CN116843883A - Dark target detection method and system based on dark transformation characteristic constant change - Google Patents
Dark target detection method and system based on dark transformation characteristic constant change Download PDFInfo
- Publication number
- CN116843883A CN116843883A CN202310755574.9A CN202310755574A CN116843883A CN 116843883 A CN116843883 A CN 116843883A CN 202310755574 A CN202310755574 A CN 202310755574A CN 116843883 A CN116843883 A CN 116843883A
- Authority
- CN
- China
- Prior art keywords
- dark
- feature
- light image
- target detection
- detector
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 82
- 230000009466 transformation Effects 0.000 title claims abstract description 19
- 230000008859 change Effects 0.000 title claims description 15
- 238000000034 method Methods 0.000 claims abstract description 32
- 238000012549 training Methods 0.000 claims abstract description 25
- 230000008569 process Effects 0.000 claims abstract description 15
- 238000003384 imaging method Methods 0.000 claims abstract description 10
- 238000006243 chemical reaction Methods 0.000 claims description 33
- 238000013528 artificial neural network Methods 0.000 claims description 13
- 238000000605 extraction Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 5
- 230000003287 optical effect Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 2
- 230000015556 catabolic process Effects 0.000 claims description 2
- 238000006731 degradation reaction Methods 0.000 claims description 2
- 238000011478 gradient descent method Methods 0.000 claims description 2
- 238000013507 mapping Methods 0.000 claims description 2
- 238000012544 monitoring process Methods 0.000 claims description 2
- 238000012800 visualization Methods 0.000 claims description 2
- 239000004973 liquid crystal related substance Substances 0.000 claims 2
- 230000001131 transforming effect Effects 0.000 abstract 1
- 230000000694 effects Effects 0.000 description 7
- 230000006978 adaptation Effects 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000002349 favourable effect Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000002238 attenuated effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/255—Detecting or recognising potential candidate objects based on visual cues, e.g. shapes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
-
- 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
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of target detection, and provides a dark target detection method and system based on dark transformation characteristics and the like, wherein the method comprises the following steps: inputting the constant light image into a physical noise model obtained by analyzing and modeling the physical sensor noise generated in the imaging process, so as to obtain a low light image subjected to dark transformation; inputting the constant light image and the low light image into a detector, wherein a feature transformation network is configured in the detector and is used for generating a first feature vector corresponding to the constant light image and the low light image and a second feature vector containing nonlinear information, which is obtained by transforming the first feature vector; determining a training loss of the detector by using the consistency loss for feature isomorphism determined according to the first feature vector and the second feature vector, optimizing the training loss, and updating the detector by the optimized training loss; and inputting the image to be detected into a detector which is updated, and obtaining a dark target detection result.
Description
Technical Field
The invention relates to the technical field of target detection, in particular to a dark target detection method and system based on dark transformation characteristics and the like.
Background
Object detection is an important task in computer vision for automatically identifying and locating the position and class of objects in images or videos, and is widely used in the fields of medical image analysis, automatic driving, object identification and the like. With the rapid development of convolutional neural networks, target detection methods based on convolutional networks, such as Faster R-CNN, YOLO and the like, have become classical algorithms in the field of target detection. However, the good performance of the algorithms is limited to a scene with normal light, and the characteristics of the edge, texture, color and other detailed features of the target can be affected under a poor illumination condition, so that the target detection algorithms can not be used for extracting sufficient detailed features, and the problems of insufficient accuracy, false detection rate, omission rate and the like of the target detection result can occur.
To solve this problem, there are currently two studies:
(1) Study in enhancing low light images:
the aim of the research is to improve the image quality in the low light scene, so that the image is adapted to a target detection algorithm suitable for detecting the constant light image, thereby improving the target detection effect; the main objectives of the low light enhancement method are to reduce noise and increase contrast, and to present more details and information of the low light image. However, low light enhancement methods are often used as a preprocessing stage for the detector, which is not only time consuming, but may also destroy the information of the original input image; because the low light enhancement method improves the overall visual effect of the target image, problems of excessive enhancement or large noise interference generated in the enhancement process easily occur in the low light enhancement process, so that the low light image loses the original target detail characteristics.
