CN112307827A - Object recognition apparatus, system and method - Google Patents
Object recognition apparatus, system and method Download PDFInfo
- Publication number
- CN112307827A CN112307827A CN201910700040.XA CN201910700040A CN112307827A CN 112307827 A CN112307827 A CN 112307827A CN 201910700040 A CN201910700040 A CN 201910700040A CN 112307827 A CN112307827 A CN 112307827A
- Authority
- CN
- China
- Prior art keywords
- error
- weighing
- counting
- type
- screening
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 25
- 238000005303 weighing Methods 0.000 claims abstract description 150
- 238000012216 screening Methods 0.000 claims abstract description 60
- 238000009826 distribution Methods 0.000 claims abstract description 44
- 238000004364 calculation method Methods 0.000 claims abstract description 13
- 238000003860 storage Methods 0.000 claims description 13
- 238000012545 processing Methods 0.000 description 27
- 238000001914 filtration Methods 0.000 description 15
- 238000004519 manufacturing process Methods 0.000 description 14
- 238000005259 measurement Methods 0.000 description 7
- 229910000975 Carbon steel Inorganic materials 0.000 description 6
- 239000010962 carbon steel Substances 0.000 description 6
- 238000012795 verification Methods 0.000 description 6
- 238000005070 sampling Methods 0.000 description 4
- 238000004806 packaging method and process Methods 0.000 description 3
- 238000007619 statistical method Methods 0.000 description 3
- BQCADISMDOOEFD-UHFFFAOYSA-N Silver Chemical compound [Ag] BQCADISMDOOEFD-UHFFFAOYSA-N 0.000 description 2
- 206010049040 Weight fluctuation Diseases 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 229910052709 silver Inorganic materials 0.000 description 2
- 239000004332 silver Substances 0.000 description 2
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24317—Piecewise classification, i.e. whereby each classification requires several discriminant rules
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
Abstract
The invention provides an object recognition device, a system and a method, wherein the object recognition device comprises: the first image matching module is used for comparing and screening first image characteristic information in the image characteristic information of each type of object with first image characteristics of the target object; the counting module is used for weighing and counting the target objects one by one according to the mass distribution information of each type of object; the calculation module is used for calculating the counting errors allowed by the objects of each type due to the statistical errors one by one according to the mass distribution information of the objects of each type and the corresponding weighing counting data; and the judging module is used for screening by judging whether the weighing counting data are within the range of the counting error allowed by the counting error of the corresponding type of object one by one. By the object recognition device, the object recognition system and the object recognition method, the object recognition efficiency can be improved.
Description
Technical Field
The present invention relates to the field of identification technologies, and in particular, to an object identification apparatus, system, and method.
Background
Due to the need for production line production multiplexing, in many applications, users need to use weighing and counting for different types of products, so that a weighing and counting system is required to identify the type of the product to be weighed, so that multiple types of products can be weighed and counted on one set of equipment. In warehouse management, when a specific amount of materials needs to be called from a warehouse for production and processing or logistics transportation of inventory products, the required materials or product types need to be identified first and then called subsequently.
Machine vision is an effective solution for identifying the type of product being weighed. The device for machine identification generally comprises a camera, a data processing system and the like. And the data processing system completes the identification of the product type according to the image data acquired by the camera. After the type identification is completed, subsequent processing of the product, such as weighing, packaging, transferring, etc., is performed.
However, as the types of products increase, the complexity of product identification through machine vision may lead to an exponential increase in the workload of processors in data processing systems, so that the reliability and speed of identification may not meet the requirements of practical applications. In this regard, existing solutions typically focus on enhancing recognition reliability through image algorithms, or increasing hardware device investment to increase the operating speed of the device. However, due to the complexity of the scheme or the increase of the implementation cost, the scheme is often not commercialized, thereby affecting the market promotion of the product.
Disclosure of Invention
The invention aims to solve the technical problem of the existing mechanical identification and provides an object identification device capable of improving the article identification efficiency.
In order to solve the above technical problem, the present invention provides an object recognition apparatus, including: the first image matching module is used for comparing and screening the first image characteristic information in the image characteristic information of each type of object with the first image characteristic of the target object; the counting module is used for weighing and counting the target objects one by one according to the mass distribution information of each type of object; the calculation module is used for calculating the counting errors allowed by the objects of each type due to the statistical errors one by one according to the mass distribution information of the objects of each type and the corresponding weighing counting data; and the judging module is used for screening by judging whether the weighing counting data are within the range of the counting error allowed by the counting error of the corresponding type of object one by one.
