CN110837848A - Characteristic pixel image identification method and system - Google Patents
Characteristic pixel image identification method and system Download PDFInfo
- Publication number
- CN110837848A CN110837848A CN201910974330.3A CN201910974330A CN110837848A CN 110837848 A CN110837848 A CN 110837848A CN 201910974330 A CN201910974330 A CN 201910974330A CN 110837848 A CN110837848 A CN 110837848A
- Authority
- CN
- China
- Prior art keywords
- image
- pixel
- recognized
- standard
- frequency distribution
- 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.)
- Withdrawn
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
- G06V10/56—Extraction of image or video features relating to colour
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to the field of image recognition, and particularly relates to a characteristic pixel image recognition method and a system, wherein the method comprises the steps of obtaining a BMP format image of an object to be recognized, obtaining the image to be recognized, extracting RGB (red, green and blue) coded values of pixel points in the image to be recognized, and obtaining pixel frequency distribution characteristics of the image to be recognized; and excavating and analyzing the standard image to obtain statistical pixel frequency distribution characteristics, establishing a standard characteristic pixel point coding library through the statistical pixel frequency distribution characteristics, segmenting and identifying the pixel frequency distribution characteristics of the image to be identified by using the standard characteristic pixel point coding library, and judging whether the image to be identified and the standard image have consistent pixel frequency distribution characteristics. The image recognition method of the invention directly uses the pixel coding characteristic frequency distribution as difference recognition, assists in the prior and later period dynamic recognition to extract object characteristic pixel coding and build a library, can gradually segment and recognize the target object in the image, and has simple image recognition process and high efficiency.
Description
Technical Field
The invention relates to the field of image recognition, in particular to a method and a system for recognizing a characteristic pixel image.
Background
At present, the inspection of defective grains of grains in China is mainly based on manual sensory inspection, and a few items are realized by means of equipment using an image analysis method as a principle, such as the content determination of yellow-grained rice of rice, the determination of the whole polished rice rate of rice and the like. By means of the latest technology at the front end, the automation degree of equipment is greatly improved, manual intervention in the detection process is reduced, the efficiency is improved, large-scale and large-scale grain sensory inspection work is completed, and the significance for acquiring large inspection data is great. However, the existing automatic identification methods mostly use a deep learning image identification technology based on a convolutional neural network, and the identification process is complex and low in efficiency.
Disclosure of Invention
The invention aims to provide a method and a system for identifying a characteristic pixel image.
The technical scheme for solving the technical problems is as follows: a characteristic pixel image recognition method comprises the following steps,
s1, acquiring a BMP format image of an object to be recognized to obtain an image to be recognized, extracting RGB (red, green and blue) code values of each pixel point in the image to be recognized to obtain pixel frequency distribution characteristics of the image to be recognized;
s2, obtaining statistical pixel frequency distribution characteristics through mining and analyzing standard images, establishing a standard characteristic pixel point coding library through the statistical pixel frequency distribution characteristics, carrying out segmentation and identification on the pixel frequency distribution characteristics of the images to be identified by utilizing the standard characteristic pixel point coding library, and judging whether the images to be identified and the standard images have consistent pixel frequency distribution characteristics.
The invention has the beneficial effects that: the invention relates to a characteristic pixel image recognition method which is different from the current popular deep learning image recognition method based on a convolutional neural network, the image recognition method of the invention directly uses the pixel coding characteristic frequency distribution as difference recognition to assist the early and late dynamic recognition to extract the characteristic object pixel coding and establish a library, the target object in the image can be segmented and recognized step by step, the actually used pixel library space is not large, and the total amount is 24 (16) bytes of 2; meanwhile, the method has strong realizability, is directly faced with image data coding processing, is not required to carry various large open source code libraries in programming, is short in code, convenient and efficient to use, and has small computation time complexity and space complexity, so that the image recognition process is simple and efficient.
On the basis of the technical scheme, the invention can be further improved as follows.
Further, in S2, specifically,
s21a, mining and analyzing the standard image to obtain statistical pixel frequency distribution characteristics, establishing a standard characteristic pixel point coding library through the statistical pixel frequency distribution characteristics, and taking the subtraction value of the coding value high-frequency state frequency sum and the coding value low-frequency state frequency sum on each channel of RGB three channels in the standard characteristic pixel point coding library as an identification frequency characteristic value;
s22a, counting the subtraction value of the encoding value high-frequency state frequency sum and the encoding value low-frequency state frequency sum on each channel of RGB three channels in the pixel frequency distribution characteristics of the image to be recognized;
s23a, when the subtraction value on each channel of the RGB three channels in the pixel frequency distribution characteristics of the image to be recognized is correspondingly matched with the recognition frequency characteristic value on the corresponding channel of the RGB three channels in the standard characteristic pixel point coding library, judging that the image to be recognized and the standard image have consistent pixel frequency distribution characteristics, otherwise, judging that the image to be recognized and the standard image do not have consistent pixel frequency distribution statistical characteristics.
Further, in S2, specifically,
s21b, mining and analyzing the standard image to obtain statistical pixel frequency distribution characteristics, establishing a standard characteristic pixel point coding library through the statistical pixel frequency distribution characteristics, performing coding difference calculation between two adjacent channels of RGB three channels of each pixel point in the standard characteristic pixel point coding library, taking a coding difference absolute value, counting the frequencies of the coding difference absolute values between the two adjacent channels of all the pixel points in the standard characteristic pixel point coding library, selecting frequencies meeting a preset rule from the frequencies of the coding difference absolute values between the two adjacent channels of all the pixel points in the standard characteristic coding pixel point library, and taking the sum of the selected frequencies as an identification frequency characteristic value;
s22b, performing coding difference operation between two adjacent channels of RGB three channels of each pixel point in the pixel frequency distribution characteristics of the image to be recognized, taking a coding difference absolute value, and counting the frequency of the coding difference absolute value between the two adjacent channels of all the pixel points in the pixel frequency distribution characteristics of the image to be recognized to obtain the frequency distribution of the coding difference of the image to be recognized;
s23b, comparing the frequency distribution of the coding difference of the image to be recognized with the recognition frequency characteristic value, when the frequency distribution of the coding difference of the image to be recognized accords with the recognition frequency characteristic value, judging that the image to be recognized and the standard image have consistent pixel frequency distribution characteristics, otherwise, judging that the image to be recognized and the standard image do not have consistent pixel frequency distribution statistical characteristics.
