than above-mentioned threshold value; Otherwise,, then be judged as certified products if less than above-mentioned threshold value.
Since produce the type of unacceptable product exist a plurality of, need provide a plurality of types with reference to image template.For certified products, though the just much of that with reference to image template of a type only is provided, the certified products that a plurality of types had better be provided are with reference to template.In addition, the unacceptable product that a plurality of types are provided is with reference to template, and to unacceptable product only provide a type with reference to template, this situation also possibly exist.
Embodiment 3:
To the image processing equipment relevant with the 3rd embodiment of the utility model be described through Fig. 9 A etc. below.
Fig. 9 A, Fig. 9 B describe the Flame Image Process in the present embodiment.As shown in the figure; All images PH shown in Fig. 9 A that filming apparatus (with reference to Fig. 1) obtains is through cutting apart automatically, shown in Fig. 9 B; Cutting is P1-Pm, use each split image produce template (being designated hereinafter simply as) with reference to the split image template as judgment standard for use.In the example of Fig. 9 A, Fig. 9 B, (be 6 row, 8 row among the figure, just m=6 * 8=48), image is shown in synoptic diagram to take m the bottle BT1-BTm that leaves the insides such as box in.In this case, the arrangement of m bottle is reserved in advance, according to this arrangement, unified cut-off rule LL according to the defined area, reflect from before all image PH, automatic segmentation comprises the appearance of m the split image P1-Pm of each bottle.In the present embodiment,, produce respectively, then successively as the benchmark of judging object with reference to the split image template for the split image of cutting apart automatically.About above-mentioned automatic cutting operation, such as with reference to image storage unit (with reference to Fig. 1) when preserving the relevant data of cut-off rule LL, according to the data of being preserved, can preserve the segmentation procedure of cutting apart automatically; Can read these corresponding data and programs with reference to image storage unit then, carry out the image segmentation operation.
To the making with reference to the split image template be described through Figure 10 A, Figure 10 B below.
At first, in order to make unacceptable product, realize providing a plurality of (or single) significantly typical unacceptable product sample among Figure 10 A with reference to the split image template.Filming apparatus (with reference to Fig. 1) is taken sample, then photograph with reference to image as a whole image PH read and come (S301a), then general image PH is positioned (S302a) and area dividing (S303a).Specifically, from read data and the program of cutting apart automatically with reference to image storage unit, the zone according to cut-off rule LL delimit is divided into m split image P1-Pm to general image PH.Then, read parametric program, extract the form parameter (S304a) of each split image P1-Pm respectively.At last, the image information that comprises above-mentioned form parameter, be saved in (S305a) in the specific region with reference to image storage unit with reference to split image.Arrive this, just produced unacceptable product with reference to the split image template.
Shown in Figure 10 B, certified products are with reference to the making of split image template, and are identical with reference to the making of split image template with above-mentioned unacceptable product.Through the operation of S301b-S305b, can produce certified products with reference to the split image template.
About the object determination methods with reference to the split image template of using aforesaid operations to improve, flow chart illustration Figure 11 describes an example of this determination methods, and Fig. 2 of it and the 1st embodiment is corresponding.
At first, filming apparatus reads (S1) to general image as the judgement object images, positions (S2) then.Read data and the program of cutting apart automatically with reference to image storage unit then, be divided into m to corresponding general image automatically and judge Object Segmentation image (S401).Secondly, read parametric program, from individual judgement Object Segmentation image, extract form parameter (S403) respectively with reference to image storage unit.
Confidence computation unit reads form parameter (S404a) with reference to the split image template and judges that the form parameter of object compares (S405a), calculates unacceptable product degree of confidence (S406a) from unacceptable product.Synchronous with above-mentioned S404a-S406a, confidence computation unit compares with reference to the form parameter that reads form parameter and judgement object the split image template from certified products, calculates certified products degree of confidence (S404b-S406b).
Identical with situation shown in Figure 2; The state judging unit will judge whether Object Segmentation image (promptly judging each partitioning portion of object) is judgement (S407a-S407d, the SQ of certified products according to unacceptable product degree of confidence and certified products degree of confidence to each; S8, S9).
In the present embodiment, even if, also can as the process object object image, judge fast and accurately to those images with abstract fuzzy concept zone.