Preferably, the count error allowed by the statistical error is:
wherein K is a preset confidence factor index sigmanStandard deviation obtained from the mass distribution of a class of objects, NnIs a weighing count, AW, of a class of objectsnIs a weighted average obtained from the mass distribution of a class of objects.
Preferably, the calculation module is further configured to calculate an allowable counting error of the weighing error according to the weighing precision error of the weighing device and the weighing average value of each type of object; the judging module is used for screening by judging whether the weighing counting data are within the allowable counting error range of the counting error of the corresponding type of object and the allowable counting error range of the weighing error one by one; the allowable counting error of the weighing error is as follows: +/-C/AWn(ii) a Wherein C is the weighing precision error of the weighing equipment, AWnThe weight average value of the class of objects is obtained according to the mass distribution information.
Preferably, the judgment module calculates: whether the weighing count + ± (the count error allowed by the statistical error + the allowed count error of the weighing error) covers at least one integer; if yes, the weighing counting data is judged to be within the range of the allowed counting error corresponding to the statistical error of the type of the object, otherwise, the weighing counting data is judged to be beyond the range of the allowed counting error.
Preferably, the object recognition apparatus further includes: the second image matching module is used for performing next comparison and screening on second image characteristic information in the image characteristic information of each type of object and second image characteristics of the target object after the judgment module finishes front wheel screening; or, performing next comparison and screening on the first image characteristic information in the image characteristic information of each type of object and the first image characteristic of the target object through enhanced identification.
In order to solve the above technical problem, the present invention also discloses an object recognition system, comprising: the object recognition device, the storage device, the image acquisition device and the weighing device; the storage device is used for storing image characteristic information and quality distribution information of various types of objects; the image acquisition device is used for acquiring an image of the target object; the weighing device is used for weighing the target object.
In order to solve the above technical problem, the present invention also discloses an object identification method, including: comparing and screening first image characteristic information in the image characteristic information of each type of object with the first image characteristic of the target object; weighing and counting the target objects one by one according to the mass distribution information of each type of object; calculating the counting error allowed by each type of object due to the statistical error one by one according to the mass distribution information of each type of object and the corresponding weighing counting data; and screening by judging whether the weighing counting data are within the range of the counting error allowed by the counting error of the corresponding type of object one by one.
Preferably, the count error allowed by the statistical error is:
wherein K is a preset confidence factor index sigmanStandard deviation obtained from the mass distribution of a class of objects, NnWeighing objects of a kindCounting, AWnIs a weighted average obtained from the mass distribution of a class of objects.
Preferably, after calculating the counting errors allowed by the statistical errors of the objects of each type one by one according to the mass distribution information of the objects of each type and the corresponding weighing counting data, the method further comprises: calculating the allowable counting error of the weighing error according to the weighing precision error of the weighing equipment and the weighing average value of each type of object; the screening by judging whether the weighing counting data are within the range of the counting error allowed by the counting error of the corresponding type of object one by one comprises: screening by judging whether the weighing counting data are within the allowable counting error of the statistical error of the corresponding type of object and the allowable counting error range of the weighing error one by one; the allowable counting error of the weighing error is as follows: +/-C/AWn(ii) a Wherein C is the weighing precision error of the weighing equipment, AWnThe weight average value of the class of objects is obtained according to the mass distribution information.
Preferably, the determining whether the weighing count data is within a range of a count error allowed by the statistical error of the corresponding type of object includes: and (3) calculating: whether the weighing count + ± (the count error allowed by the statistical error + the allowed count error of the weighing error) covers at least one integer; if yes, the weighing counting data is judged to be within the range of the allowed counting error corresponding to the statistical error of the type of the object, otherwise, the weighing counting data is judged to be beyond the range of the allowed counting error.
Preferably, after the screening, the method further includes the following steps of determining whether the weighing and counting data are within a range of a counting error allowed by the counting error of the corresponding type of object one by one: performing next comparison and screening on second image characteristic information in the image characteristic information of each type of object and second image characteristics of the target object; or, performing next comparison and screening on the first image characteristic information in the image characteristic information of each type of object and the first image characteristic of the target object through enhanced identification.
The positive progress effects of the invention are as follows: by combining two screening modes of image feature matching and weighing counting, on one hand, the image matching range is reduced by performing matching screening according to partial image features, the system consumption of image calculation can be reduced, and the identification efficiency is improved; on the other hand, the reliability of identification is also ensured and more comprehensive reliability judgment data is provided according to whether the weighing and counting result is verified within the allowable counting error range.
