CN114596571A - Intelligent lens-free character recognition system - Google Patents
Intelligent lens-free character recognition system Download PDFInfo
- Publication number
- CN114596571A CN114596571A CN202210246740.8A CN202210246740A CN114596571A CN 114596571 A CN114596571 A CN 114596571A CN 202210246740 A CN202210246740 A CN 202210246740A CN 114596571 A CN114596571 A CN 114596571A
- Authority
- CN
- China
- Prior art keywords
- character
- model
- imaging
- mask
- image
- 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
- 238000003384 imaging method Methods 0.000 claims abstract description 62
- 230000003287 optical effect Effects 0.000 claims abstract description 32
- 238000004364 calculation method Methods 0.000 claims abstract description 21
- 238000005457 optimization Methods 0.000 claims abstract description 7
- 238000012549 training Methods 0.000 claims description 35
- 238000000034 method Methods 0.000 claims description 26
- 238000013528 artificial neural network Methods 0.000 claims description 19
- 102100032202 Cornulin Human genes 0.000 claims description 3
- 101000920981 Homo sapiens Cornulin Proteins 0.000 claims description 3
- 239000004973 liquid crystal related substance Substances 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000013136 deep learning model Methods 0.000 abstract description 2
- 230000005540 biological transmission Effects 0.000 abstract 2
- 230000006870 function Effects 0.000 description 10
- 238000010586 diagram Methods 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000007 visual effect Effects 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
- Character Discrimination (AREA)
Abstract
The invention discloses an intelligent lensless character recognition system. The system comprises an optical module and a calculation imaging and intelligent character positioning recognition module, wherein the optical module consists of a mask plate with adjustable amplitude and a sensor, and the transmission light amplitude distribution of the mask plate is modeled into a two-dimensional convolution layer and can be optimized as a parameter; the calculation imaging and character positioning identification module comprises a calculation imaging model, a character positioning model and a character identification model, input data is original data obtained on a sensor after passing through the optical module, the original data is output to be a text form of predicted characters, and meanwhile, the light transmission amplitude distribution of a mask plate in the optical module and the parameters of a calculation imaging network are optimized through result feedback. The invention realizes the software and hardware integrated lens-free imaging and character recognition deep learning model optimization, improves the accuracy of character positioning and character recognition under the condition of no lens, and each module of the system has universality and strong practical applicability.
Description
Technical Field
The invention belongs to the field of lens-free imaging, and particularly relates to an intelligent lens-free character recognition system.
Background
With the rapid development and application of vision tasks, cameras are integrated on various hardware devices. Some application scenarios have strict requirements on the size of the camera, and the lens-free camera is an imaging system using a thin mask instead of a lens, so that the size of the camera can be greatly reduced.
Compared with a camera with a lens, the camera without the lens can restore an image by calculating and imaging data collected on a sensor, but the image reconstructed based on the camera without the lens has the defects of blurring and resolution, so that the camera cannot be competent for many visual tasks, and at present, the research on the detection and identification of non-single character characters based on the camera without the lens is not available.
Therefore, a need exists for a lensless text recognition system.
Disclosure of Invention
The invention provides a non-lens based character positioning and recognition system, aiming at the situation that the existing non-lens imaging technology is not applied to character positioning and recognition of non-single letters due to poor imaging quality. The identification accuracy is high, and the system method has universality.
The technical scheme adopted by the invention is as follows:
the intelligent lens-free character recognition system comprises an optical module and a calculation imaging and character positioning recognition module, wherein the optical module mainly comprises a modulation amplitude-adjustable mask plate and an optical sensor which are arranged in parallel, a target to be recognized is arranged in front of the optical module, light rays emitted by the target to be recognized are projected on the plane of the optical sensor to form a projected image (original data) after being scattered by the modulation amplitude-adjustable mask plate, and the projected image is transmitted to the calculation imaging and character positioning recognition module by the optical sensor;
the computational imaging and character recognition module comprises a computational imaging model, a character positioning model and a character recognition model which are connected in series; the input of the calculation imaging and character recognition module is a projection image obtained on the sensor after passing through the optical module, and the output is a text form of characters on the projection image.
The mask plate capable of modulating the amplitude is a binary mask plate consisting of k × k cells, the value of each cell is 1 or 0, 1 represents that light can pass through, and 0 represents that light cannot pass through.