(2) Research on field adaptation:
the aim of this study is to generate a composite low-light image by learning the distribution of the low-light scene, with which the detector is trained. However, this adaptation method is not sufficient to solve the huge semantic information gap caused by the severely changing light, so that the generated low-light image is different from the low-light image corresponding to the actual scene, and the effect of performing the target detection by this method is not ideal.
In summary, the existing method improves a dark target detection algorithm from the viewpoint of reconstructing image data, does not explore the internal relation between constant light and weak light images, and has the defect that original features of the images are lost due to reconstructing the image data, so that the target detection effect is affected.
Disclosure of Invention
The invention provides a dark target detection method and a dark target detection system based on dark transformation characteristics and the like, which are used for overcoming the defect that original characteristics of an image are lost due to reconstructed image data in the prior art, so that the target detection effect is affected.
In order to solve the technical problems, the technical scheme of the invention is as follows:
the dark target detection method based on the dark transformation characteristic constant change is characterized by comprising the following steps of:
s1: analyzing and modeling the physical sensor noise generated in the imaging process to obtain a physical noise model; inputting the constant light image into the physical noise model to obtain a low light image subjected to dark conversion;
s2: inputting the constant light image and the low light image into a detector for dark target detection, wherein a feature etc. conversion network is configured in the detector and is used for generating a first feature vector corresponding to the constant light image and the low light image and a second feature vector containing nonlinear information, which is obtained by converting the first feature vector;
s3: determining consistency loss for feature change according to the first feature vector and the second feature vector, determining training loss of the detector by using the consistency loss, optimizing the training loss, and updating the detector through the optimized training loss;
s4: and inputting the image to be detected into a detector which is updated, and obtaining a dark target detection result.
Preferably, the invention also provides a dark target detection system based on the change of the dark conversion characteristics and the like, which is applied to the dark target detection method based on the change of the dark conversion characteristics and the like. The dark target detection system based on the dark transformation characteristics and the like comprises a preprocessing module, a characteristic extraction module and a detection module which are connected in sequence.
In the technical scheme, the preprocessing module is used for acquiring a constant light image data set and generating a low light image data set by utilizing the constant light image data set;
the feature extraction module is used for training the neural network through the constant-light image and low-light image data set, inputting the image to be detected into the trained neural network, and extracting the features of the image to be detected;
and the detection module is used for outputting a detection result by inputting the characteristics of the image to be detected.
Compared with the prior art, the technical scheme of the invention has the beneficial effects that:
according to the imaging principle of a camera, physical sensor noise generated in the imaging process is analyzed and modeled, and the imaging process in an actual scene can be truly simulated, so that a normal light image is generated to be a corresponding vivid low light image through dark conversion, and then the constant characteristics of the low light environment and the image in the normal light environment are learned by utilizing characteristics and the like, so that a detector can explore the inherent relation between the normal light image and the weak light image, the more compact representation of the dark image is captured, and the target detection effect is improved;
meanwhile, the method focuses on restricting the detector to learn more light and dark invariance characteristics, rather than improving the performance of the detector through an additional module or special structural design, so that the method is suitable for various mainstream target detectors, does not influence the original good performance of the target detector when detecting the normally-light image, and can simultaneously maintain excellent performance in a normally-light environment and a low-light environment.
Drawings
Fig. 1 is a flowchart of a dark target detection method based on the dark conversion feature constant of embodiment 1;
fig. 2 is a flowchart of embodiment 2 in which a dark target detection method based on a dark conversion feature is applied to a YOLO detector;
fig. 3 is an overall frame diagram of a dark target detection system based on the dark conversion feature etc. of embodiment 3.
Detailed Description
The drawings are for illustrative purposes only and are not to be construed as limiting the present patent;
for the purpose of better illustrating the embodiments, certain elements of the drawings may be omitted, enlarged or reduced and do not represent the actual product dimensions;
it will be appreciated by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.
The technical scheme of the invention is further described below with reference to the accompanying drawings and examples.
