CN113688815A - Medicine packaging text computer recognition algorithm and device for complex illumination environment - Google Patents
Medicine packaging text computer recognition algorithm and device for complex illumination environment Download PDFInfo
- Publication number
- CN113688815A CN113688815A CN202110611060.7A CN202110611060A CN113688815A CN 113688815 A CN113688815 A CN 113688815A CN 202110611060 A CN202110611060 A CN 202110611060A CN 113688815 A CN113688815 A CN 113688815A
- Authority
- CN
- China
- Prior art keywords
- image
- medicine
- recognition algorithm
- computer recognition
- point
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 239000003814 drug Substances 0.000 title claims abstract description 57
- 238000004806 packaging method and process Methods 0.000 title claims abstract description 25
- 238000005286 illumination Methods 0.000 title claims abstract description 14
- 229940079593 drug Drugs 0.000 title description 2
- 238000004364 calculation method Methods 0.000 claims abstract description 17
- 238000012856 packing Methods 0.000 claims abstract description 17
- 238000001914 filtration Methods 0.000 claims abstract description 13
- 238000000034 method Methods 0.000 claims abstract description 13
- 238000001514 detection method Methods 0.000 claims abstract description 9
- 230000000694 effects Effects 0.000 claims abstract description 9
- 238000005728 strengthening Methods 0.000 claims abstract description 7
- 230000002708 enhancing effect Effects 0.000 claims abstract description 4
- 238000013507 mapping Methods 0.000 claims abstract description 4
- 238000013528 artificial neural network Methods 0.000 claims description 11
- 238000009877 rendering Methods 0.000 claims description 9
- 238000011176 pooling Methods 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 6
- 238000009826 distribution Methods 0.000 claims description 4
- 230000002159 abnormal effect Effects 0.000 claims description 3
- 230000009977 dual effect Effects 0.000 claims description 3
- 230000005764 inhibitory process Effects 0.000 claims description 3
- 230000014759 maintenance of location Effects 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 230000011218 segmentation Effects 0.000 claims description 3
- 230000008054 signal transmission Effects 0.000 claims description 3
- 230000001629 suppression Effects 0.000 claims description 3
- 238000000605 extraction Methods 0.000 abstract description 8
- 238000013527 convolutional neural network Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 229910000831 Steel Inorganic materials 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 210000000988 bone and bone Anatomy 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 239000010959 steel Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Images
Classifications
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration by the use of histogram techniques
-
- G06T5/70—
-
- G06T5/90—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H40/00—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices
- G16H40/20—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices for the management or administration of healthcare resources or facilities, e.g. managing hospital staff or surgery rooms
Abstract
The invention provides a medicine packaging text computer recognition algorithm and device for a complex illumination environment, and relates to the technical field of character and image recognition. The medicine packaging text computer recognition algorithm and device for the complex illumination environment comprise the following steps: s1, calculating the position of a medicine packing box in a picture; s2, correcting and enhancing the exposure rate of the image; s3, extracting the edge of the image area; s4, using 3D structure prediction in the local image; s5, filtering 2D plane information; s6, accurately positioning a color difference area of the sunken embossed seal characters; s7, a two-dimensional link mapping strengthening model is used for the color difference area; S8.3D point cloud reconstruction; s9, inputting the extracted features into a character detection module; and S10, outputting information such as medicine batch numbers and the like. By the calculation of the algorithm, characters on the medicine packaging box can be clearly displayed in a complex illumination environment, the extraction effect is good, and the method can be widely applied to batch number collection and information identification on the medicine packaging box.
Description
Technical Field
The invention relates to the technical field of character and image recognition, in particular to a medicine packaging character computer recognition algorithm and a device for a complex illumination environment.
Background
Most of the current mainstream OCR character image recognition algorithms adopt single algorithms such as artificial neural networks, bone features, feature point matching and the like to realize character recognition. Such recognition can be achieved with a relatively high degree of accuracy for inked text (e.g., plain text documents). However, when the conventional mainstream OCR character image recognition algorithm is applied to extraction of characters spanning multiple color areas (the medicine packaging box usually has several colors) on the medicine packaging box, even steel seal characters (recesses) under a complex light source, the conventional mainstream OCR character image recognition algorithm is not sufficient.