The projected image outputs a predicted reconstructed image through a computational imaging model; the character positioning model processes the input reconstructed image and outputs the position of characters in the image; after the output result of the character positioning model is input into the character recognition model, the character recognition result of the image is output;
in the training process of the computational imaging and character recognition module, only the computational imaging model participates in training, parameters need to be updated, and the character positioning model and the character recognition model do not participate in training.
The computational imaging model is a neural network of an encoder-decoder system, and specifically adopts U-NET; the character positioning model adopts any character positioning model structure, and specifically adopts CTPN; the character recognition model adopts any character recognition model structure, and particularly adopts CRNN.
Patterns on the mask plate with the adjustable amplitude are displayed through a liquid crystal display, and the patterns on the mask plate are randomly generated or determined after training optimization; the method for determining the mask plate pattern after training optimization comprises the following steps:
1) modeling the imaging process of the target to be identified and the optical module into a two-dimensional convolution layer, specifically:
m＝w*o
wherein w represents the amplitude distribution on the reticle, i.e. the value distribution of the cells on the reticle; constructing a coordinate system by taking the center point of the mask as an original point, (i, j) is the coordinate of the center point of the unit cell on the mask, and wi，jRepresenting the value of a cell with coordinates (i, j) on the mask;
o represents the scaled image on the sensor plane when the target to be identified does not pass through the reticle (i.e. o represents the scaled image on the sensor plane when the target to be identified passes through the aperture); coordinate system (x, y) table is constructed by taking plane central point of sensor as originCoordinate value, o, of pixel point of the projected image on the sensor planex，yA pixel value at (x, y) representing the sensor plane when the target to be identified does not pass through the reticle; ox+i，y+jRepresents the pixel value at (x + i, y + j) on the sensor plane;
m represents an image of the target to be identified, which is projected on a sensor plane after passing through the mask; m isx，yRepresenting a pixel value at (x, y) of a sensor plane after the target to be identified passes through the mask;
k represents the number of rows or columns of cells on the reticle, i ∈ [1, k ];
2) and carrying out binarization on the two-dimensional convolutional layer to obtain a two-dimensional neural network two-dimensional convolutional layer, wherein the result is as follows:
wherein,
wherein, wbRepresenting the result of binarization processing on w;
since the mask has only 0 and 1 values, we use a binary neural network to train, which uses a sign function to map two consecutive values to-1 or +1, then add 1 and divide by 2;
3) parameter w of two-dimensional convolution layer of binary neural networkbTraining and optimizing the model parameters and a calculation imaging and character positioning identification module;
3.1) in the training process, randomly initializing the pattern of the mask plate through circuit adjustment, and taking the result of the random initialization as the initial parameter of the binary neural network convolution layer;
3.2) training the forward propagation process of the system: fixing a target to be recognized, measuring a projected image of the target to be recognized on a plane of an optical sensor after the target to be recognized passes through a mask in a real physical scene, and taking the projected image as input of a calculation imaging and character positioning recognition module;
training of the back propagation process: calculating Loss function Loss of predicted image and real image label output by imaging and character positioning identification module, back-propagating the Loss function Loss to the binary neural network convolutional layer, updating parameter w of the binary neural network convolutional layerbAnd according to the updated parameter wbModulating the adjustable mask plate, wherein the modulation result is used as a mask plate pattern in the forward propagation process of the model during the next round of training;
3.3) obtaining the mask pattern after the training is finished, wherein the mask pattern is the optimized result.
The unit grid size of the modulable mask plate is the same as the size of the pixel points on the sensor plane; the distance d1 between the target to be recognized and the amplitude-modulatable mask plate is far greater than the distance d2 between the amplitude-modulatable mask plate and the optical sensor, and d1 is greater than 100d 2; the amplitude distribution on the reticle is therefore approximately equal to the projection of the amplitude distribution on the reticle onto the sensor plane.
The Loss function Loss of the computational imaging and character positioning recognition module in the training process is as follows:
Loss＝a×Loss1+b×Loss2；
wherein, Loss1 is to calculate the error between the predicted image output by the imaging model and the real image label (the target image to be identified); loss2 is the error between the predicted text finally output by the computed imaging and character positioning recognition module and the real character label of the target to be recognized (text information on the target image to be recognized); a and b are weights.
The invention has the following effective benefits:
the lens-free character recognition system can reduce the size limitation brought by the lens, so that the camera is more convenient to be integrated on other equipment.