Certain terms are used throughout the description and claims to refer to particular system components. As one skilled in the art will appreciate, different usage objects may represent the same component by different names. Components that differ in name but not function are not distinguished herein and are intended to be within the scope of the present invention.
Drawings
Fig. 1 is a schematic structural diagram of an object recognition system according to an embodiment of the present invention;
FIG. 2 is a schematic structural diagram of another object recognition system according to an embodiment of the present invention;
fig. 3 is a flowchart of an object recognition method according to an embodiment of the present invention.
Detailed Description
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in detail below.
Example 1
An object identification system in the automatic warehouse goods sorting system is used for identifying the types of the components. As shown in fig. 1, the object recognition system includes an object recognition data processing device, a storage device, a scale, a camera, and an input receiving device. The storage device is pre-stored with an industrial component library, which includes type codes of each component, weighing precision data of the scale, image information corresponding to each component, such as matching identification features of color identification feature shapes, textures and the like, and corresponding mass distribution information.
When a certain number of related types of components are selected from a warehouse according to the use requirement of production operation, such as 1000 carbon steel T-shaped bolts, the object recognition system firstly obtains the selection of a user on a required target component through a UI (user interface) of an input receiving device, and then obtains the image and quality distribution information of the target component through the type information of the target component in an industrial component library. Through the image of the target component, various image characteristic information such as corresponding color characteristics, texture characteristics, shape characteristics and the like can be obtained. The quality distribution information is statistical data information formed according to the weight data distribution condition of the component collected in the production process, and reflects the production weight fluctuation of the component.
The automatic warehouse goods sorting system grabs a plurality of quantities of all types of components from the warehouse, and the object recognition system shoots images of all types of components which are grabbed and taken out through the camera. The object identification data processing device detects the RGB information of the pixels within a certain distance range of a specific image sampling point by selecting a color matching mode with a small computation amount of partial image features through the first image matching module, and judges whether the color information within the distance range can be matched with the color features of the target component. For example, if the color of the carbon steel T-shaped bolt is bright silver, whether the color feature in the specific range meets the RGB value of the bright silver standard or not can be detected within a certain range of the sampling point, so that part of component types are firstly excluded by color screening of image information.
And after finishing the primary color screening, the object identification data processing device carries out the next weighing, counting and screening. The object recognition system respectively captures a plurality of residual components of each type for weighing to obtain the total weighing weight W of the components1，W2，W3……Wn. Meanwhile, a counting module in the object identification data processing device obtains a weighing mean value AW of the target component according to the mass distribution information of the target component, so that the total weight W of the target component is weighed1，W2，W3……WnRespectively dividing the obtained values by the weighing mean value AW of the target component to obtain weighing counting values N1，N2，N3……Nn. The counting module is used for weighing and counting a value N1，N2，N3……NnThe decimal is reserved and the rounding is not performed.
Because the mass distribution data is only statistical analysis data, the components of the same type can be mixed with the weighing mean AW in productionnThe calculation module of the object identification data processing device obtains the statistical standard deviation sigma according to the quality distribution information of the target component due to a certain objective errorn(ii) a Meanwhile, according to a preset credibility multiple index K, the standard deviation sigma is multiplied by the credibility multiple index KxnObtaining the error interval +/-K sigma of a single componentn. Then, by calculating the error interval x the weighing count value NnI.e. + -. K σnNnObtaining error intervals of a plurality of the components, and finally obtaining the error intervals of the components
Resulting in a count error that is allowed due to statistical errors.
Meanwhile, considering that a certain device measurement error exists in the weighing of the scale serving as the weighing device, the calculation module obtains the weighing precision error C of the scale according to the specification parameters of the scale in the storage device, for example, the precision error C of the scale is 0.05g, namely ± 0.05g, so as to further calculate the allowable counting error due to the weighing error of the weighing device itself as: +/-C/AWn。
The judgment module of the object identification data processing device calculates the following formula: weighing and counting NnWhether the counting error of the component is within the allowable range is judged by whether the counting error of the component is covered by at least one integer or not, so that whether the component of the type is possible to be the target component or not is identified, namely, calculation is carried out
Whether the resulting range of (a) covers at least one integer. If the leftmost NnAfter the integer number is calculated by the above formula, the integer number is not changed, and the judgment module of the object identification data processing device judges that the number is overIf the allowable error range is out, the identification fails, and the component of the type is removed from the range of the identification type, so that the identification range is further narrowed through weighing, counting and screening; otherwise, judging that the screening is passed. Since the weighted mean AW and the standard deviation σ are both attributes of the component, and the weighted total weight W is obtained from actual measurement, if the weighted total weight W is the correct value for the component to be formed, NnThe range of values of (a) must include integers, if not, this indicates that it is not necessarily a component of that type.
