CN115205835A - Liquefied petroleum gas label pattern recognition system - Google Patents
Liquefied petroleum gas label pattern recognition system Download PDFInfo
- Publication number
- CN115205835A CN115205835A CN202211092220.2A CN202211092220A CN115205835A CN 115205835 A CN115205835 A CN 115205835A CN 202211092220 A CN202211092220 A CN 202211092220A CN 115205835 A CN115205835 A CN 115205835A
- Authority
- CN
- China
- Prior art keywords
- image block
- image
- vector
- label
- petroleum gas
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
Abstract
The invention discloses a liquefied petroleum gas label graph recognition system, and relates to the field of image enhancement. The method comprises the following steps: the device comprises an image acquisition unit, an image processing unit, a data processing unit and an image identification unit. The method comprises the steps of obtaining an HSV color space image of a label image of the liquefied petroleum gas, segmenting a plurality of image blocks, obtaining a dictionary vector and a first sparse vector of each image block, adjusting each feature description value in the first sparse vector of each image block to determine light reflection features, calculating light reflection texture correction coefficients of each pixel point by using the feature description values of the light reflection features, combining with required enhancement degrees, carrying out image enhancement on the label image, and identifying the label of the liquefied petroleum gas according to the enhanced label image. The invention determines the required enhancement degree of each area according to the light reflection conditions of different areas to carry out self-adaptive enhancement processing on the liquefied petroleum gas label image, so that the enhanced liquefied petroleum gas label image can be clearer, and the accuracy and the identification efficiency of pattern identification are improved.
Description
Technical Field
The invention relates to the field of image enhancement, in particular to a liquefied petroleum gas label pattern recognition system.
Background
Liquefied petroleum gas is a chemical product with high risk degree, and needs to be safely managed. Security management labels are typically applied to the lpg bottles for better management. When transportation and storage management are carried out, the distinguishing management can be carried out by identifying the graphics and characters on the label of the petroleum gas bottle body.
However, a smooth label can reflect light, and particularly, the smooth label is stuck on a cambered liquefied petroleum gas cylinder body, so that the light reflecting degree is higher, the difficulty of label identification is increased due to the influence of the light reflecting, and therefore, the image of the label influenced by the light reflecting needs to be correspondingly processed.
In the prior art, a neural network is mostly constructed, a trained neural network is obtained by training the neural network, then an image to be enhanced is input into the neural network, an enhanced image is finally output, and the graph and character in the enhanced label image are further identified, so that the graph and character on the label are obtained to store liquefied petroleum gas, however, the label images are not identical in light reflection degree of each region, the whole label image is enhanced by using the neural network, a large amount of image data is needed, a large amount of time is consumed in the training process, and the image enhancement is performed through the neural network, so that the accuracy of the identification result of the enhanced image cannot be guaranteed.
Disclosure of Invention
In view of the above technical problems, the present invention provides a liquefied petroleum gas label pattern recognition system, which specifically includes:
an image acquisition unit: collecting a label image on the liquefied petroleum gas tank by using a camera;
an image processing unit: the label image acquisition unit is used for converting the label image acquired by the image acquisition unit into an HSV color space image, uniformly dividing the HSV color space image to obtain a plurality of image blocks and acquiring a vector of each image block;
a data processing unit: the image processing unit is used for inputting all image block vectors obtained by the image processing unit into the K-SVD network to obtain a dictionary matrix and each first sparse vector corresponding to each image block;
adjusting each feature description value in the first sparse vector to obtain a sparse vector of each image block after each adjustment, and acquiring a feature description value sequence of the sparse vector of each image block after multiple adjustments;
taking a difference vector of the sparse vector of each image block after each adjustment and the first sparse vector as a difference image block vector of each image block under each adjustment, and calculating the model lengths of three channel vectors in the difference image block vector of each image block under each adjustment to obtain a model length sequence of three channels of each difference image block;
obtaining three Pearson correlation coefficients of each feature description value in each image block by using the feature description value sequence of each image block and the module length sequences of three channels of the difference image block vector of the image block;
calculating the reflecting feature conformity degree of each feature description value in each image block by using the three Pearson correlation coefficients of each feature description value in each image block, and determining all reflecting features of each image block;
calculating the required enhancement degree of each image block according to each feature description value of the sparse vector of each image block and the corresponding light reflection feature conformity degree of the feature description value;
adjusting the image blocks by utilizing all light reflection characteristics of each image block, obtaining adjusted image blocks by combining a dictionary matrix, carrying out difference on each image block and the corresponding adjusted image block to obtain adjusted difference image blocks, and calculating the contrast of each pixel point in each adjusted difference image block;
calculating the reflecting texture conforming degree of each pixel point according to the contrast of each pixel point in each adjusting difference image block;
obtaining an enhancement coefficient of each pixel point in each image block according to the required enhancement degree of each image block and the reflecting texture conforming degree of each pixel point in each image block;
an image enhancement unit: the label image of the liquefied petroleum gas is enhanced by the aid of the enhancement coefficient of each pixel point in each image block obtained by the data processing unit, and the label image enhanced by the liquefied petroleum gas is obtained;
an image recognition unit: the label information recognition unit is used for recognizing the label image obtained by the image enhancement unit after the liquefied petroleum gas is enhanced.
The process of obtaining the sparse vector of each image block after each adjustment in the data processing unit is as follows:
adjusting each feature description value in the first sparse vector of each image block, wherein the adjusting method is to add 1 to the original feature description value on the basis of the last adjustment, and set the adjusting times to complete multiple adjustments of each feature description value in the first sparse vector of each image block;
keeping other description values of the sparse vector corresponding to each image block unchanged, and obtaining the sparse vector of each image block after each adjustment, namely the sparse vector of each image block after each adjustment.