The current typical OCR character recognition method has the following problems for embossed characters with reflective surface material, more and different color areas and depressions on the medicine packaging box:
1. image feature extraction: the existing technology generally uses a convolution-based neural network (CNN) as a feature extraction means, which means that a large amount of data needs to be matched to enhance the robustness of feature extraction. However, no trained mature algorithm is provided at present, which aims at the application scene of medicine box information extraction, so that the existing convolutional neural network feature extraction algorithm has a poor effect under the influence of objective factors such as blurring, distortion, multi-color background and multi-angle complex light.
2. Character detection: CTPN is one of the most common text detection models at present, and bidirectional LSTM is added in the detection network of R-CNN for semantic recognition, so that the effect of accurate detection is achieved. However, the training set of R-CNN is based on two-dimensional plane images, and the front and back logical relevance of characters of the box embossed seal characters (such as product batch numbers) is weak, so that the recognition effect of the detection on three-dimensional, especially concave embossed seal characters is very poor at present.
Disclosure of Invention
Technical problem to be solved
Aiming at the defects of the prior art, the invention provides a medicine packaging text computer recognition algorithm and device for a complex illumination environment, and solves the problem of inaccurate recognition under the influence of light in the existing text recognition.
(II) technical scheme
In order to achieve the purpose, the invention is realized by the following technical scheme: a medicine packing box character computer recognition algorithm suitable for complex illumination environment comprises the following steps:
s1, calculating the position of a medicine packing box in a picture;
1) the image acquisition system is started to operate in a real-time recording mode, and when the background is detected to change to a certain degree, images of the medicine packaging box are acquired according to the number of frames;
2) carrying out mean value filtering on the image by adopting a 3 x 3 template, and then segmenting the fonts in the image by utilizing a simple threshold segmentation method;
3) scanning from left to right in a segmented manner within a certain interval range to obtain a left edge sampling point of the medicine packaging box, and then scanning from bottom to top in a segmented manner to obtain a lower edge sampling point of the medicine packaging box;
4) respectively fitting linear equations respectively corresponding to the two point sets in the last step by using a least square method, and setting a function relation between x and y on a straight line as follows: when y is a + bx, the minimum distance formula is:
5) solving the intersection point obtained after calculation by using two linear equations obtained by the calculation in the previous step and a simultaneous equation set, wherein the intersection point is the coordinate of the lower left corner of the Chinese medicine packing box in the image;
6) solving the intersection point obtained after calculation by using two linear equations obtained by the calculation in the last step and a simultaneous equation set, wherein the intersection point is the coordinate of the upper right corner of the package of the traditional Chinese medicine in the image;
s2, correcting and enhancing the exposure rate of the image;
1) the method comprises the following steps of judging abnormal light rays and normal light rays by using a newly developed global-local discriminator, and finding out the color gamut position needing to be adjusted, wherein the formula of the relative discriminator is as follows:
2) and a self-characterized retention function is used to ensure that the characteristics of the content of the image are consistent before and after the color gamut is adjusted, and the function formula is as follows:
3) the self-attention characteristic is added into the U neural network to generate output, so that the effect of extracting the characteristics of different dimensions in each different depth layer can be achieved;
s3, extracting the edge of the image area;
1) histogram equalization is used to enhance image contrast, the formula is as follows:
2) gaussian filtering is used for removing noise, and the principle is that weighted average is carried out according to the gray values of pixel points with filtering and field points thereof according to a parameter rule generated by a Gaussian formula, so that high-frequency noise superposed in an ideal image can be effectively filtered;
3) then, by calculating the speed of the change of the pixel gray value in the image, namely the gradient, the conceptual formula is as follows:
4) carrying out non-maximum suppression on the gradient image: the gradient image obtained from the last step has a lot of interferences, such as thick and wide edges, weak edges and other problems, non-maximum value inhibition is used for searching the local maximum value of the pixel points, the gray value corresponding to the non-maximum value is set to be 0, and thus a majority of non-edge pixel points can be removed;
5) edge join using dual thresholds: the purpose of the step is to eliminate false edges, two reasonable high and low thresholds are selected, a point smaller than the low threshold is regarded as a false edge and is set as 0, a point larger than the high threshold is regarded as a strong edge and is set as 1, and pixel points in the middle area of the two thresholds are further checked;
s4, predicting a local image by using a 3D structure, prejudging each pixel point of a picture object by using the 3D structure predictor, generating a 3D coordinate Xi ═ Xi, yi, zi ] T at each pixel point through 2D convolution calculation, and further generating the surface of the 3D object, namely forming dense point cloud by using a deep neural network of two-dimensional convolution operation;
s5, filtering 2D plane information;
s6, accurately positioning a color difference area of the sunken embossed seal characters;
s7, a two-dimensional link mapping strengthening model is used for the color difference area;
1) introducing a rendering pipeline as a real rendering near-similarity comparison, performing two-dimensional projection optimization by using a rendering image, and projecting a 3D coordinate into a 2D coordinate;
2) the model is strengthened by using a loss function of a template (mask) formed by the well depth and the phase of the image, and the formula is as follows:
and
S8.3D point cloud reconstruction;
s9, inputting the extracted features into a character detection module;
and S10, outputting information such as medicine batch numbers and the like.
Preferably, (x) in the fourth step in said S1i,yi) The above equation respectively solves a and b to form a simultaneous equation set, and the parameters a and b are obtained after the solution, so that a linear equation corresponding to the point set is obtained.
Preferably, C in the first step of S2 represents the neural network of the discriminator, Xr and Xf are distributions sampled from the map, and σ represents the S function.
Preferably, in the second step of S2, ILRepresents the input low brightness picture and G (I)L) Representative of the enhanced output of the generator, phii,jRepresents the extracted feature map, I represents the ith largest pooling, j represents the jth convolutional layer after the ith largest pooling, WijAnd HijThe size of the feature map is extracted.
Preferably, M in the second step of S7kAnd ZkThe true template and the well depth image in the Kth new view angle can make the error uniformly distributed in all new view angles instead of a fixed certain N view angle through the strengthening.
The utility model provides a medicine packing box words computer recognition device suitable for complicated illumination environment, includes discernment platform, computer, the upper surface of discernment platform is provided with the discernment district, the upper surface of discernment platform and the rear that is located the discernment district are provided with the spliced pole, the lower surface of spliced pole is provided with the camera, the inside of camera is provided with the light source, the inside of camera is provided with signal transmission module, computer and camera signal connection, computer and hospital's medicine database signal connection, there is the algorithm in the inside of computer.
(III) advantageous effects
The invention provides a medicine packaging text computer recognition algorithm and a device used in a complex illumination environment. The method has the following beneficial effects:
1. according to the invention, in the image preprocessing module, the judgment of the basic information such as the position, the front and the back, the inclination and the like of the medicine packaging box is firstly realized, then the local picture is intercepted by utilizing the information acquired in the last step, and the characteristics (character information on the medicine packaging box) of the relevant information in the intercepted picture are amplified by utilizing the image enhancement technology, so that the information display is more clear.
2. According to the invention, the 2D image is converted into the 3D structure, so that the concave characters can be further highlighted, a remarkable 3D characteristic is formed, the characters are more clearly displayed, and the influence of light is effectively eliminated.
3. According to the invention, through calculation of the algorithm, characters on the medicine packaging box can be clearly displayed in a complex illumination environment, the extraction effect is good, and the method can be widely applied to batch number collection and information identification on the medicine packaging box.
Drawings
FIG. 1 is a schematic flow chart of the algorithm of the present invention;
FIG. 2 is a schematic view of the structure of the apparatus of the present invention.