Further, in S2, specifically,
s21c, mining and analyzing the standard image to obtain statistical pixel frequency distribution characteristics, and establishing a standard characteristic pixel point coding library through the statistical pixel frequency distribution characteristics; defining a local range in the standard image, and calculating the sum of subtraction absolute values of two adjacent pixel points on corresponding channels of three RGB channels in the local range in the standard image according to the standard characteristic pixel point coding library; for the whole standard image, a local range is selected in a flowing mode until the whole standard image is covered, the sum of subtraction absolute values of two adjacent pixel points in all the local ranges in the standard image on corresponding channels of three RGB channels is calculated, and the frequency of the sum of subtraction absolute values of two adjacent pixel points in all the local ranges in the standard image on corresponding channels of three RGB channels is counted; selecting frequencies which accord with a preset rule from the frequencies of the sum of subtraction absolute values of two adjacent pixel points on corresponding channels of three RGB channels in all local ranges in the standard image, and taking the sum of the selected frequencies as a recognition frequency characteristic value;
s22c, defining a local range in the image to be recognized, and calculating the sum of subtraction absolute values of two adjacent pixel points on corresponding channels of three channels of RGB in the local range in the image to be recognized according to the pixel frequency distribution characteristics of the image to be recognized; for the whole image to be recognized, a local range is selected in a flowing mode until the whole image to be recognized is covered, the sum of subtraction absolute values of two adjacent pixel points in all the local ranges in the image to be recognized on corresponding channels of three channels of RGB is calculated, and the frequency distribution of the sum of subtraction absolute values of two adjacent pixel points in all the local ranges in the image to be recognized on corresponding channels of three channels of RGB is counted;
s23c, extracting frequencies meeting preset rules from the frequency distribution of the sum of subtraction absolute values of adjacent two pixel points on corresponding channels of three channels of RGB in all local ranges of the image to be recognized, and accumulating and calculating to obtain the frequency accumulation sum of the image to be recognized; and when the frequency accumulation sum value of the image to be recognized accords with the recognition frequency characteristic value, judging that the image to be recognized and the standard image have consistent pixel frequency distribution characteristics, otherwise, judging that the image to be recognized and the standard image do not have consistent pixel frequency distribution statistical characteristics.
Further, in S1, specifically,
s11, acquiring a BMP format image of the object to be recognized to obtain the image to be recognized;
s12, removing the background in the image to be recognized to obtain a foreground image in the image to be recognized;
and S13, extracting RGB (red, green and blue) coding values of all pixel points of the foreground image in the image to be recognized to obtain pixel frequency distribution characteristics of the foreground image in the image to be recognized.
Further, in S12, specifically,
s121, shooting a background image in advance, carrying out image grid analysis on the image to be recognized by utilizing a pixel coding high-frequency state of the shot background image in advance to form an image grid set to be recognized, judging a background grid from the image grid set to be recognized, and defining the remaining image grids which are not judged as the background grids in the image grid set to be recognized as a remaining image grid set to be recognized;
s122, utilizing an accumulated background image pixel library to perform pixel identification on any to-be-identified image grid in the residual to-be-identified image grid set, marking the identified pixel position, and marking as a background pixel;
s123, when the amount of the image grids to be recognized marked as background pixels in the residual image grid set to be recognized reaches a preset proportion, judging that the image grids to be recognized marked as background pixels in the residual image grid set to be recognized are background grids;
s124, taking out the pixel code values in all the to-be-identified image grids judged as the background grids, putting the pixel code values into a background pixel library, and circularly and iteratively executing S122-S124 until no new background grids in the residual to-be-identified image grid set are identified;
and S125, removing all pixel points in the to-be-identified image grids which are judged as background grids from the to-be-identified image grid set to obtain a foreground image in the to-be-identified image.
Furthermore, the statistical pixel frequency distribution characteristics are obtained by mining and analyzing the standard image, and a standard characteristic pixel point coding library is established through the statistical pixel frequency distribution characteristics,
acquiring a BMP format image of a standard object which is the same as the object to be identified to obtain a standard image;
and extracting RGB (red, green and blue) code values of all pixel points of the foreground image in the standard image to obtain pixel frequency distribution characteristics of the foreground image in the standard image, and directly extracting the RGB code values of all pixel points of the foreground image in the standard image to build a library to obtain a standard characteristic pixel point code library.
Further, step S2 is followed by step S,
s3, when the image to be recognized and the standard image are judged to have the consistent pixel frequency distribution statistical characteristics and the picture colors of the foreground image of the image to be recognized are observed to be consistent, dividing the foreground image of the image to be recognized into grids with the same size, counting high-frequency RGB (red, green and blue) coded values of each grid in the foreground image of the whole image to be recognized, calculating a T value of normal distribution according to the frequency of all the high-frequency RGB coded values in each grid in the foreground image of the whole image to be recognized, if the T value reaches a preset threshold value, judging that the corresponding grid is a normal grid of picture colors, otherwise, judging that the corresponding grid is an abnormal grid of picture colors.
Further, when the corresponding grid is judged to be a grid with normal picture color, the RGB code value of each pixel point in the corresponding grid is added into the standard characteristic pixel point code library.
Based on the characteristic pixel image identification method, the invention also provides a characteristic pixel image identification system.
A characteristic pixel image recognition system comprises the following modules,
the pixel frequency distribution characteristic acquisition module is used for acquiring a BMP format image of an object to be recognized to obtain an image to be recognized, extracting RGB (red, green and blue) code values of all pixel points in the image to be recognized and obtaining a pixel frequency distribution characteristic of the image to be recognized;
and the segmentation identification module is used for mining and analyzing a standard image to obtain statistical pixel frequency distribution characteristics, establishing a standard characteristic pixel point coding library through the statistical pixel frequency distribution characteristics, and performing segmentation identification on the pixel frequency distribution characteristics of the image to be identified by using the standard characteristic pixel point coding library to judge whether the image to be identified and the standard image have consistent pixel frequency distribution characteristics.
The invention has the beneficial effects that: the invention relates to a characteristic pixel image recognition system which is different from a current popular deep learning image recognition system based on a convolutional neural network, wherein the image recognition system directly performs difference recognition by using pixel coding characteristic frequency distribution, assists in early-stage and later-stage dynamic recognition to extract characteristic object pixel coding and establish a library, can gradually partition and recognize a target object in an image, and has small actually used pixel library space with the total amount of 24-power (16 megabytes) of 2; meanwhile, the method has strong realizability, is directly faced with image data coding processing, is not required to carry various large open source code libraries in programming, is short in code, convenient and efficient to use, and has small computation time complexity and space complexity, so that the image recognition process is simple and efficient.
