CN107622278A - A kind of patent image-recognizing method based on Android phone platform - Google Patents
A kind of patent image-recognizing method based on Android phone platform Download PDFInfo
- Publication number
- CN107622278A CN107622278A CN201710784529.0A CN201710784529A CN107622278A CN 107622278 A CN107622278 A CN 107622278A CN 201710784529 A CN201710784529 A CN 201710784529A CN 107622278 A CN107622278 A CN 107622278A
- Authority
- CN
- China
- Prior art keywords
- image
- histogram
- method based
- database
- android
- 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.)
- Pending
Links
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Image Analysis (AREA)
- Collating Specific Patterns (AREA)
Abstract
The present invention relates to a kind of patent image-recognizing method based on Android phone platform, it is characterised in that this method comprises the following steps:S01, Database;S02, patent IMAQ;S03, patent image compare.Compared with prior art, the beneficial effects of the invention are as follows:The structure of database is more reasonable, and data manipulation is easy, and improves the efficiency of load image;IMAQ mode is more flexible, easy to operate using the shoot function of terminal device or using patent image needed for extraction in photo library;Image comparison is more accurate, and using the mode for the histogram for calculating image, " fingerprint " of image is compared, and analyzes similarity, and precision is high, and corresponding rapid, so as to meet the use demand of design patent retrieval well.
Description
Technical field
A kind of patent retrieval field, more particularly to a kind of patent image-recognizing method based on Android phone platform.
Background technology
Patent is a pith in intellectual property protection, although Chinese patent numbers are sure to occupy first now,
It is to still suffer from the problem of certain in quality and patent application.Its Patent has three kinds:Patent of invention, patent of invention, outward appearance
Design patent.Because appearance design and invention cannot ensure without examination as to substances, the stability of its patent, and reliability.
And the technology retrieval before patent application, it is effective necessary condition for lifting patent reliability.Wherein design patent, its data
Storehouse information is huge, how fast and effectively to search targetedly design patent, is that current design patent application faces
A major issue.Current patent search system is mainly by the way of character search, and the screening for figure needs
Artificial to compare, this is resulted in, and retrieval difficulty is big, and input cost is high, and easily omits, and then causes, and retrieval result is not accurate.
The content of the invention
In view of the above-mentioned state of the art, the technical problems to be solved by the invention are that providing a kind of simplify operates,
The patent image-recognizing method based on Android phone platform of raising efficiency
Technical scheme is used by the present invention solves above-mentioned technical problem:It is a kind of based on the special of Android phone platform
Sharp image-recognizing method, this method comprise the following steps:
S01, Database:All patents are preserved using Android SQLite as application program local data base
Information, including author, general information and its corresponding product patent image, and establish database;
S02, patent IMAQ:Target patent image is obtained, the acquisition target patent image can utilize Android to grasp
The shoot function for making the intelligent terminal of system is shot or using patent image needed for extraction in photo library, and passes through mobile phone end
Wireless transport module is held by the patent image transmitting collected to apps server;
S03, patent image compare:The corresponding data on each patent label filtration will be obtained from database first
Collection, patent image is loaded from apps server, patent searching image is fetched by using chain using Glide, then called
Comparison algorithm is by target patent image compared with each candidate patent image;Meanwhile create and sorted based on similarity result
Priority query, store the ID of each patent image and its corresponding likelihood, and recognition result is sent to Android behaviour
Make the intelligent terminal of system.
Further, in the step S01, establish database and comprise the following steps:
S11, first, all images are uploaded on line server;
S12, then, the respective links of each image are stored into database.
Further, in the step S03, the comparison algorithm can use the Nogata for perceiving hash algorithm or calculating image
Figure.
Further, the perception hash algorithm, comprises the following steps:
S21, the size that picture is narrowed down to 8x8, make picture include 64 pixels;
S22, simplify picture color, convert the image into gray scale;
S23, the average gray value for calculating all 64 pixels, and the gray value of each pixel is compared with average value
Compared with if greater than average value, being then recorded as 1, if less than average value, be then recorded as 0;
S24, all values for recording to obtain of comparing are put into an array, this is image " fingerprint ", then passes through ratio
Compared with corresponding value in " fingerprint " of two images, it is the same to calculate how many position in 64, according to the number of identical digit,
Analyze picture acquaintance degree.