Wherein, 1, identifying a platform; 2. an identification area; 3. a support pillar; 4. a camera; 5. a computer.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. 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.
Example (b):
as shown in fig. 1-2, an embodiment of the present invention provides a medicine packing box text computer recognition algorithm suitable for a complex lighting environment, including the following steps:
s1, calculating the position of a medicine packing box in a picture;
1) the image acquisition system is started to operate in a real-time recording mode, and when the background is detected to change to a certain degree, images of the medicine packaging box are acquired according to the number of frames;
2) carrying out mean value filtering on the image by adopting a 3 x 3 template, and then segmenting the fonts in the image by utilizing a simple threshold segmentation method;
3) scanning from left to right in a segmented manner within a certain interval range to obtain a left edge sampling point of the medicine packaging box, and then scanning from bottom to top in a segmented manner to obtain a lower edge sampling point of the medicine packaging box;
4) respectively fitting linear equations respectively corresponding to the two point sets in the last step by using a least square method, and setting a function relation between x and y on a straight line as follows: when y is a + bx, the minimum distance formula is:
5) solving the intersection point obtained after calculation by using two linear equations obtained by the calculation in the previous step and a simultaneous equation set, wherein the intersection point is the coordinate of the lower left corner of the Chinese medicine packing box in the image;
6) solving the intersection point obtained after calculation by using two linear equations obtained by the calculation in the last step and a simultaneous equation set, wherein the intersection point is the coordinate of the upper right corner of the package of the traditional Chinese medicine in the image;
s2, correcting and enhancing the exposure rate of the image;
1) the method comprises the following steps of judging abnormal light rays and normal light rays by using a newly developed global-local discriminator, and finding out the color gamut position needing to be adjusted, wherein the formula of the relative discriminator is as follows:
2) and a self-characterized retention function is used to ensure that the characteristics of the content of the image are consistent before and after the color gamut is adjusted, and the function formula is as follows:
3) the self-attention characteristic is added into the U neural network to generate output, so that the effect of extracting the characteristics of different dimensions in each different depth layer can be achieved;
s3, extracting the edge of the image area;
1) histogram equalization is used to enhance image contrast, the formula is as follows:
2) gaussian filtering is used for removing noise, and the principle is that weighted average is carried out according to the gray values of pixel points with filtering and field points thereof according to a parameter rule generated by a Gaussian formula, so that high-frequency noise superposed in an ideal image can be effectively filtered;
3) then, by calculating the speed of the change of the pixel gray value in the image, namely the gradient, the conceptual formula is as follows:
4) carrying out non-maximum suppression on the gradient image: the gradient image obtained from the last step has a lot of interferences, such as thick and wide edges, weak edges and other problems, non-maximum value inhibition is used for searching the local maximum value of the pixel points, the gray value corresponding to the non-maximum value is set to be 0, and thus a majority of non-edge pixel points can be removed;
5) edge join using dual thresholds: the purpose of the step is to eliminate false edges, two reasonable high and low thresholds are selected, a point smaller than the low threshold is regarded as a false edge and is set as 0, a point larger than the high threshold is regarded as a strong edge and is set as 1, and pixel points in the middle area of the two thresholds are further checked;
s4, predicting a local image by using a 3D structure, prejudging each pixel point of a picture object by using the 3D structure predictor, generating a 3D coordinate Xi ═ Xi, yi, zi ] T at each pixel point through 2D convolution calculation, and further generating the surface of the 3D object, namely forming dense point cloud by using a deep neural network of two-dimensional convolution operation;
s5, filtering 2D plane information;
s6, accurately positioning a color difference area of the sunken embossed seal characters;
s7, a two-dimensional link mapping strengthening model is used for the color difference area;
1) introducing a rendering pipeline as a real rendering near-similarity comparison, performing two-dimensional projection optimization by using a rendering image, and projecting a 3D coordinate into a 2D coordinate;
2) the model is strengthened by using a loss function of a template (mask) formed by the well depth and the phase of the image, and the formula is as follows:
and
S8.3D point cloud reconstruction;
s9, inputting the extracted features into a character detection module;
and S10, outputting information such as medicine batch numbers and the like.