Drawings
FIG. 1 is a flow chart of a method for feature pixel image recognition according to the present invention;
FIG. 2 is a block diagram of a feature pixel image recognition system according to the present invention.
Detailed Description
The principles and features of this invention are described below in conjunction with the following drawings, which are set forth by way of illustration only and are not intended to limit the scope of the invention.
As shown in fig. 1, a method for recognizing a characteristic pixel image includes the steps of,
s1, acquiring a BMP format image of an object to be recognized to obtain an image to be recognized, extracting RGB (red, green and blue) code values of each pixel point in the image to be recognized to obtain pixel frequency distribution characteristics of the image to be recognized;
s2, obtaining statistical pixel frequency distribution characteristics through mining and analyzing standard images, establishing a standard characteristic pixel point coding library through the statistical pixel frequency distribution characteristics, carrying out segmentation and identification on the pixel frequency distribution characteristics of the images to be identified by utilizing the standard characteristic pixel point coding library, and judging whether the images to be identified and the standard images have consistent pixel frequency distribution characteristics.
The image recognition principle in the method of the invention is as follows:
the RGB code values in BMP format photos shot by a camera are researched, and the RGB code values of pixel points forming the images have relatively consistent frequency distribution characteristics when the colors of the exterior surfaces of the objects in the same class are the same: the R value of each pixel point in the image is 0-255 frequency distribution, the G value is 0-255 frequency distribution, the B value is 0-255 frequency distribution, and the values of (R, G), (R, B) and (G, B) between every two pixel points in the RGB are (0, 0) - (255 ) frequency distribution. Different objects have different colors, so that pixel frequency distribution is different, according to the principle, the statistical distribution rule of object image codes to be identified can be mined and extracted in advance, corresponding characteristic pixel points are stored in a database, and the similar object images appearing in the photos can be segmented and identified.
The first embodiment is as follows:
in the present invention, the S2 may be specifically,
s21a, mining and analyzing the standard image to obtain statistical pixel frequency distribution characteristics, establishing a standard characteristic pixel point coding library through the statistical pixel frequency distribution characteristics, and taking the subtraction value of the coding value high-frequency state frequency sum and the coding value low-frequency state frequency sum on each channel of RGB three channels in the standard characteristic pixel point coding library as an identification frequency characteristic value;
s22a, counting the subtraction value of the encoding value high-frequency state frequency sum and the encoding value low-frequency state frequency sum on each channel of RGB three channels in the pixel frequency distribution characteristics of the image to be recognized;
s23a, when the subtraction value on each channel of the RGB three channels in the pixel frequency distribution characteristics of the image to be recognized is correspondingly matched with the recognition frequency characteristic value on the corresponding channel of the RGB three channels in the standard characteristic pixel point coding library, judging that the image to be recognized and the standard image have consistent pixel frequency distribution characteristics, otherwise, judging that the image to be recognized and the standard image do not have consistent pixel frequency distribution statistical characteristics.
In the first embodiment, the frequency distribution can be flexibly processed, such as: for each channel 256 state of the three-channel RGB code, the 256 states can be classified according to the high-order 4 bits of the code, and the 256 states are divided into 16 states to examine the frequency distribution. And in the standard characteristic pixel point coding library, taking the subtraction value of the high-frequency state frequency sum and the low-frequency state frequency sum in 16 states as the identification frequency characteristic value. When the subtraction value of the actually counted pixel coding high-frequency state frequency sum and the low-frequency state frequency sum in the foreground image of the image to be recognized reaches the recognition frequency characteristic value, the foreground image of the image to be recognized can be considered to be the image of the object corresponding to the pixel frequency distribution characteristic of the standard image. Generally speaking, if R, G, B the code values of three channel pixels all conform to the corresponding recognition frequency characteristic value, it can be determined that the foreground image of the image to be recognized is the image of the object corresponding to the pixel frequency distribution characteristic of the standard image.
Example two:
the S2 can also be embodied as,
s21b, mining and analyzing the standard image to obtain statistical pixel frequency distribution characteristics, establishing a standard characteristic pixel point coding library through the statistical pixel frequency distribution characteristics, performing coding difference calculation between two adjacent channels of RGB three channels of each pixel point in the standard characteristic pixel point coding library, taking a coding difference absolute value, counting the frequencies of the coding difference absolute values between the two adjacent channels of all the pixel points in the standard characteristic pixel point coding library, selecting frequencies meeting a preset rule from the frequencies of the coding difference absolute values between the two adjacent channels of all the pixel points in the standard characteristic coding pixel point library, and taking the sum of the selected frequencies as an identification frequency characteristic value;
s22b, performing coding difference operation between two adjacent channels of RGB three channels of each pixel point in the pixel frequency distribution characteristics of the image to be recognized, taking a coding difference absolute value, and counting the frequency of the coding difference absolute value between the two adjacent channels of all the pixel points in the pixel frequency distribution characteristics of the image to be recognized to obtain the frequency distribution of the coding difference of the image to be recognized;
s23b, comparing the frequency distribution of the coding difference of the image to be recognized with the recognition frequency characteristic value, when the frequency distribution of the coding difference of the image to be recognized accords with the recognition frequency characteristic value, judging that the image to be recognized and the standard image have consistent pixel frequency distribution characteristics, otherwise, judging that the image to be recognized and the standard image do not have consistent pixel frequency distribution statistical characteristics.
For example: recording RGB as a three-channel coding value of a certain pixel point in a standard image, respectively extracting R, G, B coding values to perform difference operation and take difference absolute values, namely | R-G | and | G-B |, then calculating all pixel points in the whole standard image as above, classifying according to the same type of samples (such as qualified grains), respectively accumulating and counting frequencies corresponding to the two difference absolute values, and selecting the sum of the corresponding frequencies at a plurality of difference absolute values with higher occurrence frequency in the qualified grains and obviously different from the imperfect grains as an identification frequency characteristic value of a judgment standard, wherein one identification frequency characteristic value corresponds to | R-G | and one identification frequency characteristic value corresponds to | G-B |. And then, respectively carrying out difference operation on RGB three-channel coding values of each pixel point of the image to be recognized, taking absolute values of differences, namely | R-G | and | G-B |, and then respectively counting the frequencies of the two absolute values of the differences to obtain two types of frequency distribution. And finally, calculating the sum of the frequencies according to the frequency statistical results of the images to be recognized and the corresponding absolute values of the difference values respectively selected by two thresholds (preset rules) of the qualified grains, comparing the sum with the characteristic values of the two recognition frequencies respectively, and preliminarily judging the images to be recognized as the qualified grains if the sum is satisfied with the characteristic values of the two recognition frequencies, otherwise, judging the images to be recognized as the unqualified grains.