The invention realizes the software and hardware integrated lens-free imaging and character recognition deep learning model optimization, improves the accuracy of character positioning and character recognition under the condition of no lens, and each module of the system has universality and strong practical applicability.
Drawings
Fig. 1 is an overall data flow of the present invention.
Fig. 2 is a schematic diagram of an optical module according to the present invention.
FIG. 3 is a schematic diagram of a computer-aided imaging and text positioning recognition module according to the present invention.
Detailed Description
The invention is described in further detail below with reference to the figures and the embodiments.
As shown in FIG. 1, the lensless text recognition system of the present invention includes an optical module, a computational imaging and text positioning recognition module. The target to be recognized obtains original data through an optical module, and the original data obtains the text form of characters through a calculation imaging and character positioning recognition module.
The optical module consists of an amplitude-modulatable mask plate and a sensor, wherein the amplitude-modulatable mask plate is a binary mask plate and is divided into k × k unit cells, the value of each unit cell is 1 or 0, 1 represents that light can pass through, 0 represents that light cannot pass through, the unit cells are placed in front of the optical sensor and are placed in parallel, an object is placed in front of the mask plate, light emitted by the object is scattered by the amplitude-modulatable mask plate, then a specific projection image (original data) is projected onto a plane where the sensor is located, and the projection image is recorded by the sensor and then transmitted to the calculation imaging and character positioning recognition module. The patterns of the mask plate with the adjustable amplitude can be controlled in real time through a circuit and can be displayed through a liquid crystal display, and the patterns on the mask plate can be generated randomly or can be optimized and updated through training; the mask obtained by training optimization can enable the recognition effect of the calculation imaging and character positioning recognition module to be better. When the mask pattern in the optical module is fixed, the fixed mask pattern is used under the condition that the mask is not optimized, and the positioning and identification results of characters can still be obtained through the original data obtained on the sensor and the calculation imaging and character positioning identification module.
As shown in fig. 2, the method of training the optimized reticle pattern is as follows:
1) modeling a process of imaging target and reticle interaction as a two-dimensional convolutional layer
m＝w*o
Wherein w represents the amplitude distribution on the reticle, i.e. the value distribution of the cells on the reticle; constructing a coordinate system by taking the center point of the mask as an original point, (i, j) is the coordinate of the center point of a unit cell on the mask, wi，jRepresenting the value of a cell with coordinates (i, j) on the mask;
o represents an image which is zoomed on a sensor plane when the target to be identified does not pass through the mask plate under the condition that the amplitude-modulatable mask plate is fully transparent; constructing a coordinate system by taking the central point of the plane of the sensor as an origin, wherein (x, y) represents the coordinate value of a pixel point on the plane of the sensor, ox，yA pixel value at (x, y) representing the sensor plane when the target to be identified does not pass through the reticle;
m represents an image of the target to be identified, which is projected on the plane of the sensor after passing through the mask plate; m isx，yRepresenting a pixel value at (x, y) of a sensor plane after the target to be identified passes through the mask;
k represents the number of rows or columns of cells on the reticle, i ∈ [1, k ].
Since the reticle cell pixel size and the sensor pixel size are the same, and d1> > d2(d1 is greater than 100 times d2), w may be approximately equal to the projection of the amplitude distribution on the reticle onto the sensor plane.
2) And (3) training and optimizing the parameters w of the two-dimensional convolutional layer as model parameters together with a subsequent calculation imaging and character positioning recognition module.
2.1) carrying out binarization on the two-dimensional convolution layer to obtain a binary neural network:
wherein, wbThe result of binarization processing of w is shown.
Since the mask has only 0 and 1 values, we train using a binary neural network that maps w to-1 or +1 using a sign function, then adds 1 and divides by 2.
2.2) in the training process, randomly initializing the pattern of the mask plate through circuit adjustment, and taking the value as an initial parameter of the binary neural network convolution layer.
In the process of forward propagation of the system: fixing the object to be measured, measuring the original data obtained on the sensor after passing through the mask in the real physical scene, and using the original data as the input of the calculation imaging and character positioning identification module,
after the calculation imaging and character positioning identification module, the error between the real label and the real label is obtained,
during the gradient back propagation: and calculating the error between the output of the imaging and character positioning identification module and the real label, calculating the gradient and performing back propagation, performing back propagation on the error to the binary neural network convolution layer, updating the binary neural network convolution layer, and modulating the adjustable mask according to the updated weight result to be used as a mask pattern in the forward propagation process of the model during the next round of training.