Example 1
The present embodiment proposes a dark target detection method and system based on the change of the dark conversion feature, as shown in fig. 1, which is a flowchart of the dark target detection method based on the change of the dark conversion feature.
The dark target detection method based on the dark conversion characteristic constant change, provided by the embodiment, comprises the following steps:
s1: analyzing and modeling the physical sensor noise generated in the imaging process to obtain a physical noise model; inputting the constant light image into the physical noise model to obtain a low light image subjected to dark conversion;
s2: inputting the constant light image and the low light image into a detector for dark target detection, wherein a feature etc. conversion network is configured in the detector and is used for generating a first feature vector corresponding to the constant light image and the low light image and a second feature vector containing nonlinear information, which is obtained by converting the first feature vector;
s3: determining consistency loss for feature change according to the first feature vector and the second feature vector, determining training loss of the detector by using the consistency loss, optimizing the training loss, and updating the detector through the optimized training loss;
s4: and inputting the image to be detected into a detector which is updated, and obtaining a dark target detection result.
In the specific implementation process, the imaging process in the actual scene is simulated more truly by analyzing and modeling the physical sensor noise generated in the imaging process; performing dark conversion processing on the constant-light image by using the physical sensor noise to generate a vivid low-light image; taking the constant light image and the corresponding low light image generated by the constant light image as training data of the detector, and optimizing the training loss of the detector; the image to be detected is input to the optimized detector, and the detection result is output.
The method comprises the steps of utilizing a transformation network such as the characteristics configured in a detector to learn the invariant characteristics of images in a low-light environment and a normal-light environment, enabling the detector to explore the internal relation between the normal-light image and the low-light image, capturing a more compact representation of the dark image, and improving the target detection effect; meanwhile, the good performance of the detector in the original detection of the constant light image is not affected, so that the detector can maintain excellent performance in the constant light environment and the low light environment.
In an alternative embodiment, the step S1 includes:
s1.1: obtaining simulated lens noise by adding random noise subject to poisson distribution; wherein the lens noise delta s The calculation formula of (1) comprises:
wherein S is the optical signal collected by the camera;
s1.2: random noise which obeys Gaussian distribution of zero mean value and fixed variance is generated, and superimposed with an original image to obtain simulated reading noise;
wherein the read noise delta r Obeying a gaussian distribution, the expression of which comprises:
δ r ～N(0，1)；
s1.3: using a constant w, lens noise delta s And read noise delta r Composing the physical noise model x noise The method comprises the steps of carrying out a first treatment on the surface of the Wherein x is noise The expression of (2) includes:
wherein the constant w represents a factor of linear degradation and x represents a true value of a pixel of the input image;
s1.4: inputting a constant light image into a physical noise model to obtain a low light image xD after dark conversion:
x D ＝wx+x noise 。
in an implementation, the pixel values of the input image are linearly attenuated by a constant w.
Wherein optionally, in step S1.3, the constant w is randomly sampled over the range of [0.01,1.0 ].
In this embodiment, the lens noise and the readout noise are main noise that occurs when photons reach the camera sensor through the lens and are converted into analog voltage signals during imaging; wherein the lens noise is caused by the uncertainty arrival of the photons captured in the camera, i.e. the volatility of the photons; since the number of photons is very limited, random fluctuations occur during the capturing process, which appear as irregular ripples of brightness on the image, i.e. lens noise; in the implementation process, the photon number of each pixel point is regarded as an independent event, and the poisson distribution is used for describing the distribution of lens noise on the assumption that the occurrence probability of each photon is equal.
The read noise is noise that occurs when electrons are converted to a digital signal by a preamplifier; the magnitude of the read noise is independent of any optical signal.
In an alternative embodiment, in step S2, the expression for consistency loss of feature invariance includes:
wherein p and p D First eigenvectors, z and z, respectively corresponding to the normally light image and the low light image D P and p respectively D The second eigenvector obtained after the conversion, cos (·,) represents the cosine similarity function, and deltach (·) is the gradient stop operation.