Example three:
the S2 can also be embodied as,
s21c, mining and analyzing the standard image to obtain statistical pixel frequency distribution characteristics, and establishing a standard characteristic pixel point coding library through the statistical pixel frequency distribution characteristics; defining a local range in the standard image, and calculating the sum of subtraction absolute values of two adjacent pixel points on corresponding channels of three RGB channels in the local range in the standard image according to the standard characteristic pixel point coding library; for the whole standard image, a local range is selected in a flowing mode until the whole standard image is covered, the sum of subtraction absolute values of two adjacent pixel points in all the local ranges in the standard image on corresponding channels of three RGB channels is calculated, and the frequency of the sum of subtraction absolute values of two adjacent pixel points in all the local ranges in the standard image on corresponding channels of three RGB channels is counted; selecting frequencies which accord with a preset rule from the frequencies of the sum of subtraction absolute values of two adjacent pixel points on corresponding channels of three RGB channels in all local ranges in the standard image, and taking the sum of the selected frequencies as a recognition frequency characteristic value;
s22c, defining a local range in the image to be recognized, and calculating the sum of subtraction absolute values of two adjacent pixel points on corresponding channels of three channels of RGB in the local range in the image to be recognized according to the pixel frequency distribution characteristics of the image to be recognized; for the whole image to be recognized, a local range is selected in a flowing mode until the whole image to be recognized is covered, the sum of subtraction absolute values of two adjacent pixel points in all the local ranges in the image to be recognized on corresponding channels of three channels of RGB is calculated, and the frequency distribution of the sum of subtraction absolute values of two adjacent pixel points in all the local ranges in the image to be recognized on corresponding channels of three channels of RGB is counted;
s23c, extracting frequencies meeting preset rules from the frequency distribution of the sum of subtraction absolute values of adjacent two pixel points on corresponding channels of three channels of RGB in all local ranges of the image to be recognized, and accumulating and calculating to obtain the frequency accumulation sum of the image to be recognized; and when the frequency accumulation sum value of the image to be recognized accords with the recognition frequency characteristic value, judging that the image to be recognized and the standard image have consistent pixel frequency distribution characteristics, otherwise, judging that the image to be recognized and the standard image do not have consistent pixel frequency distribution statistical characteristics.
For example: in a standard image, a rule is given to define a local range, two adjacent pixel points are selected in the local range, the coding values of the two adjacent pixel points in the local range of the standard image are respectively R1G1B1 and R2G2B2, the sum of subtraction absolute values of the two adjacent pixel points on a channel corresponding to RGB three channels is calculated, namely | R1-R2| + | G1-G2| + | B1-B2|, and the frequency of the sum of the subtraction absolute values is counted; the whole standard image is subjected to flow selection in a local range until the whole standard image is covered, all statistical frequency results are accumulated, and the frequency distribution of the sum of subtraction absolute values between all adjacent two pixel points in the whole standard image is calculated; for the same type samples (such as qualified grains), the statistical results of the photos of each sample are accumulated, and the overall statistical frequency distribution is calculated; and comparing and selecting the frequency of the part absolute value sum value (the sum of the subtraction absolute values) of the qualified grains, wherein the frequency of the part absolute value sum value of the qualified grains is obviously higher than that of the imperfect grains, and adding the part absolute value sum value of the qualified grains to be used as the characteristic value of the identification frequency for judging the qualified grains. Calculating the sum of absolute values of subtraction differences on channels corresponding to RGB three channels of an image to be recognized according to two adjacent pixel points in a local range determined by a given rule, and counting the frequency of the sum; the local range of flow, covering the whole image to be identified, is then counted for the frequency distribution of such "sum" (i.e. subtracting the sum of the absolute values of the differences). And extracting the frequency of the corresponding statistical sum value when the threshold value of the qualified granules is selected from the statistical result of the image to be identified, accumulating the sum, comparing the sum with the characteristic value of the identification frequency, and if the sum is satisfied, primarily judging the qualified granules, otherwise, judging the unqualified granules.
In the present invention, the step S1 is specifically,
s11, acquiring a BMP format image of the object to be recognized to obtain the image to be recognized;
s12, removing the background in the image to be recognized to obtain a foreground image in the image to be recognized;
and S13, extracting RGB (red, green and blue) coding values of all pixel points of the foreground image in the image to be recognized to obtain pixel frequency distribution characteristics of the foreground image in the image to be recognized.
In the present invention, the step S12 is specifically,
s121, shooting a background image in advance, carrying out image grid analysis on the image to be recognized by utilizing a pixel coding high-frequency state of the shot background image in advance to form an image grid set to be recognized, judging a background grid from the image grid set to be recognized, and defining the remaining image grids which are not judged as the background grids in the image grid set to be recognized as a remaining image grid set to be recognized;
s122, utilizing an accumulated background image pixel library to perform pixel identification on any to-be-identified image grid in the residual to-be-identified image grid set, marking the identified pixel position, and marking as a background pixel;
s123, when the amount of the image grids to be recognized marked as background pixels in the residual image grid set to be recognized reaches a preset proportion, judging that the image grids to be recognized marked as background pixels in the residual image grid set to be recognized are background grids;
s124, taking out the pixel code values in all the to-be-identified image grids judged as the background grids, putting the pixel code values into a background pixel library, and circularly and iteratively executing S122-S124 until no new background grids in the residual to-be-identified image grid set are identified;
and S125, removing all pixel points in the to-be-identified image grids which are judged as background grids from the to-be-identified image grid set to obtain a foreground image in the to-be-identified image.
In the invention, the statistical pixel frequency distribution characteristics are obtained by mining and analyzing the standard image, and a standard characteristic pixel point coding library is established by the statistical pixel frequency distribution characteristics,
acquiring a BMP format image of a standard object which is the same as the object to be identified to obtain a standard image; and extracting RGB (red, green and blue) code values of all pixel points of the foreground image in the standard image to obtain pixel frequency distribution characteristics of the foreground image in the standard image, and directly extracting the RGB code values of all pixel points of the foreground image in the standard image to build a library to obtain a standard characteristic pixel point code library.