And after the training is finished, fixing the mask pattern.
As shown in fig. 3, the computational imaging and text recognition module includes a computational imaging model, a text orientation model, and a text recognition model. In the using process, the module is connected in series by three models, input data is original data obtained on a sensor after passing through an optical module, and output is in a text form of predicted characters. In the training process, only the imaging model is calculated and parameters need to be updated, and the character positioning model and the character recognition model do not participate in training.
Computational imaging models are a deep learning based imaging method that can compute a predicted reconstructed image (model output) from raw data (model input) obtained on a sensor. The computational imaging model is a neural network with a structure of a coder-decoder, and particularly can adopt a network with a U-NET structure. In the training process, an image containing letters and numbers is used as a target to be recognized, original data obtained on a sensor after the image passes through an optical module is used as model input, and a training loss function consists of two parts: 1) an error Loss1 between a target image to be recognized (real image label) and a prediction image output by the model; 2) and inputting the predicted reconstructed image into the subsequent character positioning model and the character recognition model to obtain a predicted text, and calculating an error Loss2 between the predicted text and the character label of the target to be recognized. The Loss function Loss is a × Loss1+ b × Loss 2.
Gradient back propagation: the loss function calculates gradient feedback to update the calculated imaging model and the two-dimensional convolution layer, and finally the corresponding trained calculated imaging model is obtained.
The word-oriented model is a neural network model. The input data is an image and the output data is the location of the text. The model can adopt any character positioning model structure, and the specific structure can adopt the following steps: the image passes through a VGG16 network, 3 x 3 sliding windows are calculated on each line of the last convolution layer of the network, the results are connected through a BLSTM structure, and finally, a full connection layer is input and output as predicted coordinates and confidence scores. And the model is well trained and does not participate in back-propagation updating parameters of the loss function in the system.
The character recognition model is a neural network model. The input data is an alphanumeric image and the output is the result of character recognition. The model can adopt any character recognition model structure, such as CRNN: the image firstly passes through a plurality of convolution layers, a characteristic sequence is extracted, the image enters BLSTM, and finally the score of a predicted letter is output. And the model is well trained and does not participate in back-propagation updating parameters of the loss function in the system.
Claims (7)
1. An intelligent lensless character recognition system is characterized by comprising an optical module and a calculation imaging and character positioning recognition module, wherein the optical module mainly comprises a modulation amplitude mask plate and an optical sensor which are arranged in parallel, a target to be recognized is placed in front of the optical module, light rays emitted by the target to be recognized are projected on the plane of the optical sensor to form a projection image after being scattered by the modulation amplitude mask plate, and the projection image is transmitted to the calculation imaging and character positioning recognition module by the optical sensor;
patterns on the mask plate with the adjustable amplitude are displayed through a liquid crystal display, and the patterns on the mask plate are randomly generated or determined after training optimization;
the computational imaging and character recognition module comprises a computational imaging model, a character positioning model and a character recognition model which are connected in series; the input of the calculation imaging and character recognition module is a projection image obtained on the sensor after passing through the optical module, and the output is a text form of characters on the projection image.
2. An intelligent lensless text recognition system of claim 1, wherein the modulatable amplitude mask is a binary mask consisting of k x k cells, each cell having a value of 1 or 0, 1 indicating that light is allowed to pass through, and 0 indicating that light is not allowed to pass through.
3. The intelligent lensless character recognition system of claim 1, wherein the projected image is output via a computational imaging model as a predicted reconstructed image; the character positioning model processes the input reconstructed image and outputs the position of characters in the image; after the output result of the character positioning model is input into the character recognition model, the character recognition result of the image is output;
in the training process of the computational imaging and character recognition module, only the computational imaging model participates in training, and the character positioning model and the character recognition model do not participate in training.
4. The intelligent lensless character recognition system of claim 3, wherein the computational imaging model is a neural network of an encoder-decoder system, specifically employing U-NET; the character positioning model adopts any character positioning model structure, and specifically adopts CTPN; the character recognition model adopts any character recognition model structure, and particularly adopts CRNN.