The technical principle of counting and screening in the embodiment is as follows: in the process of comparing the target components with the various types of components one by one, the total weight W of the components is verified according to the current weightnCombined with the weight average AW of the target componentnAnd standard deviation sigma of target componentnSubstituted into judgment weighing and counting NnVerification is performed + - (statistical error allowed count error + weighing error allowed count error). If the current verification component and the target component do not belong to one type, the counting precision of the current verification component and the target component is poor and does not conform to the allowable range of counting errors, so that counting matching screening is completed.
After completing one round of screening of color matching screening and weighing counting, if two or more types of objects are not excluded, the object identification data processing device identifies the types of the components which are not excluded after screening again for the image information of the next round. In this case, since the number of types of the remaining components is small after the first round of screening, an image matching method with high matching accuracy and a large amount of computation can be selected. And the second image matching module of the object identification data processing device further reduces the distance range of the sampling points in color identification for more accurate matching through image enhancement identification, or calculates information such as shape/texture and the like in the image of the component, performs matching identification with the information such as the shape/texture and the like of the target type component, and screens out the component types which cannot be matched.
And finally, if the object identification system identifies and matches the only component type, the identification and matching of the target component are completed, and the information of the type and the number of the identified components is output. If the object recognition system recognizes a plurality of component types and cannot be further excluded, or does not recognize component types that can be matched, recognition error information is output.
It is understood that the order of performing the image matching filtering and the counting filtering in this embodiment may also be to perform the counting filtering first, then perform the color filtering by the first image matching module, perform the filtering of the color enhancement recognition by the second image matching module, or perform the recognition of other image features.
In the object recognition system of embodiment 1, two recognition methods, namely, image recognition and weighing counting recognition, are combined, so that on one hand, compared with the conventional mechanical visual recognition, the image matching range is reduced through weighing counting recognition, the workload and the resource consumption of a data processing system are reduced, and the recognition speed is ensured; on the other hand, by combining the two identification modes, the accuracy and the reliability of identification can be ensured.
Example 2
Before receiving the product, the packaging production line identifies the corresponding product type through the object identification system, and then performs subsequent sorting, such as counting and packaging. As shown in fig. 2, the object recognition system includes an object recognition data processing device, a camera, a scale, and a storage device. The storage device is pre-stored with the type codes of all the products that the production line may produce, the weighing precision data of the scale, the image information corresponding to each product, such as the matching identification features of color identification feature shape, texture, etc., and the corresponding mass distribution information. The mass distribution information is statistical data information formed according to the weight data distribution condition of each type of product collected from the production process, and reflects the production weight fluctuation of each type of product.
When the production line finishes the production of the products, a plurality of products are transmitted to the object recognition system and are shot by the camera. The image information of the product can be obtained through the image of the product, for example, the product is a carbon steel T-shaped bolt, the color feature of the image is sampled and identified through a first image matching module of the object identification data processing device by selecting a color matching mode with a small operation amount relative to image matching, and a part of product types can be firstly excluded through screening the color information of various types of products stored in the storage device.
And after finishing the primary color screening, the object identification data processing device carries out the next weighing, counting and screening. Firstly, the object identification system weighs the carbon steel T-shaped bolt through a scale to obtain the weighed total weight W of the carbon steel T-shaped bolt. Meanwhile, a counting module in the object identification data processing device obtains the weighing average value AW of each type of product according to the mass distribution information of each type of product in the storage device1，AW2，AW3… …, dividing the total weight W of the carbon steel T-bolts by the average AW of each type of product1，AW2，AW3… …, obtaining respective corresponding weighing and counting values N1，N2，N3… … are provided. The counting module is used for weighing and counting a value N1，N2，N3… … the decimal is kept and is not rounded.
Since the mass distribution data are only statistical analysis data, the same type of product will be produced with the weighing mean AWnThere is a certain objective error, so the calculation module of the object identification data processing device obtains the statistical standard deviation sigma according to the quality distribution information of the target productn(ii) a Meanwhile, according to a preset credibility multiple index K, the standard deviation sigma is multiplied by the credibility multiple index KxnAnd obtaining the error interval +/-K sigma of a single product. Then, by calculating the error interval x the weighing count value NnI.e. + -. K σ NnObtaining error intervals of a plurality of the products to be grabbed, and finally, obtaining the error intervals of the products to be grabbed
Resulting in a count error that is allowed due to statistical errors.