The process of acquiring the Pearson correlation coefficient in the data processing unit is as follows:
acquiring each feature description value sequence of the sparse vector of each image block after each adjustment;
acquiring a sparse vector of each image block after each adjustment and a first sparse vector of each image block, and taking a difference vector of the sparse vector of each image block after each adjustment and the first sparse vector as a difference image block vector of each image block under each adjustment to obtain a difference image block vector of each image block under each adjustment;
acquiring a vector of a difference image block vector of each image block under each adjustment in an H channel/S channel/V channel, and calculating the modular length of the vector of the difference image block vector of each image block under each adjustment in the H channel/S channel/V channel to obtain a modular length sequence of the vector of the difference image block vector of each image block under each adjustment in the H channel/S channel/V channel;
and respectively calculating a model length sequence of the difference image block vector of each image block in the H channel/S channel/V channel under each adjustment, and a Pearson correlation coefficient of each feature description value sequence of the sparse vector of each image block after each adjustment to obtain each feature description value sequence of the sparse vector of each image block after each adjustment, and respectively corresponding to three Pearson correlation coefficients of the model length sequences of three channels of the difference image block vector.
The method for adjusting the image blocks by utilizing all the light reflection characteristics of each image block and combining the dictionary matrix in the data processing unit comprises the following steps:
and zeroing the feature description values of all the light reflection features in the first sparse vector of each image block to obtain a second sparse vector, and multiplying the second sparse vector of each image block by the dictionary matrix to adjust each image block to obtain an adjusted image block of each image block after adjustment.
The method for determining all the light reflection characteristics in each image block in the data processing unit comprises the following steps:
setting a light reflection threshold, taking the characteristic corresponding to the characteristic description value with the light reflection characteristic conformity degree smaller than the light reflection threshold as a non-light reflection characteristic, and taking the characteristic corresponding to the characteristic description value with the light reflection characteristic conformity degree larger than the light reflection threshold as a light reflection characteristic.
The method of calculating the required degree of enhancement for each image block in the data processing unit is as follows:
taking the sum of products of each feature description value of the first sparse vector of each image block and the reflection conformity degree of the feature corresponding to each feature description value as the required enhancement degree of the image block;
the method for calculating the light reflection conformity degree of the characteristic corresponding to each characteristic description value comprises the following steps:
respectively taking the feature description values with the same sequence number as the dictionary matrix in the sparse vectors of all the image blocks after adjustment at each time and the mean value of the Pearson correlation coefficient of the modular length sequence of each channel as the correlation coefficient of the feature corresponding to the feature description value in each channel;
and taking the absolute value of the correlation coefficient of the feature corresponding to each feature description value in the V channel, the quotient of the correlation coefficient of the feature corresponding to each feature description value in the S channel and the sum of the correlation coefficient of the feature corresponding to each feature description value in the H channel and the zero-preventing coefficient as the light reflection coincidence degree of the feature corresponding to each feature description value.
The process of calculating the reflecting texture conformity degree of each pixel point in the data processing unit is as follows:
subtracting each image block and the corresponding adjusting image block to obtain an adjusting difference image block, and taking the variance of the gray values of each pixel point in the adjusting difference image block and the eight neighborhood pixel points as the contrast of the pixel point to obtain the contrast of each pixel point in the adjusting difference image block;
and clustering the contrast of each pixel point in the adjusted difference image block according to the category 2 to obtain the mean value of the contrasts in the two categories, taking the minimum contrast in the category with the maximum mean value as a reflection texture basic value, and taking the ratio of each pixel point in each adjusted difference image block to the reflection texture basic value as the reflection texture conformity degree of the pixel point in the adjusted difference image block.
The method for obtaining the enhancement coefficient of each pixel point in each image block in the data processing unit comprises the following steps:
calculating the product of the reflection texture correction coefficient and the reflection enhancement degree of each pixel point in each image block, and performing normalization processing on the product to serve as the enhancement coefficient of the pixel point in the image block;
and obtaining the enhancement coefficient of each pixel point in each image block.
The process of identifying the label information of the label image which is obtained by the image enhancement unit after the liquefied petroleum gas enhancement in the image identification unit is as follows:
performing image enhancement on the label image of the liquefied petroleum gas according to the gray value of each pixel point in each image block, the number of the pixel points in each image block and the enhancement coefficient of each pixel point in each image block to obtain the label image of the liquefied petroleum gas after enhancement;
and matching the label image after the enhancement of the liquefied petroleum gas with a standard label image of the liquefied petroleum gas in a database by using a template, and identifying the label information by using the template type with the maximum matching degree as the label information type.
Compared with the prior art, the embodiment of the invention has the beneficial effects that:
1. the smooth liquefied petroleum gas label is pasted on the liquefied petroleum gas tank with the cambered surface, so that the collected liquefied petroleum gas label image can reflect light, the identification of the liquefied petroleum gas label can be influenced by the reflection light, the required enhancement degree of each region is determined according to the reflection conditions of different regions, the liquefied petroleum gas label image after enhancement is subjected to self-adaptive enhancement processing, and the enhanced liquefied petroleum gas label image is clearer, so that the accuracy and the identification efficiency of pattern identification are improved.
2. Performing dictionary training by using a K-SVD algorithm in a dictionary training algorithm to obtain a sparse vector of each area in the liquefied petroleum gas label image, adjusting each feature description value of the sparse vector to obtain an adjusted image block, calculating the required enhancement degree of each image block (namely each area of the image) by analyzing the difference image block vector of the adjusted image block and the original image block, and performing image analysis by using the K-SVD algorithm as a basis to obtain the required enhancement degree of each image block more accurately with very small error.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the embodiments or the description of the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a block diagram of a system provided with a liquefied petroleum gas label pattern recognition system according to embodiment 1 of the present invention;
fig. 2 is a flowchart of a method provided by a liquefied petroleum gas label pattern recognition system according to embodiment 2 of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In the following description, for purposes of explanation and not limitation, specific details are set forth, such as particular system structures, techniques, etc. in order to provide a thorough understanding of the embodiments of the present application. It will be apparent, however, to one skilled in the art that the present application may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present application with unnecessary detail.
The terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature; in the description of the present embodiment, "a plurality" means two or more unless otherwise specified.
Example 1
The embodiment of the invention provides a liquefied petroleum gas label pattern recognition system, as shown in fig. 1, the specific contents include:
the embodiment comprises an image acquisition unit, an image processing unit, a data processing unit, an image enhancement unit and an image identification unit. The label image required in the embodiment is acquired through the image acquisition unit, the image acquired by the image acquisition unit is transmitted to the image processing unit to perform image processing on the label image, the image information in the image after each processing is extracted, the data processing unit is used for receiving the image information extracted by the image processing unit, the image information extracted by the image processing unit is analyzed and processed in the data processing unit, the gray value of each pixel point in each image block after the label image is enhanced is obtained according to the data analyzed and processed by the data processing unit, the obtained gray value of each pixel point in each image block after the pixel point is enhanced is transmitted to the image enhancement unit to output the enhanced label image, and the enhanced label image is transmitted to the image identification unit to perform image identification. The method specifically comprises the following steps:
an image acquisition unit: collecting a label image on the liquefied petroleum gas tank by using a camera;
an image processing unit: the label image acquisition unit is used for converting the label image acquired by the image acquisition unit into an HSV color space image, uniformly dividing the HSV color space image to obtain a plurality of image blocks and acquiring a vector of each image block;
a data processing unit: the image processing unit is used for inputting all image block vectors obtained by the image processing unit into the K-SVD network to obtain a dictionary matrix and each first sparse vector corresponding to each image block;
adjusting each feature description value in the first sparse vector to obtain a sparse vector of each image block after each adjustment, and acquiring a feature description value sequence of the sparse vector of each image block after multiple adjustments;
taking a difference vector of the sparse vector and the first sparse vector of each image block after each adjustment as a difference image block vector of each image block under each adjustment, and calculating the module length of three channel vectors in the difference image block vector of each image block under each adjustment to obtain a module length sequence of three channels of each difference image block;
obtaining three Pearson correlation coefficients of each feature description value in each image block by using the feature description value sequence of each image block and the modular length sequences of three channels of the difference image block vector of the image block;
calculating the reflecting characteristic conformity degree of each feature description value in each image block by using the three Pearson correlation coefficients of each feature description value in each image block, and determining all reflecting characteristics of each image block;
calculating the required enhancement degree of each image block according to each feature description value of the sparse vector of each image block and the corresponding light reflection feature conformity degree of the feature description value;
adjusting the image blocks by utilizing all light reflection characteristics of each image block, obtaining adjusted image blocks by combining a dictionary matrix, carrying out difference on each image block and the corresponding adjusted image block to obtain adjusted difference image blocks, and calculating the contrast of each pixel point in each adjusted difference image block;
calculating the reflecting texture conforming degree of each pixel point according to the contrast of each pixel point in each adjusting difference value image block;
obtaining an enhancement coefficient of each pixel point in each image block according to the required enhancement degree of each image block and the reflecting texture conforming degree of each pixel point in each image block;
an image enhancement unit: the label image of the liquefied petroleum gas is enhanced by the enhancement coefficient of each pixel point in each image block obtained by the data processing unit, so that the label image of the liquefied petroleum gas after enhancement is obtained;
an image recognition unit: the label information recognition unit is used for recognizing the label image obtained by the image enhancement unit after the liquefied petroleum gas is enhanced.
Example 2
The embodiment of the invention provides a liquefied petroleum gas label pattern recognition system, as shown in fig. 2, the specific contents include:
s101, acquiring an HSV color space image of the label image:
in the embodiment, label information identification is performed on a label image on a liquefied petroleum gas bottle body, so that the label image of the liquefied petroleum gas needs to be acquired at first, but because the liquefied petroleum gas label is pasted on a cambered liquefied petroleum tank under normal conditions and is easy to reflect light, the acquired definition of the partial area of the liquefied petroleum gas label image is low, the liquefied petroleum gas label image needs to be enhanced, and a large relation exists between the reflected image and the brightness, so that the acquired label image needs to be converted into an HSV color space image for analysis.
And arranging a camera right above the liquefied petroleum gas tank to acquire a label image on the liquefied petroleum gas tank body, and performing color space conversion on the acquired label image to obtain an HSV color space image of the label image.
S102, dictionary training is carried out on each image block:
the reason for the reflection is that the light quantity reflected by each area to the camera is different, and the light quantity reflected by some areas to the camera is larger, so that the brightness presented by the area is larger, therefore, the embodiment divides the HSV color space image of the label image to obtain a plurality of image blocks, and the image blocks with different reflection degrees are enhanced to different degrees by analyzing the brightness condition of each image block.
1. A plurality of image blocks of an HSV color space image of a label image are acquired.
Uniformly dividing HSV color space image of label image into a pixel size of a
Image block, in the present embodiment
Is set to 100, the first
An image block is marked as
The image block set composed of all image blocks is recorded as
。
2. And performing dictionary training on each image block.
Obtaining the vector of all image blocks to form an image block vector set
The vectors of all image blocks form an image block vector set as the input of a K-SVD (dictionary training algorithm) network, and the output is a dictionary matrix
And combining the first sparse vector of each image block according to the sequence in the image block set to obtain a first sparse vector set, and recording the first sparse vector set as a first sparse vector set
. Wherein, the first
First sparse vector of each image block is recorded as
Of 1 at
The image blocks can be represented as
Wherein the dictionary matrix is composed of
A dictionary is loud and composed of
A dictionary vector is noted
。
The image block vector set is obtained by expanding the pixel values of the H channel, the S channel, and the V channel of each image block, and the size of the image block set according to the present embodiment is known, where the size of the image block is known
Dimension is developed by image block H channelThe vector of arrival is then calculated,
the dimension is a vector obtained by expanding the S channel of the image block,
and the dimension is a vector obtained by expanding the V channel of the image block.