Further, the histogram for calculating image, comprises the following steps:
S31, firstly, since picture size is inconsistent, it is standard size to need to adjust picture before comparison;
S32, secondly, RGB image is converted into gray level;
S33, then, because the type of image is matrix, is converted to floating type, preferably to calculate figure by each pixel value
As histogram;
It is S34, last, compare two histograms, convert the result to percentage, and priority query is arrived into likelihood storage
In.
Further, in the step S33, for ease of the histogram of computing array, call in built-in OpenCV
CalcHist () function.
Further, in the step S34, for ease of comparing two histograms, call in built-in OpenCV
CompareHist () function, and select the standard of Pasteur's coefficient therein to be calculated, its calculation formula is as follows:
Wherein H1(I) gray scale of post is corresponded in first image histogram of expression;H2(I) second image histogram is represented
The gray scale of middle corresponding post;Represent the average gray of all posts in first image histogram;Represent second image Nogata
The average gray of all posts in figure;D represents the distance between two image histograms, and the value is reacted between two image histograms
Similitude, d values show that more greatly two image histograms are more similar.
Compared with prior art, the beneficial effects of the invention are as follows:The structure of database is more reasonable, and data manipulation is easy,
And improve the efficiency of load image;IMAQ mode is more flexible, is shone using the shoot function or utilization of terminal device
Patent image needed for extraction, easy to operate in valut;Image compare it is more accurate, using the mode for the histogram for calculating image,
" fingerprint " of image is compared, analyzes similarity, precision is high, and corresponding rapid, so as to meet appearance design well
The use demand of patent retrieval.
Embodiment
Below in conjunction with the embodiment of the present invention, it is the preferred embodiment in the embodiment of the present invention, rather than whole realities
Apply example.Based on the embodiment in the present invention, those of ordinary skill in the art are obtained under the premise of creative work is not made
Every other embodiment, belong to the scope of protection of the invention.
A kind of patent image-recognizing method based on Android phone platform of the present invention, by following three parts structure
Build and form:Part I is the early stage preparation on data lab setting;Part II is the realization of main Android activities;
Compare the main algorithm of similarity by explaining in last part.
First, the foundation of database
All patent information, including author are preserved using android SQLite as application program local data base,
General information and its corresponding product patent image.All patents are divided into two parts, and a part is related to contrast patent
Information, and another part is then the information unrelated with patent is contrasted, when the present invention establishes database, the present invention sets one
Label distinguishes them, if label is 1, it will be related information, and 0 be unrelated information.It will help the present invention with
All contrast patents are more effectively found in operation afterwards.In addition, in order to more easily manipulate database, utilize
Litepal, it is an Android storehouse of increasing income, it is allowed to is created in a simpler manner or upgrade form.
The subject matter run into when building database is how to store patent view data because this method determine as
What patent searching image and the efficiency by patent image transmitting to another patent view data form.
Therefore, the present invention uses:All patent images are uploaded on line server first, then by each patent figure
The respective links of picture are stored into database.In order to improve the efficiency of loading patent image, present invention utilizes Glide, and it is
The framework that one patent image for Android loads, this framework also effectively reduce the internal memory needed for application program.
2nd, Android Activity realization
The method that the present invention provides two kinds of acquisition target patent images, a kind of is that the camera for calling Android phone is taken pictures,
Another kind is that a kind of patent image is selected from photo library.Both modes can obtain target patent image and be transmitted
To lower one page.
Patent image comparison method of the present invention uses:It will be obtained first from database on each patent label filtration
Cup data set.As described above, loading patent image from server, the present invention fetches retrieval specially using Glide by using chain
Sharp image, then call comparison algorithm by target patent image compared with each candidate patent image.In order to store every time
Operation result, invention creates the priority query to be sorted based on similarity result, and store the ID of each patent image
And its corresponding likelihood.After handling all patent images, ten elements before the present invention takes out from priority query, Ran Hougen
All information are obtained according to ID, and using these data as grid view.