Particularly, in the present embodiment, can carry out quality to each partitioning portion of whole object thing and judge.Like this, if there is zone inferior in the part in the integral body, just can refers in particular to that part of zone inferior and fix.
Give one example again about cutting apart automatically at present.Shown in figure 12, be accommodated in the feeder container of uses such as box lunch among the container TR as object, according to the accommodation space SP1-SP6 that divides in the container TR, can carry out area dividing.The example of Figure 11 is unified to cut apart through cut-off rule LL, but can be divided into the zone of each shape, different sizes here.
Also be provided with unit in this example, be used for improving the reliability of certified products degree of confidence and unacceptable product degree of confidence.
Embodiment 4:
To the image processing equipment relevant with the 4th embodiment of the utility model be described through Figure 13 below.This image processing equipment is the variation of the 1st embodiment.About the structure of equipment, since identical with image processing equipment in the 1st example shown in Figure 1, so here diagram and explanation just have been not described herein.
Figure 13 is the abstract flow chart illustration of image processing method.In the present embodiment, only unacceptable product with reference to image template as benchmark with reference to image template, and do not use certified products with reference to image template, be different with the 1st example in this.That is to say; Extract the form parameter of judgement object images through S1-S3 after; Only compare the form parameter that is drawn into through S4 and S5, utilize S6 to calculate unacceptable product index (unacceptable product degree of confidence) then with the form parameter of unacceptable product with reference to image template.If this unacceptable product degree of confidence is greater than threshold value (S7a:Yes), judging that object is judged as unacceptable product (S8); If this unacceptable product degree of confidence is not more than threshold value (S7a:No), then judging that object is judged as certified products (S9).
In the present embodiment, even if, also can as the process object object image, judge fast and accurately to those images with abstract fuzzy concept zone.
More than be only to have used unacceptable product, do not use certified products with reference to image template with reference to image template.On the contrary, only use certified products with reference to image template, and do not use unacceptable product also to exist with reference to the situation of image template.
Also be provided with unit in the present embodiment, be used for improving the reliability of certified products degree of confidence and unacceptable product degree of confidence.
Embodiment 5:
To the image processing equipment relevant with the 5th embodiment of the utility model be described through Figure 14 below.This image processing equipment is exactly the variation of the 1st example-the 4th embodiment.About the structure of equipment, since identical with the image processing equipment of the 1st embodiment shown in Figure 1, so here diagram and explanation just have been not described herein.
The image processing equipment of present embodiment, identical with the 1st embodiment-the 4th embodiment, through the Flame Image Process shown in any example, carry out quality repeatedly and judge.And in good and bad decision operation, upgrade with reference to image template.Just to carry out learning manipulation.In this case, when upgrading, might absorb an inappropriate part as basis of reference with reference to image template with reference to image template.Or, when upgrading with reference to image template change judgment standard, consider handlebar whether suitable with reference to image template as inappropriate situation about handling.In this example, designed the supervision Elementary Function, supervisory with reference to image template, suitable to confirm with reference to image template, use as common judgment standard.
Shown in figure 14, at first, the state judging unit carries out quality and judges (S701).In S701 operation, if be updated with reference to image template, confidence computation unit will be calculated divergence (step S702) from each form parameter average.Here said divergence is the value of reflection with respect to the average degree of divergence of certain form parameter that comprises with reference to image template.Such as, in the form parameter that certified products comprise with reference to image template,, consider and include all multi-C vectors in the vector value in certain category for all multi-C vectors that color, shape, texture, size etc. show.Therefore, but be not blended into certified products with reference in the image template with reference to the form parameter of image template, then be used as the multi-C vector of the form parameter of sneaking into the different vector of the multi-C vector of other form parameter and consider if originally do not belong to certified products.According to certain rule; Quantize certified products with reference to all form parameters that image template comprises; Ask poor to the value that obtains after quantizing and all form parameter mean values then, the difference of asking is defined as the divergence of each form parameter, utilizes divergence to specify out the form parameter that should get rid of then.Confidence computation unit calculates the number (S703) that divergence surpasses the form parameter of setting.Confidence computation unit can be supervised with reference to image template always, and the number of the form parameter that calculates up to S70 reaches the number (S704) of regulation.If in S704, made the judgement that reaches the regulation book, confidence computation unit will from reference to the image template deletion corresponding form parameter (S705).Confidence computation unit repeats aforesaid operations always, finishes (S706) up to good and bad judgment processing.In above-mentioned steps, confidence computation unit is being brought into play the effect of supervision with reference to the supervision unit of image template.