S103, calculating the light reflection conforming degree of each characteristic:
because the label image on the liquefied petroleum gas tank is unclear mainly due to the situation of light reflection, so that the identification of the label image is influenced, and therefore the light reflection part needs to be subjected to higher-intensity enhancement processing, firstly, the light reflection coincidence degree of all the characteristics in the image block is calculated, the light reflection characteristics are determined according to the light reflection coincidence degree of all the characteristics, the enhancement degree required by each image block is determined according to the characteristic description value of the light reflection characteristics in each image block, and therefore, the light reflection coincidence degree of each characteristic is calculated by the step.
For convenience of description, each dictionary vector is recorded as a feature, and a vector value of a first sparse vector corresponding to each dictionary vector is recorded as a feature description value of each feature, but the features are more, and not all the features describe the situation of light reflection, so that the feature description value of each feature is adjusted to analyze the coincidence degree of each feature as a light reflection feature according to the change situation in the adjusted image block, and the specific process is as follows:
the following analysis processing is carried out on each image block in the HSV color space image of the label image on the liquefied petroleum gas tank:
1. and obtaining the modular length of each color channel vector in the difference image block vector after each adjustment.
Will be first
First sparse vector of image block
(dictionary vectors and corresponding secondSequence number of a sparse vector) feature description values
Is adjusted to
Keep at first
The first sparse vector of each image block has other feature description values (except for
Individual feature description) is not changed, and a first-time adjusted sparse vector is obtained
Then the first time after adjustment
Each image block is represented as
To obtain the first adjusted
Each image block;
after the first adjustment
Each image block
Image block vector of
And the image block before adjustment (here, the second image block without adjustment)
An original image block
) Image block vector of
The pixel values of all channels are correspondingly subtracted to obtain a difference image block vector
；
Obtaining the H (tone) channel vector in the difference image block vector
Dimensional data) for calculating the modulo length of the H (hue) channel vector
Obtaining S (saturation) channel vector in difference image block vector (
Dimensional data), calculate the modular length of the S (saturation) channel vector
Obtaining the V (brightness) channel vector in the difference image block vector (
Dimensional data), calculating the modulus length of the V (luminance) channel vector
；
Will be first
First sparse vector of image block
(dictionary vectors and corresponding first DiluentsSequence number of sparse vectors) feature description values
Is adjusted to
Keep at first
Other feature description values of the first adjusted sparse vector of the image block (except for
Feature description) is not changed, and a second-time adjusted sparse vector is obtained
Then after the second adjustment
Each image block is represented as
To obtain the second adjusted
Each image block;
after the second adjustment
An image block
Image block vector of
And the image block before adjustment (here, the first adjusted image block)
An image block
) Image block vector of
The pixel values of all channels are correspondingly subtracted to obtain a difference image block vector
；
Obtaining the H (tone) channel vector in the difference image block vector
Dimensional data) for calculating the modulo length of the H (hue) channel vector
Obtaining S (saturation) channel vector in difference image block vector (
Dimensional data), calculate the modular length of the S (saturation) channel vector
Obtaining V (brightness) channel vector in difference image block vector (
Dimensional data), calculating the modulus length of the V (luminance) channel vector
；
Adjusting the image block according to the method of obtaining the modular length of each color channel vector in the difference image block vector after each adjustment in S103, where the adjusting method is to add 1 to the original feature description value on the basis of the last adjustment, and set the adjusting times, which is set to 20 in this embodiment (that is, the finally adjusted image block is represented as the image block after final adjustment
) And obtaining the adjusted difference image block vector
Length of mode in H (hue) channel vector
Length of mode in S (saturation) channel vector
And the modulus length in the V (luminance) channel vector
；
According to the method for obtaining the modular length of each color channel vector in the difference image block vector after each adjustment in S103, the feature description values of other features are adjusted to complete the adjustment of the second step
Adjusting each feature description value in a first sparse vector set of each image block for multiple times;
2. and calculating the correlation coefficient of the feature corresponding to each feature description value under each channel.
Obtained according to step 1 in S103
First sparse vector of image block
The modulo length sequence of each color channel vector after 20 adjustments of the feature descriptor, i.e. the modulo length sequence of the H (hue) channel vector
Sequence of modular lengths of the vector in the S (saturation) channel
And a sequence of modulo lengths in the V (luminance) channel vector
And get the first
First sparse vector of image block
A feature description value sequence of 20 adjusted feature description values
；
Calculate the first
First sparse vector of image block
The technology is common knowledge, and is used for measuring the similarity between two vectors, and in the embodiment, the feature description value sequence after 20 times of adjustment of the feature description values and the Pearson correlation coefficient of the H (hue) channel vector modular length sequence, the S (saturation) channel vector modular length sequence and the V (brightness) channel vector modular length sequence are obtained without excessive explanation
First sparse vector of image block
Correlation coefficient of characteristic corresponding to characteristic description value in H (tone) channel
Correlation in S (saturation) channelNumber of
And the correlation coefficient of the V (luminance) channel
；
Obtaining correlation coefficients of the features corresponding to each feature description value of the first sparse vector of each image block on the three color channels according to the method in the step 2 in the S103;
3. calculating the degree of reflective conformity of each feature.