3rd, the explanation of comparison algorithm
1. perceive hash algorithm
Present invention selection first perceives hash algorithm, and it can be to generate one " fingerprint " per pictures
(fingerprint) fingerprint of character string, then more different pictures, comparative result is closer, just illustrates that picture is more similar.Tool
Gymnastics is made as follows:
The first step, picture is narrowed down to 8x8 size, picture is included 64 pixels.This step is to remove in picture
Details, only retain the essential information such as structure, caused by the same time it can also reduce object scaling and patent image size
Influence.Second step, simplify picture color.Patent image is converted to gray scale by this step.3rd, calculate the flat of all 64 pixels
Equal gray value, and by the gray value of each pixel compared with average value, if greater than average value, then 1 is recorded as, such as
Fruit is less than average value, then is recorded as 0.Finally, all values for recording to obtain of comparing are put into an array, this is patent image
" fingerprint ".The order of combination is inessential, as long as ensureing that all patent images use identical order.Then pass through and compare
Corresponding value in " fingerprint " of two patent images, it is the same to calculate how many position in 64.This in theory, equivalent to
Calculate " Hamming distance ", identical digit is more, it was demonstrated that picture is more similar.
The advantages of algorithm, is that it is not influenceed by patent image scaled, and is easily achieved.But due to calculating
Method mainly calculates similarity by comparing the pixel of two figures, so requirement of the algorithm to inquiring about patent image is higher, this
Mean to differ larger with artwork for inquiring about patent image.In view of the situation, the present invention is main in second algorithm
If comparing whether picture is similar by calculating the histogram of patent image, because the histogram of patent image is preferably from complete
Office describes the content of patent image.
2. histogram comparison method
This method calls some functions from OpenCV.OpenCV is the work(for being used for handling and analyzing digital patent image
The powerful Open Source Code storehouse of energy.The main theory of this algorithm is by calculating being averaged for the L2-Norm distances of each patent image
Value obtains the normalization histogram of each patent image, and will obtain and be compared between two histograms.Concrete operations are such as
Under:
Firstly, since patent picture size is inconsistent, so it is standard size to need to adjust picture before comparison.Its
It is secondary, RGB patent images are converted into gray level, because in this case, the edge of object is main source, and reduces face
Color information can also accelerate process.Then, because the type of patent image is matrix, the present invention is converted to each pixel value floating
Point-type, preferably to calculate patent image histogram.To achieve it, the present invention is called in built-in OpenCV
CalcHist () function, it can be by source patent image transmitting to required histogram.Final step is to compare two Nogatas
Figure, OpenCV also provide such algorithm.CompareHist () function provides the method for four kinds of calculating similarities, this hair
The bright standard that have selected Pasteur's coefficient therein calculates distance between each post.The distance describes two same grayscale ranks
Spatial diversity, the main formulas for obtaining distance is as follows:
Wherein H1(I) gray scale of post is corresponded in first image histogram of expression;H2(I) second image histogram is represented
The gray scale of middle corresponding post;Represent the average gray of all posts in first image histogram;Represent second image Nogata
The average gray of all posts in figure;D represents the distance between two image histograms, and the value is reacted between two image histograms
Similitude, d values show that more greatly two image histograms are more similar.
Formula respectively obtains the result of the average normalized histogram of two patent images first, obtain each post value and
Different sums between average result, then divided by each histogram standard variance.The distance describe two histograms it
Between correlation, while also reflects the similitude between patent image.Finally, the present invention converts the result to percentage, and
By likelihood storage into priority query.
The beneficial effects of the invention are as follows:The structure of database is more reasonable, and data manipulation is easy, and improves load image
Efficiency;IMAQ mode is more flexible, using special needed for extraction in the shoot function or utilization photo library of terminal device
Sharp image, it is easy to operate;Image comparison is more accurate, and using the mode for the histogram for calculating image, " fingerprint " of image is entered
Row compares, and analyzes similarity, and precision is high, and corresponding rapid, so as to meet the use need of design patent retrieval well
Ask.