Optimizing the consistency loss, that is, reducing the consistency loss, can increase cosine similarity between the first feature vector and the second feature vector, that is, can increase feature similarity between the normally light image and the low light image; the aim of changing the characteristics can be achieved by minimizing the consistency loss.
In an alternative embodiment, the specific process of step S3 includes:
s3.1: adding a supervision detection loss L S Combined with supervision to detect loss L S Loss of consistency L DTE Determining a training loss of the detector; wherein the monitoring detects the loss L S The expression of (2) includes:
L S ＝L box +L cla +L obj
wherein L is box 、L cls And L obj Representing position loss, classification loss, and confidence loss, respectively;
the training loss expression of the detector includes:
L＝L S +α·L DTE
wherein alpha is a balance factor of a preset value;
s3.2: optimizing the training loss using a random gradient descent method, and updating the detector with parameters that minimize the training loss to obtain an updated detector.
In the embodiment, the consistency loss and the supervision detection loss are optimized at the same time, the purpose of changing characteristics and the like is achieved by optimizing the consistency loss, the characteristic distance between the constant light image and the low light image is minimized, the detector can explore the internal relation between the constant light image and the low light image, the more compact representation of the dark image is captured, and the target detection effect is improved; the detection capability of the detector is improved by optimizing the supervised detection loss.
In an alternative embodiment, the feature-level transformation network includes a backbone neural network, a projection layer, and a prediction layer; wherein the backbone neural network is used for feature extraction, the projection layer is used for mapping the first feature vector to the same feature space, the prediction layer is used for converting the first feature vector mapped to the same feature space to obtain a second feature vector containing nonlinear information.
Wherein optionally the backbone neural network comprises DarkNet-53.
Further optionally, the projection layer includes a projection MLP header, and the prediction layer includes a prediction MLP header.
Example 2
The present embodiment applies the dark target detection method based on the dark conversion feature etc. proposed in embodiment 1, and proposes the following performance comparative example.
As shown in fig. 2, a flowchart of the present embodiment is a flowchart of applying the dark target detection method based on the dark conversion feature etc. to the YOLO detector.
Dark target detection methods based on the dark transformation feature alike are compared on the ExDark dataset together with other excellent dark target detection methods including MBLLEN (multi-branched dim light enhanced network model) and Zero-DCE (Zero-Reference Deep Curve Estimation, zero reference depth curve evaluation).
In comparison, a dark target detection method based on dark conversion characteristics and the like is instantiated on the YOLOv3 detector, and other excellent dark target detection methods are implemented; wherein the YOLOv3 detector is provided with a DarkNet-53 main neural network and a detection head of YOLO.
The quantitative experimental results (AP) and the total average results (mAP) of this example are shown in table 1.
It can be seen from table 1 that the method proposed by the present invention gives the best results in most categories compared to other methods. Moreover, the method provided by the invention obviously improves the basic detector YOLOv3 and realizes the improvement of 3.4% mAP, which indicates that after the method provided by the invention is applied, the detector can learn more favorable information for detecting dark objects, and the result further proves that the characteristic learning capability of the detector on low-light images is favorable for improving through learning the variable characteristics such as dark transformation and the like.
Table 1 comparison of the results of the various methods
Example 3
The present embodiment proposes a dark target detection system based on the change of the dark conversion feature, etc., and the dark target detection method based on the change of the dark conversion feature, etc. proposed in embodiment 1 is applied.
As shown in fig. 3, an overall frame diagram of the dark target detection system based on the dark conversion feature etc. of the present embodiment is shown.
The dark target detection system based on the dark transformation feature is characterized by comprising:
the preprocessing module is used for acquiring a constant light image data set and generating a low light image data set by utilizing the constant light image data set;
the feature extraction module is used for training the neural network through the constant-light image and low-light image data set, inputting the image to be detected into the trained neural network, and extracting the features of the image to be detected;
and the detection module is used for outputting a detection result by inputting the characteristics of the image to be detected.
Optionally, the dark target detection system based on the change of the dark conversion features and the like further comprises a visualization module for visually displaying the detection result output by the detection module, so that a user can intuitively know the detection result, and the system is convenient to operate by the user.