(x) in the fourth step in said S1i,yi) In the first step of S2, C represents the neural network of the discriminator, Xr and Xf are the distribution sampled from the graph, sigma represents the S function, and in the second step of S2, I is the distribution of the samples in the graph, whereinLRepresents the input low brightness picture and G (I)L) Representative of the enhanced output of the generator, phii,jRepresents the extracted feature map, I represents the ith largest pooling, j represents the jth convolutional layer after the ith largest pooling, WijAnd HijIs the size of the extracted feature map, M in the second step of S7kAnd ZkThe true template and the well depth image in the Kth new view angle can make the error uniformly distributed in all new view angles instead of a fixed certain N view angle through the strengthening.
The utility model provides a medicine packing box chinese character computer recognition device suitable for complicated illumination environment, including discernment platform 1, computer 6, discernment platform 1 is used for placing the medicine, the upper surface of discernment platform 1 is provided with identification area 2, place the medicine in here, the upper surface of discernment platform 1 and the rear that is located identification area 2 are provided with spliced pole 3, conveniently connect camera 4, camera 4's inside is provided with the light source, spliced pole 3's lower surface is provided with camera 4, shoot the medicine box surface, camera 4's inside is provided with signal transmission module, transmit the photo, computer 5 and camera 4 signal connection, computer 5 and hospital's medicine database signal connection, conveniently to medicine information output, type.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (6)
1. A medicine packing box character computer recognition algorithm suitable for complex illumination environment is characterized in that: the method comprises the following steps:
s1, calculating the position of a medicine packing box in a picture;
1) the image acquisition system is started to operate in a real-time recording mode, and when the background is detected to change to a certain degree, images of the medicine packaging box are acquired according to the number of frames;
2) carrying out mean value filtering on the image by adopting a 3 x 3 template, and then segmenting the fonts in the image by utilizing a simple threshold segmentation method;
3) scanning from left to right in a segmented manner within a certain interval range to obtain a left edge sampling point of the medicine packaging box, and then scanning from bottom to top in a segmented manner to obtain a lower edge sampling point of the medicine packaging box;
4) respectively fitting linear equations respectively corresponding to the two point sets in the last step by using a least square method, and setting a function relation between x and y on a straight line as follows: y is a + bx, the minimum distance formula is
5) Solving the intersection point obtained after calculation by using two linear equations obtained by the calculation in the previous step and a simultaneous equation set, wherein the intersection point is the coordinate of the lower left corner of the Chinese medicine packing box in the image;
6) solving the intersection point obtained after calculation by using two linear equations obtained by the calculation in the last step and a simultaneous equation set, wherein the intersection point is the coordinate of the upper right corner of the package of the traditional Chinese medicine in the image;
s2, correcting and enhancing the exposure rate of the image;
1) the newly developed global-local discriminator is used for judging abnormal light rays and normal light rays and finding out the color gamut position needing to be adjusted, and the formula of the relative discriminator is as follows:
2) and a self-characterized retention function is used to ensure that the characteristics of the content of the image are consistent before and after the color gamut is adjusted, and the function formula is as follows:
3) the self-attention characteristic is added into the U neural network to generate output, so that the effect of extracting the characteristics of different dimensions in each different depth layer can be achieved;
s3, extracting the edge of the image area;