In the invention, the standard characteristic pixel point coding library is composed as follows: the standard feature pixel point code library has a size of 24 bytes (total 16 megabytes) of 2, wherein the address value corresponds to the code value of 24 bits of the RGB pixel, and the address corresponds to the stored 1-byte content (binary 8 bits). This content is a flag that this pixel appears in a different identifier. For example, for rice, the object to be identified includes background color, qualified rice grains, scab grains, sprouting grains, mildew grains, worm-eaten grains, immature grains, and the like. 1 byte has 8 bits, each bit can take two symbols of '0' and '1', with '0' indicating no occurrence and '1' indicating occurrence. When imperfect wheat grains are identified, a 16-megabyte wheat standard characteristic pixel point coding library can be established.
In the invention, the standard characteristic pixel point coding library is established in the following way:
the library building method comprises two modes: (1) and taking a picture by using the sample to extract pixels and build a library. (2) And dynamically adding pixels to build a library by utilizing statistical inference in the identification process.
Taking background pixel extraction as an example:
(1) and establishing a library for taking picture of prime points in advance
The imperfect grains of the grain are identified by taking a picture, and the material and the color of the background can be manually selected in advance. After the background color, the optical photographing environment and the illumination intensity are selected, photographing is carried out in advance to collect background color pixel points in the background photo, and a library is directly extracted and built.
(2) Extracting pixel points by utilizing statistical inference in the software identification process to build a library
In the actual working process, the optical environment fluctuates within a certain allowable range, and especially, dust of grain particles is gradually adsorbed on the background material. Therefore, the pixel points extracted by shooting and framing in advance cannot cover all the actually-appearing background pixel points. At this time, when a certain local area is judged to be a background color area by utilizing the pixel points of the established background pixel library and combining a statistical inference method, all the pixel points of the area can be extracted in time, the pixel points which cannot be identified as the background color in the pixel library are supplemented, and the library is dynamically established.
When the grain particle pixel points are extracted and a database is built, background pixels of pictures shot by classified samples are removed, and then extraction is started. Along with the increase of sample photo accumulation, the extracted pixel point coding value is gradually increased to a certain amount, and the identification requirements of basic pixels of qualified grains and imperfect grains can be basically met.
In the invention, the standard characteristic pixel point coding library is used as follows:
in the actual working process, when a target containing grain grains is identified, the standard characteristic pixel point coding library is compared with the statistical distribution characteristics of actual pixel points in the picture for combined use, and finally, a target area pattern is researched and judged to identify grain qualified grains and grain imperfect grains.
Taking background area identification as an example. The background pixel value statistical distribution characteristics obtained in advance from the background sample picture are utilized, the background area which can be identified is compared and identified from the picture to be identified, and then the pixel point in the picture is compared and supplemented with the identification by the existing background standard characteristic pixel point coding library.
If the local scab of the scab particles is identified, a background area is generally removed from a picture, after a foreground is extracted, the grain particle picture pixel to be identified can be compared with the scab pixel in a standard feature pixel point coding library by means of the actual statistical feature of the foreground pixel value after the grain particle picture is preliminarily judged to be the scab particles, identification statistics is carried out, and when the occurrence quantity ratio of the pixels exceeds a threshold value, the grain particle picture pixel can be judged to be the scab particles. It should be noted that, because the outer surface color pixel on the upper portion of the lesion grain is the same as the outer surface color pixel of the qualified grain, the same pixel is identified repeatedly when the library is built. When the local lesion spot area is identified, the pixels only marked as lesion spots need to be selected for comparison and confirmation, so as to reduce misjudgment.
In addition, when the image to be recognized and the standard image are judged to have the consistent pixel frequency distribution statistical characteristics and the picture colors of the foreground image of the image to be recognized are observed to be consistent, the RGB codes of the pixels in the foreground image of the image to be recognized are relatively close, and a large number of high-frequency states of RGB code values can appear along with the increase of the picture; at this time, the foreground image of the image to be recognized may be divided into grids of the same size, a high-frequency RGB code value appearing in the entire foreground image of the image to be recognized in each grid is counted, a T value of normal distribution is calculated by a frequency of appearance of all the high-frequency RGB code values in each grid in the entire foreground image of the image to be recognized, if the T value reaches a preset threshold value, the corresponding grid is determined to be a normal grid of picture color, otherwise, the corresponding grid is determined to be an abnormal grid of picture color. Therefore, the grid picture can be divided, and a foundation is laid for detailed image identification. The grid picture segmentation is a single-color area image identification technology, can be applied to background area identification segmentation, and also can be applied to continuous area identification segmentation with single color after a foreground area is extracted. In the photo of imperfect grain recognition, the background color is single, and the covered area is generally larger. When the background pixel feature library accumulates fewer feature pixels, the background area can be identified by the technology. In addition, the qualified grains of the grains (such as the qualified grains of the rice) also have certain surface regions with single color, background regions are removed from the photos, and after foreground grain particle images are extracted, the color consistency judgment can be carried out on the regions after gridding by the technology. Find the grid area of the abnormal color. Some of the sprouted kernels, except for the sprouted embryo, showed little difference in the surface of the pellet from the acceptable kernels, which helps to locate the germ area.
When the corresponding grid is judged to be a grid with normal picture color, adding the RGB code value of each pixel point in the corresponding grid into the standard characteristic pixel point code library, and directly using the RGB code value for image segmentation and identification in a new photo; therefore, dynamic construction of a standard characteristic pixel point coding library can be realized. The pixels with the same coding value can appear in different areas of the same photo for multiple times, or appear in different photos, such as background pixels with single color. When the pixel point does not appear in the standard characteristic pixel point coding library, the pixel point is identified by a grid identification method and then recorded in the standard characteristic pixel point coding library, and the pixel point can be used for identifying the pixel point in the grid to be identified or can be used for segmenting and identifying the picture identification in other new pictures.
The patterns with the same or similar visual colors are formed by a plurality of pixel points with the same coding value, for example, the background color with single color in the grain photos basically keeps unchanged in the shot different grain photos, so that the pixel points with the same coding value exist in the different photo backgrounds. By utilizing the characteristics, background pixels are extracted to build a library, and the pixel codes can be used for background recognition of other new photos.