The same or similar reference numerals correspond to the same or similar components;
the terms describing the positional relationship in the drawings are merely illustrative, and are not to be construed as limiting the present patent;
it is to be understood that the above examples of the present invention are provided by way of illustration only and not by way of limitation of the embodiments of the present invention. Other variations or modifications of the above teachings will be apparent to those of ordinary skill in the art. It is not necessary here nor is it exhaustive of all embodiments. Any modification, equivalent replacement, improvement, etc. which come within the spirit and principles of the invention are desired to be protected by the following claims.
Claims (10)
1. The dark target detection method based on the dark transformation characteristic constant change is characterized by comprising the following steps of:
s1: analyzing and modeling the physical sensor noise generated in the imaging process to obtain a physical noise model; inputting the constant light image into the physical noise model to obtain a low light image subjected to dark conversion;
s2: inputting the constant light image and the low light image into a detector for dark target detection, wherein a feature etc. conversion network is configured in the detector and is used for generating a first feature vector corresponding to the constant light image and the low light image and a second feature vector containing nonlinear information, which is obtained by converting the first feature vector;
s3: determining consistency loss for feature change according to the first feature vector and the second feature vector, determining training loss of the detector by using the consistency loss, optimizing the training loss, and updating the detector through the optimized training loss;
s4: and inputting the image to be detected into a detector which is updated, and obtaining a dark target detection result.
2. The dark target detection method based on the dark conversion feature invariance according to claim 1, wherein the S1 step includes:
s1.1: obtaining simulated lens noise by adding random noise subject to poisson distribution; wherein the lens noise delta s The calculation formula of (1) comprises:
wherein S is the optical signal collected by the camera;
s1.2: random noise which obeys Gaussian distribution of zero mean value and fixed variance is generated, and superimposed with an original image to obtain simulated reading noise;
wherein the read noise delta r Obeying a gaussian distribution, the expression of which comprises:
δ r ～N(0,1)；
s1.3: using a constant w, lens noise delta s And read noise delta r Composing the physical noise model x noise The method comprises the steps of carrying out a first treatment on the surface of the Wherein, the liquid crystal display device comprises a liquid crystal display device, noise the expression of (2) includes:
wherein the constant w represents a factor of linear degradation, representing a true value of a pixel of the input image;
s1.4: inputting the constant light image into a physical noise model to obtain a low light image x after dark conversion D ：
x D ＝x+x noise 。
3. The dark target detection method based on the dark conversion feature alike according to claim 2, wherein in step S1.3, the constant w is randomly sampled within the range of [0.01,1.0 ].
4. The dark target detection method based on the dark conversion feature invariance according to claim 1, wherein in the S2 step, the expression for the consistency loss of the feature invariance includes:
wherein p and p D First eigenvectors, z and z, respectively corresponding to the normally light image and the low light image D P and p respectively D The second eigenvector obtained after the conversion, cos (·,) represents the cosine similarity function, and deltach (·) is the gradient stop operation.
5. The dark target detection method based on the dark transformation feature invariance according to claim 4, wherein the specific process of step S3 comprises:
s3.1: adding a supervision detection loss L S Combined with supervision to detect loss L S Loss of consistency L DTE Determining a training loss of the detector; wherein the monitoring detects the loss L S The expression of (2) includes:
L S ＝L box +L cla +L obj
wherein L is box 、L cla And L obj Representing position loss, classification loss, and confidence loss, respectively;
the training loss expression of the detector includes:
L＝L S +α·L DTE
wherein alpha is a balance factor of a preset value;
s3.2: optimizing the training loss using a random gradient descent method, and updating the detector with parameters that minimize the training loss to obtain an updated detector.
6. The dark target detection method based on the dark transformation feature isomorphism according to any one of claims 1 to 5, wherein the feature isomorphism transformation network comprises a backbone neural network, a projection layer and a prediction layer; wherein the backbone neural network is used for feature extraction, the projection layer is used for mapping the first feature vector to the same feature space, the prediction layer is used for converting the first feature vector mapped to the same feature space to obtain a second feature vector containing nonlinear information.