It is obvious to a person skilled in the art that the invention is not restricted to the details of above-mentioned one exemplary embodiment, Er Qie
In the case of without departing substantially from spirit or essential attributes of the invention, the present invention can be realized in other specific forms.Therefore, no matter
From the point of view of which point, embodiment all should be regarded as exemplary, and be nonrestrictive, the scope of the present invention is by appended power
Profit requires rather than described above limits, it is intended that all in the implication and scope of the equivalency of claim by falling
Change is included in the present invention.
Claims (7)
1. a kind of patent image-recognizing method based on Android phone platform, it is characterised in that this method includes following step
Suddenly:
S01, Database:All patent letters are preserved using Android SQLite as application program local data base
Breath, including author, general information and its corresponding product patent image, and establish database;
S02, patent IMAQ:Target patent image is obtained, the acquisition target patent image can utilize Android operations system
The shoot function of the intelligent terminal of system shot or using in photo library extraction needed for patent image, and by mobile phone terminal without
Line transport module is by the patent image transmitting collected to apps server;
S03, patent image compare:The corresponding data collection on each patent label filtration will be obtained from database first, from
Apps server loads patent image, and patent searching image is fetched by using chain using Glide, then calls and compares calculation
Method is by target patent image compared with each candidate patent image;Meanwhile create based on the preferential of similarity result sequence
Level queue, stores the ID of each patent image and its corresponding likelihood, and recognition result is sent into Android operation system
Intelligent terminal.
2. a kind of patent image-recognizing method based on Android phone platform according to claim 1, its feature exist
In in the step S01, establishing database and comprise the following steps:
S11, first, all images are uploaded on line server;
S12, then, the respective links of each image are stored into database.
3. a kind of patent image-recognizing method based on Android phone platform according to claim 1, its feature exist
In in the step S03, the comparison algorithm can use the histogram for perceiving hash algorithm or calculating image.
4. a kind of patent image-recognizing method based on Android phone platform according to claim 3, its feature exist
In the perception hash algorithm, comprising the following steps:
S21, the size that picture is narrowed down to 8x8, make picture include 64 pixels;
S22, simplify picture color, convert the image into gray scale;
S23, the average gray value for calculating all 64 pixels, and by the gray value of each pixel compared with average value, such as
Fruit is more than average value, then is recorded as 1, if less than average value, is then recorded as 0;
S24, all values for recording to obtain of comparing are put into an array, this is image " fingerprint ", then by comparing two
Corresponding value in " fingerprint " of individual image, calculate how many position in 64 be it is the same, according to identical digit number, analysis
Picture phase knowledge and magnanimity.
5. a kind of patent image-recognizing method based on Android phone platform according to claim 3, its feature exist
In, it is described calculate image histogram, comprise the following steps:
S31, firstly, since picture size is inconsistent, it is standard size to need to adjust picture before comparison;
S32, secondly, RGB image is converted into gray level;
S33, then, because the type of image is matrix, floating type is converted to by each pixel value, straight preferably to calculate image
Fang Tu;
It is S34, last, compare two histograms, convert the result to percentage, and by likelihood storage into priority query.
6. a kind of patent image-recognizing method based on Android phone platform according to claim 5, its feature exist
In in the step S33, for ease of the histogram of computing array, calling calcHist () function in built-in OpenCV.