The following describes the quality of grain particles in detail by taking the method of the present invention as an example.
And (3) image recognition of imperfect rice grains:
1. acquiring a rice grain image to be identified (a camera photo JPG format is converted into a BMP format, and if the camera photo is an industrial camera, the BMP format is directly adopted);
2. identifying the background of the rice particle image to be identified;
(21) when the background image can be manually selected and is relatively fixed, the background image can be shot in advance, the high-frequency state of the pixel coding of the background image is utilized to perform image grid analysis on the rice grain image to be identified, and some exact background grids are intensively judged from the rice grain image grid to be identified;
(22) utilizing the accumulated background image pixel library to perform pixel identification on the remaining undetermined rice particle image grids to be identified, and marking the identified pixel positions;
(23) when the amount of the background pixels marked in the rice grain image grid to be identified reaches a preset proportion, the rice grain image grid to be identified can be judged as the background grid;
(24) and (5) taking out the pixel code values in the rice grain image grid to be identified, which are judged as the background grid in the previous three steps, putting the pixel code values into a background pixel library, and iteratively executing the steps (22), (23) and (24) until no new background grid is identified.
3. And eliminating all pixel points in the to-be-identified rice grain image grids which are judged as background grids, and identifying and matting to obtain a rice grain foreground image.
4. Analyzing the rice grain image to be identified, and identifying the grain attribute (qualified or imperfect grain)
(41) And (3) setting a standard by utilizing the pixel frequency distribution characteristics of the qualified rice grain images, and directly judging the rice grain images to be identified which obviously do not accord with the characteristics as imperfect grains (mildewed grains, scab grains, germinated grains, immature grains and wormhole grains).
(42) Further analyzing the image close to the standard to find whether small buds, small mildew spots, disease spots and small insect eyes exist or not, and judging the image to be an imperfect particle after identification;
among these, the method of further analyzing the image close to this standard is as follows: and judging by means of the pixel identification established by the characteristic pixel point coding library. The bud granules are determined as examples, and pixel points of the buds of the bud granules are formed, are concentrated in the bud area and basically cannot appear in other imperfect granules. When pixels in photo particles to be recognized are compared with pixel library pixel points, the pixels are marked as the appearance of the sprouting particles, when the appearance amount of the pixel points meets a given threshold value, the pixels are marked on the photo particles, and if the pixel points are determined to be adjacent and appear in small parts, the grain particles can be judged to be the sprouting particles according with the appearance characteristics of the sprouting particles. Similarly, after comparing the pixel points of the mildew spots and the pixel points of the disease spots by means of the characteristic pixel point coding library, the given threshold value can be compared according to the occurrence amount to make judgment. Generally, when the sprouting is obvious to change the surface color or mildew and scab spots are large, the statistical characteristics of the pixel frequency of the sprouting are greatly different from those of qualified granules, and the sprouting can be directly identified without the step.
(43) The remaining rice grain particles judged to be qualified.
The invention relates to a characteristic pixel image recognition method which is different from the current popular deep learning image recognition method based on a convolutional neural network, the image recognition method of the invention directly uses the pixel coding characteristic frequency distribution as difference recognition to assist the early and late dynamic recognition to extract the characteristic object pixel coding and establish a library, the target object in the image can be segmented and recognized step by step, the actually used pixel library space is not large, and the total amount is 24 (16) bytes of 2; meanwhile, the method has strong realizability, is directly faced with image data coding processing, is not required to carry various large open source code libraries in programming, is short in code, convenient and efficient to use, and has small computation time complexity and space complexity, so that the image recognition process is simple and efficient.
Based on the characteristic pixel image identification method, the invention also provides a characteristic pixel image identification system.
As shown in fig. 2, a characteristic pixel image recognition system includes the following modules,
the pixel frequency distribution characteristic acquisition module is used for acquiring a BMP format image of an object to be recognized to obtain an image to be recognized, extracting RGB (red, green and blue) code values of all pixel points in the image to be recognized and obtaining a pixel frequency distribution characteristic of the image to be recognized;
and the segmentation identification module is used for mining and analyzing a standard image to obtain statistical pixel frequency distribution characteristics, establishing a standard characteristic pixel point coding library through the statistical pixel frequency distribution characteristics, and performing segmentation identification on the pixel frequency distribution characteristics of the image to be identified by using the standard characteristic pixel point coding library to judge whether the image to be identified and the standard image have consistent pixel frequency distribution characteristics.
The specific functions of the above modules correspond to the specific steps in the method of the present invention.
The invention relates to a characteristic pixel image recognition system which is different from a current popular deep learning image recognition system based on a convolutional neural network, wherein the image recognition system directly performs difference recognition on pixel coding characteristic frequency distribution, assists in early-stage and later-stage dynamic recognition to extract characteristic object pixel coding and establish a library, can gradually partition and recognize a target object in an image, and has small actually used pixel library space and the total amount of 24 powers (16 million) of 2; meanwhile, the method has strong realizability, is directly faced with image data coding processing, is not required to carry various large open source code libraries in programming, is short in code, convenient and efficient to use, and has small computation time complexity and space complexity, so that the image recognition process is simple and efficient.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (10)
1. A method for recognizing a characteristic pixel image is characterized in that: comprises the following steps of (a) carrying out,
s1, acquiring a BMP format image of an object to be recognized to obtain an image to be recognized, extracting RGB (red, green and blue) code values of each pixel point in the image to be recognized to obtain pixel frequency distribution characteristics of the image to be recognized;
s2, obtaining statistical pixel frequency distribution characteristics through mining and analyzing standard images, establishing a standard characteristic pixel point coding library through the statistical pixel frequency distribution characteristics, carrying out segmentation and identification on the pixel frequency distribution characteristics of the images to be identified by utilizing the standard characteristic pixel point coding library, and judging whether the images to be identified and the standard images have consistent pixel frequency distribution characteristics.
2. The feature pixel image recognition method according to claim 1, characterized in that: specifically, the step S2 is,
s21a, mining and analyzing the standard image to obtain statistical pixel frequency distribution characteristics, establishing a standard characteristic pixel point coding library through the statistical pixel frequency distribution characteristics, and taking the subtraction value of the coding value high-frequency state frequency sum and the coding value low-frequency state frequency sum on each channel of RGB three channels in the standard characteristic pixel point coding library as an identification frequency characteristic value;
s22a, counting the subtraction value of the encoding value high-frequency state frequency sum and the encoding value low-frequency state frequency sum on each channel of RGB three channels in the pixel frequency distribution characteristics of the image to be recognized;
s23a, when the subtraction value on each channel of the RGB three channels in the pixel frequency distribution characteristics of the image to be recognized is correspondingly matched with the recognition frequency characteristic value on the corresponding channel of the RGB three channels in the standard characteristic pixel point coding library, judging that the image to be recognized and the standard image have consistent pixel frequency distribution characteristics, otherwise, judging that the image to be recognized and the standard image do not have consistent pixel frequency distribution statistical characteristics.