For the value after the quantizing of above-mentioned each form parameter, can use the certified products degree of confidence.With reference to the form parameter of image template, be the relevant information of typical unacceptable product or certified products all, unacceptable product degree of confidence and certified products degree of confidence should be all unusual height.
Therefore, obtain the standard deviation that degree of confidence distributes,,, just get rid of this parameter, can be used as judgment standard like this and guarantee proper state if also promptly there is the very large deviate of divergence to exist if the extremely deviate of deviation average is arranged in the parameter.Surpass the situation of divergence relevant regulations value, refuse exactly to be included into the zone that degree of confidence is deferred to the situation of normal distribution to it.
In the present embodiment, even if, also can as the process object object image, judge fast and accurately to those images with abstract fuzzy concept zone.
Particularly, whether present embodiment, is supervised suitable as judgment standard commonly used with reference to image template through upgrading with reference to image template or carrying out the relevant study of judgment standard.
In conjunction with above each embodiment the utility model is described, but to have more than be to be defined in each above-mentioned embodiment to the utility model.For example,, considered with form parameter to be the basis, situation about stipulating through various computing method, can carry out various adjustment to this according to the character of judging object about the color of image, shape etc. about defective moral character certainty factor.For example, can carry out weighting to various parameters such as color, shapes.Specifically; Judge whether object is good,, will strengthen the judgement proportion of color parameter key element in this case compared with other key element owing to the color distortion of judging object shows huge difference; The priority processing color parameter can be judged with this more exactly.
In above-mentioned embodiment; In order to detect defective,, adopt unacceptable product degree of confidence and certified products degree of confidence about degree of confidence; Though stipulate with first degree of confidence and second degree of confidence; But should only not be directed against unacceptable product and certified products, and should and judge purpose etc., stipulate from multiple viewpoint to different judgement objects.Such as; Use the position of microscopic examination life entity; The situation of identification given shape, detect situation that the special shape insect is arranged in getting out of the wood, in the image that obtains specific shape whether exist, the thing of given shape exist what or the like all need to judge, adopt compound these to judge the degree of confidence of demands.That is to say that the unacceptable product as above-mentioned, certified products will calculate degree of confidence from other viewpoint.Also have, can not only use a degree of confidence (only drawing), or two degree of confidence (drawing from unacceptable product degree of confidence and certified products degree of confidence) judge, and the degree of confidence of using more than three is judged from the unacceptable product degree of confidence.In addition, can not be confined to single image, should be able to judge single image or many images that synthetic processing obtains.
Claims (9)
1. image processing equipment, this image processing equipment can be judged the state of this object from shooting has the image that object obtained of certain form, it comprises:
Its image is taken and obtained to filming apparatus (1) to the object with certain form;
With reference to image storage unit (2), when judging the object state,, and store as the object of reference form parameter the related information parametersization of benchmark with reference to image based on object images;
Confidence computation unit (3) is represented with respect to the height of the approximation of reference substance form parameter the object form parameter with numerical value, calculate degree of confidence;
State judging unit (4) based on the degree of confidence that confidence computation unit is calculated, is judged the state of object.
2. image processing equipment according to claim 1; It is characterized in that: said with reference to also having object of reference form parameter extraction unit in the image storage unit; Take sample through said filming apparatus; With the information parameterization of the sample image that obtains, extract with reference to the said object of reference form parameter that should store in the image storage unit said.
3. image processing equipment according to claim 1 and 2; It is characterized in that: said confidence computation unit calculate respectively first degree of confidence and with this first degree of confidence various criterion under second degree of confidence, said state judging unit is judged the state of said object according to this first degree of confidence and second degree of confidence.
4. image processing equipment according to claim 3; It is characterized in that: said with reference to image storage unit, to photographed image-related information that I haven't seen you for ages will represent the unacceptable product state and the photographed image-related information of representing the certified products state one notes as said object of reference form parameter.