1) histogram equalization is used to enhance image contrast, the formula is as follows:
2) gaussian filtering is used for removing noise, and the principle is that weighted average is carried out according to the gray values of pixel points with filtering and field points thereof according to a parameter rule generated by a Gaussian formula, so that high-frequency noise superposed in an ideal image can be effectively filtered;
3) then, by calculating the speed of the change of the pixel gray value in the image, namely the gradient, the conceptual formula is as follows:
4) carrying out non-maximum suppression on the gradient image: the gradient image obtained from the last step has a lot of interferences, such as thick and wide edges, weak edges and other problems, non-maximum value inhibition is used for searching the local maximum value of the pixel points, the gray value corresponding to the non-maximum value is set to be 0, and thus a majority of non-edge pixel points can be removed;
5) edge join using dual thresholds: the purpose of the step is to eliminate false edges, two reasonable high and low thresholds are selected, a point smaller than the low threshold is regarded as a false edge and is set as 0, a point larger than the high threshold is regarded as a strong edge and is set as 1, and pixel points in the middle area of the two thresholds are further checked;
s4, predicting a local image by using a 3D structure, prejudging each pixel point of a picture object by using the 3D structure predictor, generating a 3D coordinate Xi ═ Xi, yi, zi ] T at each pixel point through 2D convolution calculation, and further generating the surface of the 3D object, namely forming dense point cloud by using a deep neural network of two-dimensional convolution operation;
s5, filtering 2D plane information;
s6, accurately positioning a color difference area of the sunken embossed seal characters;
s7, a two-dimensional link mapping strengthening model is used for the color difference area;
1) introducing a rendering pipeline as a rendering near-similarity comparison, performing two-dimensional projection optimization by using a rendering image, and projecting a 3D coordinate into a 2D coordinate;
2) the model is strengthened by using a loss function of a template (mask) formed by the well depth and the phase of the image, and the formula is as follows:
and
S8.3D point cloud reconstruction;
s9, inputting the extracted features into a character detection module;
and S10, outputting information such as medicine batch numbers and the like.
2. The computer recognition algorithm for medicine packing boxes and characters in complex lighting environments according to claim 1, wherein the computer recognition algorithm comprises the following steps: (x) in the fourth step in said S1i,yi) The above equation respectively solves a and b to form a simultaneous equation set, and the parameters a and b are obtained after the solution, so that a linear equation corresponding to the point set is obtained.
3. The computer recognition algorithm for medicine packing boxes and characters in complex lighting environments according to claim 1, wherein the computer recognition algorithm comprises the following steps: in the first step of S2, C represents the neural network of the discriminator, Xr and Xf are distributions sampled from the graph, and σ represents the S function.
4. The computer recognition algorithm for medicine packing boxes and characters in complex lighting environments according to claim 1, wherein the computer recognition algorithm comprises the following steps: in the second step in S2, wherein ILRepresents the input low brightness picture and G (I)L) Representative of the enhanced output of the generator, phii,jRepresents the extracted feature map, I represents the ith largest pooling, j represents the jth convolutional layer after the ith largest pooling, WijAnd HijThe size of the feature map is extracted.
5. The computer recognition algorithm for medicine packing boxes and characters in complex lighting environments according to claim 1, wherein the computer recognition algorithm comprises the following steps: m in the second step in S7kAnd ZkThe true template and the well depth image in the Kth new view angle can make the error uniformly distributed in all new view angles instead of a fixed certain N view angle through the strengthening.