5. The intelligent lensless text recognition system of claim 1, wherein the method for determining the reticle pattern after training optimization comprises the steps of:
1) modeling the imaging process of the target to be identified and the optical module into a two-dimensional convolution layer, specifically:
m＝w*o
wherein w represents the amplitude distribution on the reticle, i.e. the value distribution of the cells on the reticle; constructing a coordinate system by taking the center point of the mask as an original point, (i, j) is the coordinate of the center point of the unit cell on the mask, and wi，jRepresenting the value of a cell with coordinates (i, j) on the mask;
o represents an image zoomed on the sensor plane when the target to be identified does not pass through the mask; constructing a coordinate system by taking the central point of the sensor plane as an origin, (x, y) representing the coordinate value of the pixel point of the projection image on the sensor plane, ox，yA pixel value at (x, y) representing the sensor plane when the target to be identified does not pass through the reticle; ox+i，y+jRepresents the pixel value at (x + i, y + j) on the sensor plane;
m represents an image of the target to be identified, which is projected on a sensor plane after passing through the mask; m isx，yRepresenting a pixel value at (x, y) of a sensor plane after the target to be identified passes through the mask;
k represents the number of rows or columns of cells on the reticle, i ∈ [1, k ];
2) and carrying out binarization on the two-dimensional convolutional layer to obtain a two-dimensional neural network two-dimensional convolutional layer, wherein the result is as follows:
wherein,
wherein, wbRepresenting the result of binarization processing on w;
3) parameter w of two-dimensional convolution layer of binary neural networkbTraining and optimizing the model parameters and a calculation imaging and character positioning identification module;
3.1) in the training process, randomly initializing the pattern of the mask plate through circuit adjustment, and taking the result of the random initialization as the initial parameter of the binary neural network convolution layer;
3.2) training the forward propagation process of the system: fixing a target to be recognized, measuring a projected image of the target to be recognized on a plane of an optical sensor after the target to be recognized passes through a mask in a real physical scene, and taking the projected image as input of a calculation imaging and character positioning recognition module;
training of the back propagation process: calculating Loss function Loss of predicted image and real image label output by imaging and character positioning identification module, back-propagating the Loss function Loss to the binary neural network convolutional layer, updating parameter w of the binary neural network convolutional layerbAnd according to the updated parameter wbModulating the adjustable mask plate, wherein the modulation result is used as a mask plate pattern in the forward propagation process of the model during the next round of training;
3.3) obtaining the mask pattern after the training is finished, wherein the mask pattern is the optimized result.
6. The intelligent lensless text recognition system of claim 5, wherein the unit cell size of the modulatable reticle is the same as the pixel size on the sensor plane; the distance d1 between the target to be identified and the amplitude-modulatable mask plate is far greater than the distance d2 between the amplitude-modulatable mask plate and the optical sensor, and d1 is more than 100d 2; the amplitude distribution on the reticle is therefore approximately equal to the projection of the amplitude distribution on the reticle onto the sensor plane.
7. The intelligent lensless character recognition system of any one of claims 3 and 5, wherein the Loss function Loss of the computational imaging and character positioning recognition module during training is:
Loss＝a×Loss1+b×Loss2；
wherein, Loss1 is to calculate the error between the predicted image and the real image label output by the imaging model; loss2 is the error between the predicted text finally output by the computed imaging and character positioning recognition module and the real character label of the target to be recognized; a and b are weights.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210246740.8A CN114596571B (en) | 2022-03-14 | 2022-03-14 | Intelligent lens-free character recognition system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210246740.8A CN114596571B (en) | 2022-03-14 | 2022-03-14 | Intelligent lens-free character recognition system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114596571A true CN114596571A (en) | 2022-06-07 |
| CN114596571B CN114596571B (en) | 2024-10-18 |
Family
ID=81818206
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210246740.8A Active CN114596571B (en) | 2022-03-14 | 2022-03-14 | Intelligent lens-free character recognition system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114596571B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6211508B1 (en) * | 1999-01-28 | 2001-04-03 | Syscan, Inc. | Lensless optical system in imaging sensing module |