Obtaining the correlation coefficients of the features corresponding to the feature description values of the first sparse vector corresponding to the serial number of each image block obtained in the previous step on the three color channels, and calculating the correlation coefficients of the features corresponding to the feature description values of the first sparse vector corresponding to the serial number of each image block obtained in the previous step
The characteristic corresponding to each characteristic description value is taken as an example:
calculating the number of each image block
The mean value of the correlation coefficients of the characteristic corresponding to the characteristic description value in the H (tone) channel is used as the second value
H (hue) channel correlation coefficient of feature corresponding to feature description value
Calculating the number of each image block
The mean value of the correlation coefficient of the feature corresponding to the feature description value in the S (tone) channel is used as the second value
S (tone) channel correlation coefficient of feature corresponding to feature description value
Calculating the number of each image block
The mean value of the correlation coefficient of the feature corresponding to the feature description value in the V (tone) channel is used as the second value
V (hue) channel correlation coefficient of feature corresponding to feature description value
；
The calculation formula of the degree of light reflection coincidence of each feature is as follows:
in the formula:
is shown as
The degree of light reflection coincidence of the characteristic corresponding to each characteristic description value,
is as follows
The V (tone) channel correlation coefficient of the feature corresponding to each feature description value,
is as follows
The S (tone) channel correlation coefficient of the feature corresponding to each feature description value,
is as follows
The H (hue) channel correlation coefficient of the feature corresponding to each feature description value,
is a zero-proof coefficient; the reflective feature should have a change in reflective feature that only causes a change in brightness of the image, and does not cause a change in saturation and chroma, when
The larger the value, the more the change of the feature description value causes the relevant change of the pixels of the lightness channel, but does not cause the relevant change of the pixels of the chroma channel and the saturation channel, thereby
The degree of conforming to the reflective characteristics is greater. When in use
The smaller the value of the feature description is, the less the change of the feature description value is, the less the pixels of the name channel are subjected to relevant change, and the more the pixels of the chroma channel and the saturation channel are subjected to relevant change.
S104, calculating the required enhancement degree of each image block:
for each image block, the larger the proportion of the reflective features in the image block is, the larger the enhancement degree required by the image block is, and the smaller the proportion of the reflective features in the image block is, the smaller the enhancement degree required by the image block is, so that the required enhancement degree of each image block is calculated according to the reflective features obtained in the step S103, the required enhancement degree of each image block is used as the basis of each pixel point in the image block, and the enhancement coefficient of each pixel point is determined through subsequent processing to complete the enhancement processing of the label image.
The required enhancement degree of the image block is calculated as follows:
in the formula:
denotes the first
The desired degree of enhancement of an image block,
is shown as
The degree of light reflection coincidence of the characteristic corresponding to each characteristic description value,
is shown as
The first in the image block
The value of the individual characteristic description is,
a serial number representing a characteristic description value,
representing the number of feature description values; when an image block is enhanced, if the reflection degree of the image block is low, the degree required to be enhanced is low, so that the enhancement degree required by the image block is calculated according to the reflection conformity degree of the reflection characteristic in each image block and the characteristic description value of the characteristic, when the reflection conformity degree is low and the characteristic description value is small, the enhancement degree required by the image block is low, and otherwise, the enhancement degree required by the image block is highThe degree of enhancement.
S105, calculating a correction coefficient of the reflective texture:
since the reflective texture generally has a pixel with a large reflective difference, that is, the pixel and surrounding pixels have a large reflective difference, and thus a large pixel value difference exists between the two pixels, this embodiment calculates the enhancement coefficient of each pixel in the image block by calculating the correction coefficient of the reflective texture, calculates the enhancement coefficient of each pixel in the image block by comparing each pixel with the neighboring pixels, and enhances each pixel according to the enhancement coefficient of each pixel, so the correction coefficient of the reflective texture is calculated first in this step.
First, the reflection characteristics of each image block are obtained, and the reflection threshold is set according to the embodiment
When it comes to
When the characteristic description value corresponds to the characteristic, the characteristic is the light reflection characteristic,
then, the characteristic corresponding to the characteristic description value is a non-reflective characteristic, and all reflective characteristics in each image block are obtained according to the judging method;
and zeroing the feature description values of all the light reflection features in the first sparse vector of each image block to obtain a second sparse vector, multiplying the second sparse vector of each image block by the dictionary matrix, and adjusting each image block to obtain an adjusted image block and an adjusted image block vector of each image block after adjustment. To a first order
Each image block is taken as an example: obtaining an adjusted image block of each image block
And adjusting the image block vector
Wherein
To block the original image
And adjusting the image block
Obtaining an adjusted difference image block by differencing corresponding pixels
And adjusting each pixel value in the difference image block to reflect the change condition of each pixel value after the reflection characteristic is removed, wherein the pixel value is changed greatly only after the reflection characteristic at the reflection pixel is removed, and the pixel value is not changed greatly after the reflection characteristic at the non-reflection pixel is removed.
Because the reflection texture generally has pixel points with larger reflection difference, that is, the pixel points have larger reflection difference with surrounding pixel points, so that larger pixel value difference exists between the two pixel points, the adjustment difference image blocks of other image blocks are obtained by calculation in the same way, and the contrast value of each pixel in the adjustment difference image block is obtained, and the specific contrast value calculation method is as follows: taking the variance of the gray values of each pixel point in the adjusted difference image block and the eight neighborhood pixel points thereof as the contrast of the pixel point to obtain the contrast of each pixel point in the adjusted difference image block;
clustering the contrast of each pixel point in the adjusted difference image block according to the category 2 to obtain the average value of the contrasts in the two categories, and taking the minimum contrast in the category with the maximum average value as the reflection texture basic value
Taking the ratio of each pixel point in each adjustment difference image block to the reflection texture basic value as the adjustment difference imageThe reflecting texture conformity degree of the pixel point in the block is adjusted, and the reflecting texture conformity degree calculation formula of each pixel point in the difference image block is adjusted as follows:
in the formula:
is shown as
In a block of an image
The degree of conformity of the reflection textures of the individual pixel points,
is shown as
In a block of an image
The contrast of each pixel point is determined,
is the reflection texture basic value;
the larger the pixel point is, the larger the junction between the reflective pixel point and the non-reflective pixel point is, the higher the possibility that the pixel point has reflective texture is.