5. image processing equipment according to claim 4 is characterized in that: saidly both comprised the image information of reflection unacceptable product state with reference to image storage unit, and comprised the image information of reflection certified products state again; Forming of the unacceptable product that the image information of reflection unacceptable product state is obtained by only from unacceptable product, sampling with reference to image template; The image information of reflection certified products state is by constituting with reference to image template of certified products that only sampling is obtained from certified products.
6. image processing equipment according to claim 3 is characterized in that: when said state judging unit was the state of benchmark judgement object with two threshold values, first threshold value relevant with said degree of confidence was preferential judgment standard; When occurring, use second threshold value as benchmark than the little value of said first threshold value.
7. image processing equipment according to claim 3 is characterized in that: said object of reference form parameter comprises a parameter in formal parameter, color parameter, structural parameters and the size parameter at least.
8. image processing equipment according to claim 3 is characterized in that: this image processing equipment also comprises unit, and this unit is saidly learnt with reference to image information with reference to said in the image storage unit through updating stored in.
9. image processing equipment according to claim 8 is characterized in that: said unit is from surpass the divergence of setting with reference to deletion in the said object of reference form parameter that comprises the image information, as the criterion of state of the same type.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201220195299 CN202584162U (en) | 2012-05-04 | 2012-05-04 | Image processing device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201220195299 CN202584162U (en) | 2012-05-04 | 2012-05-04 | Image processing device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN202584162U true CN202584162U (en) | 2012-12-05 |
Family
ID=47253553
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN 201220195299 Expired - Fee Related CN202584162U (en) | 2012-05-04 | 2012-05-04 | Image processing device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN202584162U (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103606160A (en) * | 2013-12-02 | 2014-02-26 | 苏州比特速浪电子科技有限公司 | Image processing device and image processing method |
-
2012
- 2012-05-04 CN CN 201220195299 patent/CN202584162U/en not_active Expired - Fee Related
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103606160A (en) * | 2013-12-02 | 2014-02-26 | 苏州比特速浪电子科技有限公司 | Image processing device and image processing method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103383774A (en) | Image processing method and device using same | |
CN109447169B (en) | Image processing method, training method and device of model thereof and electronic system | |
CN111353413A (en) | Low-missing-report-rate defect identification method for power transmission equipment | |
CN103544691A (en) | Image processing method and unit | |
JP6197659B2 (en) | Detection control device, program, and detection system | |
CN113962274B (en) | Abnormity identification method and device, electronic equipment and storage medium | |
CN110909640A (en) | Method and device for determining water level line, storage medium and electronic device | |
CN112819821A (en) | Cell nucleus image detection method | |
CN115131283A (en) | Defect detection and model training method, device, equipment and medium for target object | |
CN114548208A (en) | Improved plant seed real-time classification detection method based on YOLOv5 | |
CN112036400A (en) | Method for constructing network for target detection and target detection method and system | |
CN111126264A (en) | Image processing method, device, equipment and storage medium | |
CN114841920A (en) | Flame identification method and device based on image processing and electronic equipment | |
CN114511820A (en) | Goods shelf commodity detection method and device, computer equipment and storage medium | |
CN106651803B (en) | Method and device for identifying house type data | |
CN113378864B (en) | Method, device and equipment for determining anchor frame parameters and readable storage medium | |
CN111275616A (en) | Low-altitude aerial image splicing method and device | |
CN202584162U (en) | Image processing device | |
CN114067314A (en) | Neural network-based peanut mildew identification method and system | |
CN113034446A (en) | Automatic transformer substation equipment defect identification method and system | |
CN117523395A (en) | Building earthquake damage information extraction method, equipment and medium based on remote sensing image | |
CN116363344A (en) | Wheat spike counting method and system based on improved YOLOv5s | |
CN113160414B (en) | Automatic goods allowance recognition method, device, electronic equipment and computer readable medium | |
CN112257710B (en) | Picture gradient detection method and device with text plane | |
CN114511862A (en) | Form identification method and device and electronic equipment |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20121205 Termination date: 20150504 |
|
EXPY | Termination of patent right or utility model |