Meanwhile, the calculation module stores the error according to the fact that the weighing scale serving as the weighing equipment has certain equipment measurement error in the process of measuring the weightIn the device, the specification parameters of the scale obtain the weighing precision error C of the scale, for example, the precision error C of the scale is 0.05g, namely +/-0.05 g, so that the allowable counting error of the weighing equipment is further calculated as: +/-C/AWn。
The judgment module of the object identification data processing device calculates the following formula: weighing and counting Nn+ (allowable counting error for statistical error + allowable counting error for weighing error), whether or not to cover at least one integer to determine whether or not its counting error is within an allowable range, thereby identifying whether or not the type of product is likely to be a target product, i.e., calculating
Whether the resulting range of (a) covers at least one integer. If the leftmost NnAfter the integer number is calculated by the formula, the integer number is not changed, a judgment module of the object identification data processing device judges that the error range is beyond the allowable error range, the identification fails, and products of the type are removed from the range of the identification type, so that the identification range is further reduced through weighing, counting and screening; otherwise, judging that the screening is passed. Due to the mean value AW and standard deviation sigmanAll belonging to the product, the weighed total weight W being obtained according to actual measurements, if the weighed total weight W is the correct value for the product to be formed, NnMust include integers, and if no integers are present, this indicates that it is not necessarily a product of that type.
The technical principle of counting and screening in the embodiment is as follows: in the process of comparing the target product with various types of products one by one, the weight average AW of the currently verified product is combined with the total weight W of the target productnAnd the standard deviation sigma of the currently verified productnSubstituted into judgment weighing and counting NnVerification is performed + - (statistical error allowed count error + weighing error allowed count error). If the current verification product and the target product do not belong to one type, the counting precision of the current verification product and the target product is poor, and counting matching screening is finished.
After completing one round of screening of color matching screening and weighing counting, if two or more types of objects are not excluded, the object identification data processing device re-identifies the image information of the next round for the product types which are not excluded after screening. In this case, since the number of types of products remaining after the first round of screening is small, an image matching method with high matching accuracy and a large amount of computation can be selected. And the second image matching module of the object identification data processing device further reduces the distance range of the sampling points in color identification for more accurate matching through image enhancement identification, or calculates information such as shape/texture and the like in the product image, performs matching identification with the information such as the shape/texture and the like of a target type product, and screens out the product types which cannot be matched.
And finally, matching the unique product type by the object identification system, completing the identification and matching of the product, and outputting the product type. If the object recognition system recognizes a plurality of product types and cannot be further excluded or does not recognize a matching product type, recognition error information is output.
It is understood that the order of performing the image matching filtering and the counting filtering in this embodiment may also be to perform the counting filtering first, then perform the color filtering by the first image matching module, perform the filtering of the color enhancement recognition by the second image matching module, or perform the recognition of other image features.
Example 3
An object recognition method of the present embodiment is applied to the execution of the object recognition data processing apparatus of the above-described embodiment 1 or 2 to perform screening recognition of a target object. As shown in fig. 3, the object recognition method may include:
step S101, matching the color features of the target object image with the color features of each type of object.
After the pre-stored target object image is obtained by shooting and acquiring the image of the target object through the camera or receiving the target object type information input by an operator, a color matching mode with a small operation amount of partial image features is selected, the color features of the target object image are matched with the color features of various types of objects stored in the storage device, and color-based identification and screening are completed.
And step S102, weighing and counting the target objects one by one according to the mass distribution information of each type of object.
If the type of the target object is designated by an operator, respectively grabbing a plurality of the remaining objects of each type for weighing, and obtaining the total weighing weight W of the objects1，W2，W3……Wn. Meanwhile, according to the mass distribution information of the target object, the weighing average value AW of the target object is obtained, so that the total weight W is weighed1，W2，W3……WnRespectively dividing the obtained values by the weighing average value AW of the target object to obtain weighing counting values N1，N2，N3……Nn. Weighing scale value N1，N2，N3……NnThe decimal is reserved and the rounding is not performed.
If the type of the target object is unknown, the target object is weighed by a scale first, and the weighed total weight W of the target object is obtained. Meanwhile, the weighing average value AW of each type of product is obtained according to the mass distribution information of each type of object in the storage device1，AW2，AW3… …, so that the weighted total weight W of the target object is divided by the weighted average AW of each type of object1，AW2，AW3… …, obtaining respective corresponding weighing and counting values N1，N2，N3… … are provided. Weighing scale value N1，N2，N3… … the decimal is kept and is not rounded.