7. The dark target detection method based on dark transformation feature invariance according to claim 6, wherein the backbone neural network comprises dark net-53.
8. The dark target detection method based on dark transform feature invariance according to claim 6, wherein the projection layer comprises a projection MLP header and the prediction layer comprises a prediction MLP header.
9. A dark target detection system based on the dark conversion feature isomorphism, applying the dark target detection method based on the dark conversion feature isomorphism according to any one of claims 1 to 8, characterized by comprising:
the preprocessing module is used for acquiring a constant light image data set and generating a low light image data set by utilizing the constant light image data set;
the feature extraction module is used for training the neural network through the constant-light image and low-light image data set, inputting the image to be detected into the trained neural network, and extracting the features of the image to be detected;
and the detection module is used for outputting a detection result by inputting the characteristics of the image to be detected.
10. The dark target detection system based on the dark conversion feature alike according to claim 9, further comprising a visualization module for visually displaying the detection result output by the detection module.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310755574.9A CN116843883A (en) | 2023-06-25 | 2023-06-25 | Dark target detection method and system based on dark transformation characteristic constant change |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310755574.9A CN116843883A (en) | 2023-06-25 | 2023-06-25 | Dark target detection method and system based on dark transformation characteristic constant change |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116843883A true CN116843883A (en) | 2023-10-03 |
Family
ID=88160939
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310755574.9A Pending CN116843883A (en) | 2023-06-25 | 2023-06-25 | Dark target detection method and system based on dark transformation characteristic constant change |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116843883A (en) |
-
2023
- 2023-06-25 CN CN202310755574.9A patent/CN116843883A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111209952B (en) | Underwater target detection method based on improved SSD and migration learning | |
| Park et al. | Single image haze removal with WLS-based edge-preserving smoothing filter | |
| WO2022237397A1 (en) | Method and apparatus for evaluating image authenticity, computer device, and storage medium | |
| CN113591968A (en) | Infrared weak and small target detection method based on asymmetric attention feature fusion | |
| CN112580521B (en) | Multi-feature true and false video detection method based on MAML (maximum likelihood markup language) element learning algorithm | |
| CN111612741A (en) | Accurate non-reference image quality evaluation method based on distortion recognition | |
| CN116402852A (en) | Dynamic high-speed target tracking method and device based on event camera | |
| CN111985314A (en) | ViBe and improved LBP-based smoke detection method | |
| Babu et al. | An efficient image dahazing using Googlenet based convolution neural networks | |
| WO2020087434A1 (en) | Method and device for evaluating resolution of face image | |
| Shit et al. | An encoder‐decoder based CNN architecture using end to end dehaze and detection network for proper image visualization and detection | |
| CN110910497B (en) | Method and system for realizing augmented reality map | |
| Fathy et al. | Benchmarking of pre-processing methods employed in facial image analysis | |
| CN117409244A (en) | SCKConv multi-scale feature fusion enhanced low-illumination small target detection method | |
| Roy et al. | WLMS-based Transmission Refined self-adjusted no reference weather independent image visibility improvement | |
| CN116843883A (en) | Dark target detection method and system based on dark transformation characteristic constant change | |
| Li et al. | Multi-scale fusion framework via retinex and transmittance optimization for underwater image enhancement | |
| CN117809168B (en) | Method and device for detecting inherent attribute characteristics based on underwater target | |
| CN113537397B (en) | Target detection and image definition joint learning method based on multi-scale feature fusion | |
| CN114998990B (en) | Method and device for identifying safety behaviors of personnel on construction site | |
| CN116452469B (en) | Image defogging processing method and device based on deep learning | |
| Thike et al. | Vehicle detection using upper local ternary features with SVM classification | |
| CN110096951B (en) | Video saliency detection method and device based on Boolean diagram theory and storage device | |
| Nabila | Person Re-Identification using Background Subtraction and Siamese Network for Pose Varians | |
| KONG et al. | CF-StyleGAN: Near-infrared image colorization of SE attention StyleGAN via color features |
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 |