| US20050103862A1 (en) * | 2003-11-13 | 2005-05-19 | Metrologic Instruments, Inc. | Hand-supportable imaging-based bar code symbol reader employing automatic object presence and range detection to control the generation of near-field and far-field wide-area illumination during bar code symbol imaging operations |
| CN109447078A (en) * | 2018-10-23 | 2019-03-08 | 四川大学 | A kind of detection recognition method of natural scene image sensitivity text |
| US20200072728A1 (en) * | 2015-12-28 | 2020-03-05 | Commissariat A L'energie Atomique Et Aux Energies Alternatives | Device and method for bimodal observation of an object |
| WO2021076075A2 (en) * | 2019-10-14 | 2021-04-22 | İzmi̇r Yüksek Teknoloji̇ Ensti̇tüsü | Cell viability analysis and counting from holograms by using deep learning and appropriate lensless holographic microscope |
| US20220398766A1 (en) * | 2020-03-02 | 2022-12-15 | Panasonic Intellectual Property Corporation Of America | Information processing method, information processing system, and information processing device |
-
2022
- 2022-03-14 CN CN202210246740.8A patent/CN114596571B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6211508B1 (en) * | 1999-01-28 | 2001-04-03 | Syscan, Inc. | Lensless optical system in imaging sensing module |
| US20050103862A1 (en) * | 2003-11-13 | 2005-05-19 | Metrologic Instruments, Inc. | Hand-supportable imaging-based bar code symbol reader employing automatic object presence and range detection to control the generation of near-field and far-field wide-area illumination during bar code symbol imaging operations |
| US20200072728A1 (en) * | 2015-12-28 | 2020-03-05 | Commissariat A L'energie Atomique Et Aux Energies Alternatives | Device and method for bimodal observation of an object |
| CN109447078A (en) * | 2018-10-23 | 2019-03-08 | 四川大学 | A kind of detection recognition method of natural scene image sensitivity text |
| WO2021076075A2 (en) * | 2019-10-14 | 2021-04-22 | İzmi̇r Yüksek Teknoloji̇ Ensti̇tüsü | Cell viability analysis and counting from holograms by using deep learning and appropriate lensless holographic microscope |
| US20220398766A1 (en) * | 2020-03-02 | 2022-12-15 | Panasonic Intellectual Property Corporation Of America | Information processing method, information processing system, and information processing device |
Non-Patent Citations (2)
| Title |
|---|
| YUNZHE LI等: "Deep speckle correlation a deep learning approach toward scalable imaging through scattering media", OPTICA, vol. 5, no. 10, 20 October 2018 (2018-10-20), pages 1181 - 1190 * |
| 曾旭: "基于LCOS空间光调制器的无透镜光学相关器的理论与技术研究", 中国博士学位论文全文数据库 (信息科技辑), no. 2013, 15 September 2013 (2013-09-15), pages 135 - 19 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114596571B (en) | 2024-10-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111709409B (en) | Face living body detection method, device, equipment and medium | |
| CN112464912B (en) | Robot end face detection method based on YOLO-RGGNet | |
| CN108229519A (en) | The method, apparatus and system of image classification | |
| CN106780546B (en) | The personal identification method of motion blur encoded point based on convolutional neural networks | |
| CN111260687B (en) | Aerial video target tracking method based on semantic perception network and related filtering | |
| CN112634369A (en) | Space and or graph model generation method and device, electronic equipment and storage medium | |
| CN111292408A (en) | Shadow generation method based on attention mechanism | |
| CN112465977B (en) | Method for repairing three-dimensional model water surface loophole based on dense point cloud | |
| CN112613668A (en) | Scenic spot dangerous area management and control method based on artificial intelligence | |
| CN116030519A (en) | Learning attention detection and assessment method for live broadcast teaching platform | |
| Feng | Mask RCNN-based single shot multibox detector for gesture recognition in physical education | |
| CN114596571B (en) | Intelligent lens-free character recognition system | |
| CN117351460A (en) | Pond obstacle recognition and positioning method based on deep learning and binocular vision | |
| CN106934339A (en) | A kind of target following, the extracting method of tracking target distinguishing feature and device | |
| CN117058476A (en) | Target detection method based on random uncertainty | |
| CN114898447A (en) | Personalized fixation point detection method and device based on self-attention mechanism | |
| CN110619677B (en) | Method and device for reconstructing particles in three-dimensional flow field, electronic equipment and storage medium | |
| CN113705304A (en) | Image processing method and device, storage medium and computer equipment | |
| CN106874938A (en) | A kind of method for tracking target and device | |
| CN117274823B (en) | Visual transducer landslide identification method based on DEM feature enhancement | |
| Areni et al. | Multi-Object Detection Using Single Shot Multibox Detector MobileNet for People with Visual Impairments. | |
| CN115862089B (en) | Security monitoring method, device, equipment and medium based on face recognition | |
| CN114842071B (en) | Automatic conversion method for image viewing angle | |
| CN112197747B (en) | Method and apparatus for assisting target detection using wireless positioning | |
| CN113610115A (en) | Efficient face alignment method based on gray level image |
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 |