Step S103, calculating the allowable counting error of each type of object due to the statistical error.
Since the mass distribution data is only statistical analysis data, objects of the same type will be in production with the weighing mean AWnThe calculation module of the object identification data processing device obtains the statistical standard deviation sigma according to the quality distribution information of the target component due to a certain objective errorn(ii) a Meanwhile, according to a preset credibility multiple index K, the standard deviation sigma is multiplied by the credibility multiple index KxnObtaining the error of a single objectInterval + -K sigman. Then, by calculating the error interval x the weighing count value NnI.e. + -. K σnNnObtaining error intervals of a plurality of captured objects, and finally obtaining the error intervals of the captured objects
Resulting in a count error that is allowed due to statistical errors.
Step S104, calculating the allowable counting error of the weighing error.
Similarly, since the weighing device itself will have a certain device measurement error during the weight measurement, the weighing precision error C of the scale is obtained according to the specification parameters of the scale, so as to further calculate the allowable counting error due to the weighing error of the weighing device itself as follows: +/-C/AWn。
Here, step S103 and step S104 are not absolutely sequential in order of implementation.
And step S105, judging whether the weighing counting data is within the allowable counting error range of the corresponding type of object.
By the calculation formula: weighing and counting Nn+ (allowable counting error for statistical error + allowable counting error for weighing error), whether to cover at least one integer to determine whether the counting error is within an allowable range, thereby identifying whether the type of object is likely to be the target object. I.e. calculating
Whether the resulting range of (a) covers at least one integer. If the leftmost NnAfter the integer number is calculated by the formula, the integer number is not changed, a judgment module of the object identification data processing device judges that the error range is beyond the allowable error range, the identification fails, and the component of the type is removed from the range of the identification type, so that the identification range is further reduced by weighing, counting and screening; otherwise, judging that the screening is passed. Since the weighted mean AW and the standard deviation σ are both attributes of the object, the weighted total weight W is obtained from actual measurement, and if the weighted total weight W is a correct value for the component to be formed，NnThe range of values of (a) must include integers, if not, this indicates that it is not necessarily a component of that type.
And step S106, if two or more types of objects are not excluded, performing the next round of image matching screening until the matching is completed, or outputting identification error information.
If more than two types of objects are not excluded, the identification of the image information of the next round is carried out again on the object types which are not excluded after the screening. At this time, since the number of the remaining object types is small after the first round of screening, an image matching method with high matching accuracy and large computation amount can be selected. Through image enhancement and recognition, more accurate color matching is adopted, or other image characteristics in the object image, such as shape/texture and other information, are calculated, matched and recognized with the target object, and screening is carried out.
Although the idea of the invention has been described by means of the above specific embodiments, it will be appreciated by those skilled in the art that further modifications are conceivable on the basis of the teaching of the invention, for example by combining features of the embodiments with each other and/or interchanging functional units between the embodiments. Therefore, all fall within the scope of protection of the present application.
It is understood that, in the present embodiment, the order of performing the image matching filtering of step S101 and the counting filtering of steps S102 to S105 may also be performing the counting filtering first, then performing the color filtering, and then performing the filtering of the color enhancement recognition, or recognizing other image features.
Those skilled in the art will appreciate that all or part of the steps in the methods of the above embodiments may be implemented by hardware related to instructions of a program, which may be stored on a computer readable medium, which may include: read Only Memory (ROM), Random Access Memory (RAM), magnetic or optical disks, and the like.
While the invention has been described with reference to a number of specific embodiments, it will be understood by those skilled in the art that the foregoing embodiments are merely illustrative of the invention, and various changes in and substitutions of equivalents may be made without departing from the spirit of the invention. Therefore, changes and modifications to the above-described embodiments within the spirit and scope of the present invention will fall within the scope of the claims of the present application.
Claims (11)
1. An object recognition apparatus, comprising:
the first image matching module is used for comparing and screening the first image characteristic information in the image characteristic information of each type of object with the first image characteristic of the target object;
the counting module is used for weighing and counting the target objects one by one according to the mass distribution information of each type of object;
the calculation module is used for calculating the counting errors allowed by the objects of each type due to the statistical errors one by one according to the mass distribution information of the objects of each type and the corresponding weighing counting data;
and the judging module is used for screening by judging whether the weighing counting data are within the range of the counting error allowed by the counting error of the corresponding type of object one by one.