S106, performing image enhancement on the label image:
in order to ensure the accuracy of label identification, the acquired label images on the liquefied petroleum gas tank need to be subjected to self-adaptive enhancement processing, so that the enhancement coefficient of each pixel point is calculated according to different light reflection degrees to enhance each pixel point, the purpose of self-adaptively enhancing the label images on the liquefied petroleum gas tank is achieved, the enhanced label images on the liquefied petroleum gas tank achieve a better enhancement effect, and the accuracy of label identification is ensured.
1. And calculating the enhancement coefficient of each pixel point in each image block.
Firstly, according to the reflective texture conformity degree of each pixel point in each image block, obtaining the reflective texture correction coefficient of each pixel point in each image block as follows:
wherein:
is shown as
In a block of an image
The reflection texture correction coefficient of each pixel point,
is shown as
In a block of an image
The degree of conformity of the reflection textures of the pixel points; when the reflective texture coincidence degree of a pixel point is higher, the possibility that the pixel point has the reflective texture is higher, the pixel point at the reflective texture position basically does not need to be enhanced, because the reflective texture is not the image texture, and only other information of the image can be interfered after the reflective texture is enhanced, so the reflective texture correction coefficient value is lower.
Calculating the product of the reflection texture correction coefficient and the reflection enhancement degree of each pixel point in each image block, and performing normalization processing on the product to serve as the enhancement coefficient of the pixel point in the image block, wherein the calculation formula of the enhancement coefficient of the pixel point is as follows:
in the formula:
is shown as
In each image block
The enhancement coefficient of each pixel point is determined,
is shown as
In a block of an image
The reflection texture correction coefficient of each pixel point,
denotes the first
The desired degree of enhancement of an image block,
the enhancement coefficient of each pixel point is normalized by the hyperbolic tangent function, the enhancement coefficient of each pixel point is determined according to the required enhancement degree of each image block and the reflective texture correction coefficient of each pixel point, and the normalization processing is performed to keep the result between (0) and (1), so that the calculation amount is reduced.
2. And performing image enhancement on the label image.
Performing self-adaptive enhancement on each pixel point according to the gray value of each pixel point in each image block and the enhancement coefficient, wherein the calculation formula of the gray value of each pixel point in each image block in the enhanced label image is as follows:
in the formula:
for the enhanced label image
In a block of an image
The gray value of each pixel point is calculated,
is the first in the label image
In a block of an image
The gray value of each pixel point is calculated,
is shown as
In a block of an image
The enhancement coefficient of each pixel point is calculated,
which represents the minimum value of the sum of the values,
the maximum value is represented by the number of bits,
the conventional technical means of the skilled person is to modify the gray value by utilizing gray stretching to achieve the purpose of enhancing the image for the number of the pixel points contained in each image block,
for the gray stretching algorithm, an enhancement coefficient is added to calculate the gray value of each pixel point after the gray value change (i.e. the gray value of the pixel point after the image enhancement).
And calculating the gray value of each pixel point in each image block in the enhanced label image after gray stretching of each pixel point in each image block according to the method of S106, completing enhancement processing on the label image, and obtaining the label image of the liquefied petroleum gas after the label image is enhanced.
S107, identifying label information:
and template matching is carried out on the enhanced label image and the standard label image in the database, the template type of the standard label image with the maximum matching degree with the enhanced label image in the database is used as the label information type, the process of identifying the label of the liquefied petroleum gas is completed according to the label information type, the text content on the label is obtained, the liquefied petroleum gas tank is correspondingly transported and stored, and the safety management of the liquefied petroleum gas tank is ensured.
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 (9)
1. Liquefied petroleum gas label figure identification system, its characterized in that includes:
an image acquisition unit: collecting a label image on the liquefied petroleum gas tank by using a camera;
an image processing unit: the label image acquisition unit is used for converting the label image acquired by the image acquisition unit into an HSV color space image, uniformly dividing the label image to obtain a plurality of image blocks and acquiring a vector of each image block;
a data processing unit: the image processing unit is used for inputting all image block vectors obtained by the image processing unit into the K-SVD network to obtain a dictionary matrix and each first sparse vector corresponding to each image block;
adjusting each feature description value in the first sparse vector to obtain a sparse vector of each image block after each adjustment, and acquiring a feature description value sequence of the sparse vector of each image block after multiple adjustments;
taking a difference vector of the sparse vector of each image block after each adjustment and the first sparse vector as a difference image block vector of each image block under each adjustment, and calculating the model lengths of three channel vectors in the difference image block vector of each image block under each adjustment to obtain a model length sequence of three channels of each difference image block;
obtaining three Pearson correlation coefficients of each feature description value in each image block by using the feature description value sequence of each image block and the module length sequences of three channels of the difference image block vector of the image block;
calculating the reflecting characteristic conformity degree of each feature description value in each image block by using the three Pearson correlation coefficients of each feature description value in each image block, and determining all reflecting characteristics of each image block;
calculating the required enhancement degree of each image block according to each feature description value of the sparse vector of each image block and the light reflection feature conformity degree of the corresponding feature description value;
adjusting the image blocks by utilizing all light reflection characteristics of each image block, obtaining adjusted image blocks by combining a dictionary matrix, carrying out difference on each image block and the corresponding adjusted image block to obtain adjusted difference image blocks, and calculating the contrast of each pixel point in each adjusted difference image block;
calculating the reflecting texture conforming degree of each pixel point according to the contrast of each pixel point in each adjusting difference image block;
obtaining an enhancement coefficient of each pixel point in each image block according to the required enhancement degree of each image block and the reflecting texture conforming degree of each pixel point in each image block;
an image enhancement unit: the label image of the liquefied petroleum gas is enhanced by the aid of the enhancement coefficient of each pixel point in each image block obtained by the data processing unit, and the label image enhanced by the liquefied petroleum gas is obtained;
an image recognition unit: the label information recognition unit is used for recognizing the label image obtained by the image enhancement unit after the liquefied petroleum gas is enhanced.