3. The feature pixel image recognition method according to claim 1, characterized in that: specifically, the step S2 is,
s21b, mining and analyzing the standard image to obtain statistical pixel frequency distribution characteristics, establishing a standard characteristic pixel point coding library through the statistical pixel frequency distribution characteristics, performing coding difference operation between two adjacent channels of RGB three channels of each pixel point in the standard characteristic pixel point coding library, taking a coding difference absolute value, counting the frequencies of the coding difference absolute values between the two adjacent channels of all the pixel points in the standard characteristic pixel point coding library, selecting frequencies meeting a preset rule from the frequencies of the coding difference absolute values between the two adjacent channels of all the pixel points in the standard characteristic coding pixel point library, and taking the sum of the selected frequencies as an identification frequency characteristic value;
s22b, performing coding difference operation between two adjacent channels of RGB three channels of each pixel point in the pixel frequency distribution characteristics of the image to be recognized, taking a coding difference absolute value, and counting the frequency of the coding difference absolute value between the two adjacent channels of all the pixel points in the pixel frequency distribution characteristics of the image to be recognized to obtain the frequency distribution of the coding difference of the image to be recognized;
s23b, comparing the frequency distribution of the coding difference of the image to be recognized with the recognition frequency characteristic value, when the frequency distribution of the coding difference of the image to be recognized accords with the recognition frequency characteristic value, judging that the image to be recognized and the standard image have consistent pixel frequency distribution characteristics, otherwise, judging that the image to be recognized and the standard image do not have consistent pixel frequency distribution statistical characteristics.
4. The feature pixel image recognition method according to claim 1, characterized in that: specifically, the step S2 is,
s21c, mining and analyzing the standard image to obtain statistical pixel frequency distribution characteristics, and establishing a standard characteristic pixel point coding library through the statistical pixel frequency distribution characteristics; defining a local range in the standard image, and calculating the sum of subtraction absolute values of two adjacent pixel points on corresponding channels of three RGB channels in the local range in the standard image according to the standard characteristic pixel point coding library; for the whole standard image, a local range is selected in a flowing mode until the whole standard image is covered, the sum of subtraction absolute values of two adjacent pixel points in all the local ranges in the standard image on corresponding channels of three RGB channels is calculated, and the frequency of the sum of subtraction absolute values of two adjacent pixel points in all the local ranges in the standard image on corresponding channels of three RGB channels is counted; selecting frequencies which accord with a preset rule from the frequencies of the sum of subtraction absolute values of two adjacent pixel points on corresponding channels of three RGB channels in all local ranges in the standard image, and taking the sum of the selected frequencies as a recognition frequency characteristic value;
s22c, defining a local range in the image to be recognized, and calculating the sum of subtraction absolute values of two adjacent pixel points on corresponding channels of three channels of RGB in the local range in the image to be recognized according to the pixel frequency distribution characteristics of the image to be recognized; for the whole image to be recognized, a local range is selected in a flowing mode until the whole image to be recognized is covered, the sum of subtraction absolute values of two adjacent pixel points in all the local ranges in the image to be recognized on corresponding channels of three channels of RGB is calculated, and the frequency distribution of the sum of subtraction absolute values of two adjacent pixel points in all the local ranges in the image to be recognized on corresponding channels of three channels of RGB is counted;
s23c, extracting frequencies meeting preset rules from the frequency distribution of the sum of subtraction absolute values of adjacent two pixel points on corresponding channels of three channels of RGB in all local ranges of the image to be recognized, and accumulating and calculating to obtain the frequency accumulation sum of the image to be recognized; and when the frequency accumulation sum value of the image to be recognized accords with the recognition frequency characteristic value, judging that the image to be recognized and the standard image have consistent pixel frequency distribution characteristics, otherwise, judging that the image to be recognized and the standard image do not have consistent pixel frequency distribution statistical characteristics.
5. The feature pixel image recognition method according to any one of claims 1 to 4, characterized in that: specifically, the step S1 is,
s11, acquiring a BMP format image of the object to be recognized to obtain the image to be recognized;
s12, removing the background in the image to be recognized to obtain a foreground image in the image to be recognized;
and S13, extracting RGB (red, green and blue) coding values of all pixel points of the foreground image in the image to be recognized to obtain pixel frequency distribution characteristics of the foreground image in the image to be recognized.
6. The feature pixel image recognition method according to claim 5, characterized in that: specifically, the step S12 is,
s121, shooting a background image in advance, carrying out image grid analysis on the image to be recognized by utilizing a pixel coding high-frequency state of the shot background image in advance to form an image grid set to be recognized, judging a background grid from the image grid set to be recognized, and defining the remaining image grids which are not judged as the background grids in the image grid set to be recognized as a remaining image grid set to be recognized;
s122, utilizing an accumulated background image pixel library to perform pixel identification on any to-be-identified image grid in the residual to-be-identified image grid set, marking the identified pixel position, and marking as a background pixel;
s123, when the amount of the image grids to be recognized marked as background pixels in the residual image grid set to be recognized reaches a preset proportion, judging that the image grids to be recognized marked as background pixels in the residual image grid set to be recognized are background grids;
s124, taking out the pixel code values in all the to-be-identified image grids judged as the background grids, putting the pixel code values into a background pixel library, and circularly and iteratively executing S122-S124 until no new background grids in the residual to-be-identified image grid set are identified;
and S125, removing all pixel points in the to-be-identified image grids which are judged as background grids from the to-be-identified image grid set to obtain a foreground image in the to-be-identified image.
7. The feature pixel image recognition method according to any one of claims 1 to 4, characterized in that: the statistical pixel frequency distribution characteristics are obtained by mining and analyzing the standard image, and a standard characteristic pixel point coding library is established through the statistical pixel frequency distribution characteristics,
acquiring a BMP format image of a standard object which is the same as the object to be identified to obtain a standard image;
and extracting RGB (red, green and blue) code values of all pixel points of the foreground image in the standard image to obtain pixel frequency distribution characteristics of the foreground image in the standard image, and directly extracting the RGB code values of all pixel points of the foreground image in the standard image to build a library to obtain a standard characteristic pixel point code library.