7. a kind of patent image-recognizing method based on Android phone platform according to claim 5, its feature exist
In, in the step S34, for ease of comparing two histograms, compareHist () function in built-in OpenCV is called,
And selecting the standard of Pasteur's coefficient therein to be calculated, its calculation formula is as follows:
Wherein H1(I) gray scale of post is corresponded in first image histogram of expression;H2(I) represent right in second image histogram
Ying Zhu gray scale;Represent the average gray of all posts in first image histogram;Represent in second image histogram
The average gray of all posts;D represents the distance between two image histograms, and the value reacts similar between two image histograms
Property, d values show that more greatly two image histograms are more similar.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710784529.0A CN107622278A (en) | 2017-09-01 | 2017-09-01 | A kind of patent image-recognizing method based on Android phone platform |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710784529.0A CN107622278A (en) | 2017-09-01 | 2017-09-01 | A kind of patent image-recognizing method based on Android phone platform |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107622278A true CN107622278A (en) | 2018-01-23 |
Family
ID=61088466
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710784529.0A Pending CN107622278A (en) | 2017-09-01 | 2017-09-01 | A kind of patent image-recognizing method based on Android phone platform |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107622278A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110414201A (en) * | 2018-04-26 | 2019-11-05 | 仁宝电脑工业股份有限公司 | Electronic device and its operating method with fingerprint authentication function |
CN112163105A (en) * | 2020-07-13 | 2021-01-01 | 北京国电通网络技术有限公司 | Image data storage method and device, electronic equipment and storage medium |
CN113283617A (en) * | 2021-05-21 | 2021-08-20 | 东华大学 | Clothing reconstruction method and server system thereof |
US11501344B2 (en) * | 2019-10-14 | 2022-11-15 | Bottomline Technologies Limited | Partial perceptual image hashing for invoice deconstruction |
-
2017
- 2017-09-01 CN CN201710784529.0A patent/CN107622278A/en active Pending
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110414201A (en) * | 2018-04-26 | 2019-11-05 | 仁宝电脑工业股份有限公司 | Electronic device and its operating method with fingerprint authentication function |
US11501344B2 (en) * | 2019-10-14 | 2022-11-15 | Bottomline Technologies Limited | Partial perceptual image hashing for invoice deconstruction |
CN112163105A (en) * | 2020-07-13 | 2021-01-01 | 北京国电通网络技术有限公司 | Image data storage method and device, electronic equipment and storage medium |
CN112163105B (en) * | 2020-07-13 | 2024-02-09 | 北京国电通网络技术有限公司 | Image data storage method and device, electronic equipment and storage medium |
CN113283617A (en) * | 2021-05-21 | 2021-08-20 | 东华大学 | Clothing reconstruction method and server system thereof |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10803359B2 (en) | Image recognition method, apparatus, server, and storage medium | |
CN107622278A (en) | A kind of patent image-recognizing method based on Android phone platform | |
CN109410239A (en) | A kind of text image super resolution ratio reconstruction method generating confrontation network based on condition | |
US8232996B2 (en) | Image learning, automatic annotation, retrieval method, and device | |
CN108491889A (en) | Image, semantic dividing method, device and computer readable storage medium | |
CN111325271B (en) | Image classification method and device | |
Varga et al. | Fully automatic image colorization based on Convolutional Neural Network | |
CN111696101A (en) | Light-weight solanaceae disease identification method based on SE-Inception | |
KR101384627B1 (en) | Method for the automatic segmentation of object area in video | |
CN110348322A (en) | Human face in-vivo detection method and equipment based on multi-feature fusion | |
CN109271930B (en) | Micro-expression recognition method, device and storage medium | |
CN104361348B (en) | A kind of flowers recognition methods on intelligent terminal | |
US20210117687A1 (en) | Image processing method, image processing device, and storage medium | |
CN113627402B (en) | Image identification method and related device | |
CN113642400A (en) | Graph convolution action recognition method, device and equipment based on 2S-AGCN | |
CN112561976A (en) | Image dominant color feature extraction method, image retrieval method, storage medium and device | |
CN115761356A (en) | Image recognition method and device, electronic equipment and storage medium | |
CN116189162A (en) | Ship plate detection and identification method and device, electronic equipment and storage medium | |
CN115482529A (en) | Method, equipment, storage medium and device for recognizing fruit image in near scene | |
CN113011506B (en) | Texture image classification method based on deep fractal spectrum network | |
CN109740432A (en) | Plant monitoring method, device, computer equipment and storage medium | |
CN114694133B (en) | Text recognition method based on combination of image processing and deep learning | |
CN110796650A (en) | Image quality evaluation method and device, electronic equipment and storage medium | |
CN113052176A (en) | Character recognition model training method, device and system | |
CN111259734A (en) | Spatial house type identification method and device, computer equipment and storage medium |
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 |