2. The liquefied petroleum gas label pattern recognition system as claimed in claim 1, wherein the data processing unit obtains the sparse vector of each image block after each adjustment as follows:
adjusting each feature description value in the first sparse vector of each image block, wherein the adjusting method is to add 1 to the original feature description value on the basis of the last adjustment, and set the adjusting times to complete multiple adjustments of each feature description value in the first sparse vector of each image block;
keeping other description values of the sparse vector corresponding to each image block unchanged, and obtaining the sparse vector of each image block after each adjustment, namely the sparse vector of each image block after each adjustment.
3. The liquefied petroleum gas label pattern recognition system as claimed in claim 1, wherein the data processing unit obtains the Pearson correlation coefficient by:
acquiring each feature description value sequence of the sparse vector of each image block after each adjustment;
acquiring a sparse vector of each image block after each adjustment and a first sparse vector of each image block, and taking a difference vector of the sparse vector of each image block after each adjustment and the first sparse vector as a difference image block vector of each image block under each adjustment to obtain a difference image block vector of each image block under each adjustment;
obtaining a vector of a difference image block vector of each image block under each adjustment in an H channel/S channel/V channel, and calculating the modular length of the vector of the difference image block vector of each image block under each adjustment in the H channel/S channel/V channel to obtain a modular length sequence of the vector of the difference image block vector of each image block under each adjustment in the H channel/S channel/V channel;
and respectively calculating the module length sequence of the difference image block vector of each image block under each adjustment in the H channel/S channel/V channel, and the Pearson correlation coefficient of each feature description value sequence of the sparse vector of each image block after each adjustment to obtain each feature description value sequence of the sparse vector of each image block after each adjustment, and respectively corresponding to the three Pearson correlation coefficients of the module length sequences of the three channels of the difference image block vector.
4. The liquefied petroleum gas label pattern recognition system as claimed in claim 1, wherein the method for adjusting each image block by using all light reflection characteristics of the image block and combining the dictionary matrix to obtain the adjusted image block by the data processing unit comprises:
and zeroing the feature description values of all the light reflection features in the first sparse vector of each image block to obtain a second sparse vector, and multiplying the second sparse vector of each image block by the dictionary matrix to adjust each image block to obtain an adjusted image block of each image block after adjustment.
5. The liquefied petroleum gas label pattern recognition system as claimed in claim 1, wherein the data processing unit determines all the light reflection features in each image block by:
setting a light reflection threshold value, taking the characteristic corresponding to the characteristic description value with the light reflection characteristic conformity degree smaller than the light reflection threshold value as a non-light reflection characteristic, and taking the characteristic corresponding to the characteristic description value with the light reflection characteristic conformity degree larger than the light reflection threshold value as a light reflection characteristic.
6. The liquefied petroleum gas label pattern recognition system as claimed in claim 1, wherein the data processing unit calculates the required degree of enhancement for each image block as follows:
taking the sum of products of each feature description value of the first sparse vector of each image block and the reflecting coincidence degree of the feature corresponding to each feature description value as the required enhancement degree of the image block;
the method for calculating the light reflection conformity degree of the characteristic corresponding to each characteristic description value comprises the following steps:
respectively taking the feature description values with the same sequence number as the dictionary matrix in the sparse vectors of all the image blocks after adjustment at each time and the mean value of the Pearson correlation coefficient of the modular length sequence of each channel as the correlation coefficient of the feature corresponding to the feature description value in each channel;
and taking the absolute value of the correlation coefficient of the feature corresponding to each feature description value in the V channel, the quotient of the correlation coefficient of the feature corresponding to each feature description value in the S channel and the sum of the correlation coefficient of the feature corresponding to each feature description value in the H channel and the zero-preventing coefficient as the light reflection coincidence degree of the feature corresponding to each feature description value.
7. The liquefied petroleum gas label pattern recognition system as claimed in claim 1, wherein the data processing unit calculates the degree of conformity of the reflective texture of each pixel as follows:
subtracting each image block and the corresponding adjusting image block to obtain an adjusting difference image block, and taking the variance of the gray values of each pixel point in the adjusting difference image block and the eight neighborhood pixel points as the contrast of the pixel point to obtain the contrast of each pixel point in the adjusting difference image block;
and clustering the contrast of each pixel point in the adjusted difference image block according to the category 2 to obtain the mean value of the contrasts in the two categories, taking the minimum contrast in the category with the maximum mean value as a reflection texture basic value, and taking the ratio of each pixel point in each adjusted difference image block to the reflection texture basic value as the reflection texture conformity degree of the pixel point in the adjusted difference image block.
8. The liquefied petroleum gas label pattern recognition system as claimed in claim 1, wherein the method for the data processing unit to obtain the enhancement coefficient of each pixel point in each image block is as follows:
calculating the product of the reflection texture correction coefficient and the reflection enhancement degree of each pixel point in each image block, and performing normalization processing on the product to serve as the enhancement coefficient of the pixel point in the image block;
and obtaining the enhancement coefficient of each pixel point in each image block.