8. The feature pixel image recognition method according to any one of claims 1 to 4, characterized in that: the step S2 is followed by the step S,
s3, when the image to be recognized and the standard image are judged to have the consistent pixel frequency distribution statistical characteristics and the picture colors of the foreground image of the image to be recognized are observed to be consistent, dividing the foreground image of the image to be recognized into grids with the same size, counting high-frequency RGB (red, green and blue) coded values of each grid in the foreground image of the whole image to be recognized, calculating a T value of normal distribution according to the frequency of all the high-frequency RGB coded values in each grid in the foreground image of the whole image to be recognized, if the T value reaches a preset threshold value, judging that the corresponding grid is a normal grid of picture colors, otherwise, judging that the corresponding grid is an abnormal grid of picture colors.
9. The feature pixel image recognition method according to claim 8, characterized in that: and when the corresponding grid is judged to be the normal grid of the picture color, adding the RGB code value of each pixel point in the corresponding grid into the standard characteristic pixel point code library.
10. A feature pixel image recognition system, characterized by: comprises the following modules which are used for realizing the functions of the system,
the pixel frequency distribution characteristic acquisition module is used for acquiring a BMP format image of an object to be recognized to obtain an image to be recognized, extracting RGB (red, green and blue) code values of all pixel points in the image to be recognized and obtaining a pixel frequency distribution characteristic of the image to be recognized;
and the segmentation identification module is used for mining and analyzing a standard image to obtain statistical pixel frequency distribution characteristics, establishing a standard characteristic pixel point coding library through the statistical pixel frequency distribution characteristics, and performing segmentation identification on the pixel frequency distribution characteristics of the image to be identified by using the standard characteristic pixel point coding library to judge whether the image to be identified and the standard image have consistent pixel frequency distribution characteristics.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910974330.3A CN110837848A (en) | 2019-10-14 | 2019-10-14 | Characteristic pixel image identification method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910974330.3A CN110837848A (en) | 2019-10-14 | 2019-10-14 | Characteristic pixel image identification method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110837848A true CN110837848A (en) | 2020-02-25 |
Family
ID=69575318
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910974330.3A Withdrawn CN110837848A (en) | 2019-10-14 | 2019-10-14 | Characteristic pixel image identification method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110837848A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111814808A (en) * | 2020-07-17 | 2020-10-23 | 安徽萤瞳科技有限公司 | Method for identifying yellowing materials for rice color sorting |
| CN113012189A (en) * | 2021-03-31 | 2021-06-22 | 影石创新科技股份有限公司 | Image recognition method and device, computer equipment and storage medium |
| CN113109240A (en) * | 2021-04-08 | 2021-07-13 | 国家粮食和物资储备局标准质量中心 | Method and system for determining imperfect grains of grains implemented by computer |
| CN116645591A (en) * | 2023-05-31 | 2023-08-25 | 杭州数盒魔方科技有限公司 | Pixel value-based electronic contract seal picture PS trace identification method and system |
-
2019
- 2019-10-14 CN CN201910974330.3A patent/CN110837848A/en not_active Withdrawn
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111814808A (en) * | 2020-07-17 | 2020-10-23 | 安徽萤瞳科技有限公司 | Method for identifying yellowing materials for rice color sorting |
| CN113012189A (en) * | 2021-03-31 | 2021-06-22 | 影石创新科技股份有限公司 | Image recognition method and device, computer equipment and storage medium |
| CN113109240A (en) * | 2021-04-08 | 2021-07-13 | 国家粮食和物资储备局标准质量中心 | Method and system for determining imperfect grains of grains implemented by computer |
| CN113109240B (en) * | 2021-04-08 | 2022-09-09 | 国家粮食和物资储备局标准质量中心 | Method and system for determining imperfect grains of grains implemented by computer |
| CN116645591A (en) * | 2023-05-31 | 2023-08-25 | 杭州数盒魔方科技有限公司 | Pixel value-based electronic contract seal picture PS trace identification method and system |
| CN116645591B (en) * | 2023-05-31 | 2024-01-05 | 杭州数盒魔方科技有限公司 | Pixel value-based electronic contract seal picture PS trace identification method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110837848A (en) | Characteristic pixel image identification method and system | |
| CN109154978B (en) | System and method for detecting plant diseases | |
| CN107909138B (en) | Android platform-based circle-like particle counting method | |
| Crabb et al. | Real-time foreground segmentation via range and color imaging | |
| Soriano et al. | Skin detection in video under changing illumination conditions | |
| CN110009638B (en) | Bridge inhaul cable image appearance defect detection method based on local statistical characteristics | |
| CN113139521B (en) | Pedestrian boundary crossing monitoring method for electric power monitoring | |
| CN111784605B (en) | Image noise reduction method based on region guidance, computer device and computer readable storage medium | |
| AU2010238543B2 (en) | Method for video object detection | |
| CN106548160A (en) | A kind of face smile detection method | |
| CN108563979B (en) | Method for judging rice blast disease conditions based on aerial farmland images | |
| Almogdady et al. | A flower recognition system based on image processing and neural networks | |
| CN108280409B (en) | Large-space video smoke detection method based on multi-feature fusion | |
| CN105184812A (en) | Target tracking-based pedestrian loitering detection algorithm | |
| CN109685045A (en) | A kind of Moving Targets Based on Video Streams tracking and system | |
| WO2009105812A1 (en) | Spatio-activity based mode matching field of the invention | |
| CN106610987A (en) | Video image retrieval method, device and system | |
| CN110569859B (en) | Color feature extraction method for clothing image | |
| Saxen et al. | Color-based skin segmentation: An evaluation of the state of the art | |
| US20040240733A1 (en) | Image transmission system, image transmission unit and method for describing texture or a texture-like region | |
| CN109859236A (en) | Mobile object detection method, calculates equipment and storage medium at system | |
| Kurniawati et al. | Texture analysis for diagnosing paddy disease | |
| Alvarado-Robles et al. | An approach for shadow detection in aerial images based on multi-channel statistics | |
| Chaki et al. | Image color feature extraction techniques: fundamentals and applications | |
| EP3223240B1 (en) | Generating sparse sample histograms |
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 | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20200225 |