2. The object recognition apparatus of claim 1, wherein the count error allowed due to the statistical error is:
wherein K is a preset confidence factor index sigmanStandard deviation obtained from the mass distribution of a class of objects, NnIs a weighing count, AW, of a class of objectsnIs a weighted average obtained from the mass distribution of a class of objects.
3. The object recognition apparatus of claim 1,
the calculation module is also used for calculating the allowable counting error of the weighing error according to the weighing precision error of the weighing equipment and the weighing average value of each type of object;
the judging module is used for screening by judging whether the weighing counting data are within the allowable counting error range of the counting error of the corresponding type of object and the allowable counting error range of the weighing error one by one;
the allowable counting error of the weighing error is as follows: +/-C/AWn(ii) a Wherein C is the weighing precision error of the weighing equipment, AWnThe weight average value of the class of objects is obtained according to the mass distribution information.
4. The object recognition apparatus of claim 3, wherein the determination module calculates: whether the weighing count + ± (the count error allowed by the statistical error + the allowed count error of the weighing error) covers at least one integer; if yes, the weighing counting data is judged to be within the range of the allowed counting error corresponding to the statistical error of the type of the object, otherwise, the weighing counting data is judged to be beyond the range of the allowed counting error.
5. The object recognition apparatus of claim 1, further comprising:
the second image matching module is used for performing next comparison and screening on second image characteristic information in the image characteristic information of each type of object and second image characteristics of the target object after the judgment module finishes front wheel screening;
or, performing next comparison and screening on the first image characteristic information in the image characteristic information of each type of object and the first image characteristic of the target object through enhanced identification.
6. An object recognition system, comprising: an object recognition arrangement according to any one of claims 1 to 5, storage means, image acquisition means and weighing means;
the storage device is used for storing image characteristic information and quality distribution information of various types of objects;
the image acquisition device is used for acquiring an image of the target object;
the weighing device is used for weighing the target object.
7. An object recognition method, comprising:
comparing and screening first image characteristic information in the image characteristic information of each type of object with the first image characteristic of the target object;
weighing and counting the target objects one by one according to the mass distribution information of each type of object;
calculating the counting error allowed by each type of object due to the statistical error one by one according to the mass distribution information of each type of object and the corresponding weighing counting data;
and screening by judging whether the weighing counting data are within the range of the counting error allowed by the counting error of the corresponding type of object one by one.
8. The object recognition method of claim 7, wherein the count error allowed due to the statistical error is:
wherein K is a preset confidence factor index sigmanStandard deviation obtained from the mass distribution of a class of objects, NnIs a weighing count, AW, of a class of objectsnIs a weighted average obtained from the mass distribution of a class of objects.
9. The object recognition method of claim 7,
after calculating the counting errors allowed by the counting errors of the objects of each type one by one according to the mass distribution information of the objects of each type and the corresponding weighing counting data, the method further comprises the following steps: calculating the allowable counting error of the weighing error according to the weighing precision error of the weighing equipment and the weighing average value of each type of object;
the screening by judging whether the weighing counting data are within the range of the counting error allowed by the counting error of the corresponding type of object one by one comprises: screening by judging whether the weighing counting data are within the allowable counting error of the statistical error of the corresponding type of object and the allowable counting error range of the weighing error one by one;
the allowable counting error of the weighing error is as follows: +/-C/AWn(ii) a Wherein C is the weighing precision error of the weighing equipment, AWnThe weight average value of the class of objects is obtained according to the mass distribution information.
10. The object recognition method of claim 9, wherein the determining whether the weighing count data is within a count error allowable for the statistical error for the type of object comprises:
and (3) calculating: whether the weighing count + ± (the count error allowed by the statistical error + the allowed count error of the weighing error) covers at least one integer; if yes, the weighing counting data is judged to be within the range of the allowed counting error corresponding to the statistical error of the type of the object, otherwise, the weighing counting data is judged to be beyond the range of the allowed counting error.