9. The liquefied petroleum gas label pattern recognition system as claimed in claim 1, wherein the image recognition unit performs label information recognition on the liquefied petroleum gas enhanced label image obtained by the image enhancement unit as follows:
performing image enhancement on the label image of the liquefied petroleum gas according to the gray value of each pixel point in each image block, the number of the pixel points in each image block and the enhancement coefficient of each pixel point in each image block to obtain the label image of the liquefied petroleum gas after enhancement;
and matching the label image after the enhancement of the liquefied petroleum gas with a standard label image of the liquefied petroleum gas in a database by using a template, and identifying the label information by using the template type with the maximum matching degree as the label information type.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211092220.2A CN115205835B (en) | 2022-09-08 | 2022-09-08 | Liquefied petroleum gas label pattern recognition system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211092220.2A CN115205835B (en) | 2022-09-08 | 2022-09-08 | Liquefied petroleum gas label pattern recognition system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115205835A true CN115205835A (en) | 2022-10-18 |
| CN115205835B CN115205835B (en) | 2022-12-20 |
Family
ID=83571975
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211092220.2A Active CN115205835B (en) | 2022-09-08 | 2022-09-08 | Liquefied petroleum gas label pattern recognition system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115205835B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116862908A (en) * | 2023-09-01 | 2023-10-10 | 菏泽学院 | Metal gear meshing defect detection method based on image processing |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2010021633A1 (en) * | 2008-08-22 | 2010-02-25 | Hewlett-Packard Development Company, L.P. | Image analysis method and system |
| US20160253782A1 (en) * | 2013-11-01 | 2016-09-01 | Eizo Corporation | Image quality enhancing apparatus, image display apparatus, image quality enhancing method, and computer readable storage medium |
| WO2018049598A1 (en) * | 2016-09-14 | 2018-03-22 | 深圳大学 | Ocular fundus image enhancement method and system |
| CN107993204A (en) * | 2017-11-28 | 2018-05-04 | 重庆大学 | A kind of MRI image reconstructing method based on image block enhancing rarefaction representation |
| CN114926459A (en) * | 2022-06-21 | 2022-08-19 | 上海市计量测试技术研究院 | Image quality evaluation method, system and computer readable medium |
| CN114972339A (en) * | 2022-07-27 | 2022-08-30 | 金成技术股份有限公司 | Data enhancement system for bulldozer structural member production abnormity detection |
-
2022
- 2022-09-08 CN CN202211092220.2A patent/CN115205835B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2010021633A1 (en) * | 2008-08-22 | 2010-02-25 | Hewlett-Packard Development Company, L.P. | Image analysis method and system |
| US20160253782A1 (en) * | 2013-11-01 | 2016-09-01 | Eizo Corporation | Image quality enhancing apparatus, image display apparatus, image quality enhancing method, and computer readable storage medium |
| WO2018049598A1 (en) * | 2016-09-14 | 2018-03-22 | 深圳大学 | Ocular fundus image enhancement method and system |
| CN107993204A (en) * | 2017-11-28 | 2018-05-04 | 重庆大学 | A kind of MRI image reconstructing method based on image block enhancing rarefaction representation |
| CN114926459A (en) * | 2022-06-21 | 2022-08-19 | 上海市计量测试技术研究院 | Image quality evaluation method, system and computer readable medium |
| CN114972339A (en) * | 2022-07-27 | 2022-08-30 | 金成技术股份有限公司 | Data enhancement system for bulldozer structural member production abnormity detection |
Non-Patent Citations (2)
| Title |
|---|
| YANMIN HE等: ""Multi-stage image denoising based on correlation coefficient matching and sparse dictionary pruning"", 《SIGNAL PROCESSING》 * |
| 陈晔等: ""耦合了聚类中心约束项的稀疏表示图像去噪"", 《电子技术与软件工程》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116862908A (en) * | 2023-09-01 | 2023-10-10 | 菏泽学院 | Metal gear meshing defect detection method based on image processing |
| CN116862908B (en) * | 2023-09-01 | 2023-11-24 | 菏泽学院 | Metal gear meshing defect detection method based on image processing |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115205835B (en) | 2022-12-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108549873B (en) | Three-dimensional face recognition method and three-dimensional face recognition system | |
| CN107977671B (en) | Tongue picture classification method based on multitask convolutional neural network | |
| CN109829914B (en) | Method and device for detecting product defects | |
| US8184915B2 (en) | Device and method for fast computation of region based image features | |
| US20030068083A1 (en) | Face detecting method depending on image | |
| CN110533732B (en) | Method and device for identifying skin color in image, electronic equipment and storage medium | |
| US20110268319A1 (en) | Detecting and tracking objects in digital images | |
| CN111461036B (en) | Real-time pedestrian detection method using background modeling to enhance data | |
| CN109740572A (en) | A kind of human face in-vivo detection method based on partial color textural characteristics | |
| CN111652213A (en) | Ship water gauge reading identification method based on deep learning | |
| CN113592886B (en) | Drawing examining method and device for building drawing, electronic equipment and medium | |
| CN115205835B (en) | Liquefied petroleum gas label pattern recognition system | |
| CN108875727B (en) | The detection method and device of graph-text identification, storage medium, processor | |
| CN111784675A (en) | Method and device for processing article texture information, storage medium and electronic equipment | |
| CN110909582B (en) | Face recognition method and equipment | |
| CN110766708B (en) | Image comparison method based on contour similarity | |
| CN111144466B (en) | Image sample self-adaptive depth measurement learning method | |
| JPH11306325A (en) | Method and device for object detection | |
| CN111199228A (en) | License plate positioning method and device | |
| CN116030453A (en) | Digital ammeter identification method, device and equipment | |
| CN114897906A (en) | Modified plastic appearance defect control method based on computer vision | |
| CN112070116B (en) | Automatic artistic drawing classification system and method based on support vector machine | |
| CN115393748A (en) | Method for detecting infringement trademark based on Logo recognition | |
| Amelia | Age Estimation on Human Face Image Using Support Vector Regression and Texture-Based Features | |
| Nouri et al. | Global visual saliency: Geometric and colorimetrie saliency fusion and its applications for 3D colored meshes |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: Liquefied petroleum gas label graphic recognition system Effective date of registration: 20230920 Granted publication date: 20221220 Pledgee: Jinan Branch of Qingdao Bank Co.,Ltd. Pledgor: SHANDONG TELIAN INFO TECH CO.,LTD. Registration number: Y2023370000108 |