6. A medicine packing box character computer recognition device suitable for complex illumination environment, includes identification platform (1), computer (5), its characterized in that: the upper surface of discernment platform (1) is provided with discernment district (2), the upper surface of discernment platform (1) and the rear that is located discernment district (2) are provided with spliced pole (3), the lower surface of spliced pole (3) is provided with camera (4), the inside of camera (4) is provided with the light source, the inside of camera (4) is provided with signal transmission module, computer (5) and camera (4) signal connection, computer (6) and hospital's medicine database signal connection.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110611060.7A CN113688815A (en) | 2021-06-01 | 2021-06-01 | Medicine packaging text computer recognition algorithm and device for complex illumination environment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110611060.7A CN113688815A (en) | 2021-06-01 | 2021-06-01 | Medicine packaging text computer recognition algorithm and device for complex illumination environment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113688815A true CN113688815A (en) | 2021-11-23 |
Family
ID=78576424
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110611060.7A Pending CN113688815A (en) | 2021-06-01 | 2021-06-01 | Medicine packaging text computer recognition algorithm and device for complex illumination environment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113688815A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108548820A (en) * | 2018-03-28 | 2018-09-18 | 浙江理工大学 | Cosmetics paper labels defect inspection method |
| CN110059694A (en) * | 2019-04-19 | 2019-07-26 | 山东大学 | The intelligent identification Method of lteral data under power industry complex scene |
| WO2019153739A1 (en) * | 2018-02-09 | 2019-08-15 | 深圳壹账通智能科技有限公司 | Identity authentication method, device, and apparatus based on face recognition, and storage medium |
| CN110766020A (en) * | 2019-10-30 | 2020-02-07 | 哈尔滨工业大学 | System and method for detecting and identifying multi-language natural scene text |
| WO2021051604A1 (en) * | 2019-09-19 | 2021-03-25 | 平安科技（深圳）有限公司 | Method for identifying text region of osd, and device and storage medium |
-
2021
- 2021-06-01 CN CN202110611060.7A patent/CN113688815A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019153739A1 (en) * | 2018-02-09 | 2019-08-15 | 深圳壹账通智能科技有限公司 | Identity authentication method, device, and apparatus based on face recognition, and storage medium |
| CN108548820A (en) * | 2018-03-28 | 2018-09-18 | 浙江理工大学 | Cosmetics paper labels defect inspection method |
| CN110059694A (en) * | 2019-04-19 | 2019-07-26 | 山东大学 | The intelligent identification Method of lteral data under power industry complex scene |
| WO2021051604A1 (en) * | 2019-09-19 | 2021-03-25 | 平安科技（深圳）有限公司 | Method for identifying text region of osd, and device and storage medium |
| CN110766020A (en) * | 2019-10-30 | 2020-02-07 | 哈尔滨工业大学 | System and method for detecting and identifying multi-language natural scene text |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110264416B (en) | Sparse point cloud segmentation method and device | |
| CN110569704B (en) | Multi-strategy self-adaptive lane line detection method based on stereoscopic vision | |
| CN108573276B (en) | Change detection method based on high-resolution remote sensing image | |
| CN113450307B (en) | Product edge defect detection method | |
| CN109961049B (en) | Cigarette brand identification method under complex scene | |
| CN104050471B (en) | Natural scene character detection method and system | |
| CN104778721B (en) | The distance measurement method of conspicuousness target in a kind of binocular image | |
| CN110032949A (en) | A kind of target detection and localization method based on lightweight convolutional neural networks | |
| CN109949340A (en) | Target scale adaptive tracking method based on OpenCV | |
| CN106530281B (en) | Unmanned plane image fuzzy Judgment method and system based on edge feature | |
| CN105844621A (en) | Method for detecting quality of printed matter | |
| CN110570457B (en) | Three-dimensional object detection and tracking method based on stream data | |
| CN105279772B (en) | A kind of trackability method of discrimination of infrared sequence image | |
| CN110263794B (en) | Training method of target recognition model based on data enhancement | |
| CN112488046B (en) | Lane line extraction method based on high-resolution images of unmanned aerial vehicle | |
| CN113297988B (en) | Object attitude estimation method based on domain migration and depth completion | |
| CN113808180B (en) | Heterologous image registration method, system and device | |
| CN111027511A (en) | Remote sensing image ship detection method based on region of interest block extraction | |
| CN110189375A (en) | A kind of images steganalysis method based on monocular vision measurement | |
| CN113313107B (en) | Intelligent detection and identification method for multiple types of diseases on cable surface of cable-stayed bridge | |
| CN115222884A (en) | Space object analysis and modeling optimization method based on artificial intelligence | |
| CN116703895B (en) | Small sample 3D visual detection method and system based on generation countermeasure network | |
| CN110516527B (en) | Visual SLAM loop detection improvement method based on instance segmentation | |
| CN113688815A (en) | Medicine packaging text computer recognition algorithm and device for complex illumination environment | |
| CN115239607A (en) | Method and system for self-adaptive fusion of infrared and visible light images |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20211123 |
|
| WD01 | Invention patent application deemed withdrawn after publication |