11. The object recognition method according to claim 7, wherein after the screening by determining one by one whether the weighing count data is within a range of a count error allowed by the statistical error of the corresponding type of object, further comprising:
performing next comparison and screening on second image characteristic information in the image characteristic information of each type of object and second image characteristics of the target object; or, performing next comparison and screening on the first image characteristic information in the image characteristic information of each type of object and the first image characteristic of the target object through enhanced identification.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910700040.XA CN112307827B (en) | 2019-07-31 | Object recognition apparatus, system and method | |
| PCT/CN2020/101598 WO2021017796A1 (en) | 2019-07-31 | 2020-07-13 | Object recognition apparatus, system and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910700040.XA CN112307827B (en) | 2019-07-31 | Object recognition apparatus, system and method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112307827A true CN112307827A (en) | 2021-02-02 |
| CN112307827B CN112307827B (en) | 2024-04-26 |
Family
ID=
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113407571A (en) * | 2021-06-30 | 2021-09-17 | 重庆博尔德医疗科技股份有限公司 | Counting method based on weighing principle |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070076956A1 (en) * | 2005-10-04 | 2007-04-05 | Fujifilm Corporation | Automatic counting apparatus, method, and recording medium on which the program is recorded |
| CN102324042A (en) * | 2011-09-13 | 2012-01-18 | 盛乐信息技术（上海）有限公司 | Visual identifying system and visual identity method |
| JP2015106408A (en) * | 2013-12-02 | 2015-06-08 | 蘇州比特速浪電子科技有限公司 | Image processing apparatus and image processing method |
| CN108537994A (en) * | 2018-03-12 | 2018-09-14 | 深兰科技（上海）有限公司 | View-based access control model identifies and the intelligent commodity settlement system and method for weight induction technology |
| CN108765797A (en) * | 2018-05-25 | 2018-11-06 | 合肥市闵葵电力工程有限公司 | A kind of supermarket market automatic identification is weighed settlement method and system |
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070076956A1 (en) * | 2005-10-04 | 2007-04-05 | Fujifilm Corporation | Automatic counting apparatus, method, and recording medium on which the program is recorded |
| CN102324042A (en) * | 2011-09-13 | 2012-01-18 | 盛乐信息技术（上海）有限公司 | Visual identifying system and visual identity method |
| JP2015106408A (en) * | 2013-12-02 | 2015-06-08 | 蘇州比特速浪電子科技有限公司 | Image processing apparatus and image processing method |
| CN108537994A (en) * | 2018-03-12 | 2018-09-14 | 深兰科技（上海）有限公司 | View-based access control model identifies and the intelligent commodity settlement system and method for weight induction technology |
| CN108765797A (en) * | 2018-05-25 | 2018-11-06 | 合肥市闵葵电力工程有限公司 | A kind of supermarket market automatic identification is weighed settlement method and system |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113407571A (en) * | 2021-06-30 | 2021-09-17 | 重庆博尔德医疗科技股份有限公司 | Counting method based on weighing principle |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2021017796A1 (en) | 2021-02-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110517265B (en) | Method and device for detecting surface defects of product and storage medium | |
| CN113240630B (en) | Speckle image quality evaluation method and device, terminal equipment and readable storage medium | |
| EP4027311A1 (en) | Vision- and gravity sensing-based product identification method, device, and system | |
| JP6688962B2 (en) | Judgment device, judgment method, and judgment program | |
| CN110599675B (en) | Method and device for detecting number of paper money in money box, terminal equipment and storage medium | |
| CN110490516A (en) | A kind of unmanned store shelf article detection method and device | |
| WO2021017796A1 (en) | Object recognition apparatus, system and method | |
| JP2021510226A5 (en) | ||
| WO2021017797A1 (en) | Counting check device,counting system and method | |
| TWI591545B (en) | Fingerprint image detecting device and method thereof | |
| CN110795324A (en) | Data processing method and device | |
| CN112513946A (en) | Count the game currency | |
| CN107092772B (en) | Method and device for determining characteristic curve of sensor | |
| US10134122B2 (en) | System and method of distributed processing for machine-vision analysis | |
| CN112307827B (en) | Object recognition apparatus, system and method | |
| CN109064457B (en) | Quantity accumulation method and system based on vision | |
| CN111640096B (en) | Method, device and terminal for detecting appearance of electronic product | |
| CN113537196B (en) | Picture identification method and device, computer equipment and computer readable storage medium | |
| CN112304411A (en) | Weighing system with object recognition and weighing method | |
| CN115479904A (en) | Method and system for rapidly detecting feed tannin | |
| CN115147541A (en) | Ladle detection method and device, electronic equipment and computer readable storage medium | |
| KR101557421B1 (en) | Method and Apparatus for Identifying Money | |
| CN107389677B (en) | Method and device for detecting quality of flannelette fluff | |
| CN114820003A (en) | Pricing information abnormity identification method and device, electronic equipment and storage medium | |
| CN112598616A (en) | Electronic equipment exposure parameter determining method and imaging method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |