CN111523536A - Self-adaptive road surface projectile intelligent detection method based on fast RCNN - Google Patents
Self-adaptive road surface projectile intelligent detection method based on fast RCNN Download PDFInfo
- Publication number
- CN111523536A CN111523536A CN202010279716.5A CN202010279716A CN111523536A CN 111523536 A CN111523536 A CN 111523536A CN 202010279716 A CN202010279716 A CN 202010279716A CN 111523536 A CN111523536 A CN 111523536A
- Authority
- CN
- China
- Prior art keywords
- road surface
- sample set
- detection
- target
- 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.)
- Pending
Links
Images
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
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
Abstract
The invention discloses a self-adaptive intelligent detection method for road surface sprinkles based on fast RCNN, which relates to the technical field of image processing and comprises the following steps: step 1, collecting road surface throwing object images to form a source sample set, and preprocessing the source sample set to obtain a training sample set; step 2, training a fast RCNN target detection network by utilizing a training sample set to obtain an initial pavement tossing object detector; step 3, deploying an initial road surface throwing object detector on the target road section to detect the road surface throwing objects, and recording a detection result; and 4, clustering and manually intervening the detection results, then labeling to obtain a target sample set, and carrying out online fine adjustment on the initial road surface projectile detector by using the target sample set, so that the accuracy of the target road section is improved, and the target road surface projectile detector is obtained. The method has higher accuracy and efficiency for identifying the road surface sprinkled objects, and can improve the identification precision of the unseen road surface scene and the unseen road surface sprinkled objects.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a self-adaptive intelligent detection method for road surface sprinkled objects based on fast RCNN.
Background
In recent years, the inclination of policy in China and the current situation of increasing amount of garbage catalyze the development of road cleaning industry, so that the garbage disposal market is further increased. The cost of collecting and transporting one ton of garbage from the front end to the post-treatment is about 300 plus 500 yuan, so that the treatment market of domestic garbage in China can reach 1000 hundred million per year, and the market can reach about 500 million per year by one road cleaning item. Meanwhile, road maintenance and cleaning intellectualization are also an important component of intelligent traffic construction, which are beneficial to developing road appearance and road appearance construction, effectively breaking the refuse surrounding, promoting refuse classification and collection, and improving maintenance work efficiency, and are also the requirements of the era of social development. The intelligent recognition of road surface spilled objects and garbage can assist road management units to reasonably plan and design road cleaning management modes, improve the quality of sanitation cleaning operation, and effectively reduce the road cleaning management cost.
At present, the intelligent garbage recognition algorithm based on computer vision mainly has two types: the first is a convolutional neural network model for improving CaffeNet in the application of improved CaffeNet model in water surface garbage recognition, which is proposed in the document to improve the accuracy of water surface garbage recognition. The model improves the size of convolution kernels and the number of the convolution kernels, and increases a layer of sparse structure, so that the capability of extracting the characteristics of the network model is enhanced, and the complexity of the network is reduced. The method can reduce the influence of water surface ripples, object reflection, bridges and the like on water surface garbage recognition, and has a good water surface garbage recognition effect. But the robustness of the method is poor, and the method is difficult to be directly applied to more complex environments such as a road surface and the like; the garbage recognition and detection algorithm based on SSD, which is proposed in the 'garbage recognition classification research based on SSD algorithm', utilizes data enhancement to improve the robustness of a model, can achieve the purpose of quickly and accurately recognizing different types of garbage, and the method has good real-time performance but insufficient recognition capability on small targets.
It can be seen that, for the detection of small targets, the current common method at home and abroad is to use a multi-scale image pyramid or to upsample the small targets into large targets for detection. However, these two processing methods will seriously affect the time efficiency of the algorithm, thereby making it difficult to deploy and run the model on the edge computing device with low computational power.
Another problem is that the object to be identified must generally have a more uniform and invariant visual characteristic. The environment on a road surface is complex and road surface projectiles are a vague and difficult to exhaustively define. They tend to have visual characteristics that are closely related (e.g., visual characteristics of wood chips and leaves are close) and have large intra-class differences (e.g., plastic bags of different sizes and colors are very different). Therefore, the existing target detection algorithm and the deployment and use mode thereof can bring the problems of missing detection and false detection.
Therefore, those skilled in the art are dedicated to developing an adaptive intelligent detection method for a road surface projectile based on fast RCNN, which overcomes the problem of insufficient generalization capability of the prior art, and can adaptively adjust and complete training aiming at different road surface scenes and emerging road surface projectiles, so as to gradually improve the accuracy of projectile identification.
Disclosure of Invention
In view of the above-mentioned defects in the prior art, the technical problems to be solved by the present invention are that when there are many road surface sprinklers and most of them are small targets, both the recognition accuracy and the time efficiency cannot be considered, and the false detection and the missing detection of the algorithm in the face of new road environment and sprinklers that do not appear in the training data cannot be considered.
In order to achieve the aim, the invention provides an adaptive intelligent detection method for a road surface projectile based on fast RCNN, which comprises the following steps:
step 1, collecting road surface throwing object images to form a source sample set, and preprocessing the source sample set to obtain a training sample set;
step 2, training a fast RCNN target detection network by using the training sample set to obtain an initial road surface projectile detector;
step 3, deploying the initial road surface throwing object detector on the target road section to detect the road surface throwing objects, and recording the detection result;
and 4, clustering and manually intervening the detection results and then labeling to obtain a target sample set, and carrying out online fine adjustment on the initial road surface projectile detector by using the target sample set, so that the accuracy of the target road section is improved, and the target road surface projectile detector is obtained.
Further, the step 1 further comprises:
and counting the occurrence frequency of various sprinklers in the source sample set, sequencing, and selecting the first N categories with the highest occurrence frequency as the N sprinklers to be detected.
And labeling each sample in the training sample set to obtain a sample label corresponding to the training sample set.
Further, the pre-processing comprises:
horizontally turning the image to obtain an image to be further processed;
randomly carrying out translation transformation, rotation transformation and color transformation on the image to be further processed to obtain an image to be normalized;
and carrying out normalization processing on the image to be normalized.
Further, the step 2 further comprises:
step 2.1, training the Faster RCNN target detection network by adopting an SGD random gradient descent algorithm, setting the maximum iteration times, calculating a network error in a minimum batch mode (mini-batch) in each iteration, and updating parameters in the Faster RCNN target detection network by using the network error;
and 2.2, terminating the training when the preset maximum iteration times are reached or the error rate on the verification set is not reduced any more, and obtaining the initial pavement tossing object detector.
Further, the network error is an error of the class label and the sample label output by the Faster RCNN target detection network.
Further, the step 3 comprises:
step 3.1, collecting the road surface throwing object image again as an original detection image;
and 3.2, preprocessing the original detection image to obtain a detection image, and forming a detection sample set.
3.3, extracting the characteristics of the detection image by using a backbone network in the initial tossing object detector, and extracting all levels of characteristics of the network to form a characteristic pyramid;
step 3.4, feature classification;
step 3.5, drawing a detection result of the sprinkled object in the detection image, and outputting the result to finish the detection of the sprinkled object in the current picture;
and 3.6, storing the original detection image and the detection result.
Further, the feature classification is to classify the features by using a classification network in the initial projectile detector, a threshold is set, if an output value is greater than the threshold, it is determined that the image contains a projectile to be detected, otherwise, it is determined as a background.
Further, the detection result is coordinates and confidence of the target output by the initial road surface projectile detector.
Further, the step 4 comprises:
step 4.1, setting a confidence threshold, traversing the confidence, directly adding the sample with the confidence greater than the confidence threshold into the training sample set, and adding the sample with the confidence less than the confidence threshold into a post-processing sample set;
step 4.2, re-labeling the targets in the post-processing sample set, and adding the targets into the training sample set to obtain a target sample set;
and 4.3, repeating the step 2.1 and the step 2.2 on the initial pavement tossing detector by using the target sample set for fine adjustment. And obtaining the target pavement tossing object detector after the preset iteration times are finished.
Furthermore, the collecting of the road surface throwing object images is to intercept N images containing road surface throwing object samples from videos shot by the sweeper at each road section, and the image mode is RGB.
Compared with the prior art, the invention at least has the following beneficial technical effects:
1. the accuracy and efficiency of identifying the road surface throwing objects by using the fast RCNN algorithm are high, and real-time target detection and identification of videos or images shot by the garbage sweeper can be realized by using specific low-power-consumption edge GPU computing equipment (such as NVIDIA AGX Xavier). The recognition result provided by the algorithm can lay a good foundation for the pavement intelligent evaluation and scoring in the later period;
2. the iterative training method adopted by the method has good generalization capability and can improve the identification precision of the unseen road scene and the unseen road sprinkled object.
The conception, the specific structure and the technical effects of the present invention will be further described with reference to the accompanying drawings to fully understand the objects, the features and the effects of the present invention.
Drawings
FIG. 1 is a flow chart of a method in accordance with a preferred embodiment of the present invention;
FIG. 2 is a diagram of raw data for a preferred embodiment of the present invention;
FIG. 3 is a diagram of training sample effects in accordance with a preferred embodiment of the present invention;
FIG. 4 is a diagram illustrating the detection effect of the preferred embodiment of the present invention.
Detailed Description
The technical contents of the preferred embodiments of the present invention will be more clearly and easily understood by referring to the drawings attached to the specification. The present invention may be embodied in many different forms of embodiments and the scope of the invention is not limited to the embodiments set forth herein.
In the drawings, structurally identical elements are represented by like reference numerals, and structurally or functionally similar elements are represented by like reference numerals throughout the several views. The size and thickness of each component shown in the drawings are arbitrarily illustrated, and the present invention is not limited to the size and thickness of each component. The thickness of the components may be exaggerated where appropriate in the figures to improve clarity.
As shown in fig. 1, the invention is a flow chart of an adaptive intelligent detection method for road surface sprinkled objects based on fast RCNN, comprising the following steps:
step 1, collecting road surface throwing object images to form a source sample set, and preprocessing the source sample set to obtain a training sample set;
step 2, training a fast RCNN target detection network by utilizing a training sample set to obtain an initial pavement tossing object detector;
step 3, deploying an initial road surface throwing object detector on the target road section to detect the road surface throwing objects, and recording a detection result;
and 4, clustering and manually intervening the detection results, then labeling to obtain a target sample set, and carrying out online fine adjustment on the initial road surface projectile detector by using the target sample set, so that the accuracy of the target road section is improved, and the target road surface projectile detector is obtained.
Wherein the pretreatment comprises: horizontally turning the image; randomly performing translation transformation, rotation transformation and color transformation on the image; and (5) carrying out image normalization processing. Meanwhile, the image is in RGB mode.
The step 1 further comprises: counting the occurrence frequency of various sprinklers in the source sample set, sequencing, and selecting the first N categories with the highest occurrence frequency as N sprinklers to be detected; and labeling each sample in the training sample set to obtain a sample label corresponding to the training sample set.
In this example, 778 images containing samples of road surface projectiles were collected from video captured by a camera mounted on a refuse sweeper. The original image pixel size is 4096 × 2160, and small images of the area covering the road pavement body are cut out and uniformly scaled to 700 × 700, as shown in the original data diagram of fig. 2.
Counting each category, selecting 7 categories with the highest frequency of occurrence as categories of the pavement throwers to be identified, wherein the categories are as follows: stones, paper boxes, paper sheets, plastic bags, plastic bottles, wood chips, and leaves.
To increase the robustness of the detector, the image is subjected to random horizontal flipping, translation transformation (transformation range between negative 5 pixels to positive 5 pixels and inclusive of negative 5 pixels and positive 5 pixels) and rotation transformation (transformation range between negative 30 degrees to positive 30 degrees and inclusive of negative 30 degrees and positive 30 degrees). And then, dividing the pixel value of the image by 255, and normalizing the value range of the pixel between 0 and 1 and including 0 and 1 to obtain a training sample set. At the same time, the samples appearing in the training sample set are labeled.
The step 2 further comprises:
2.1, training a Faster RCNN target detection network by adopting an SGD random gradient descent algorithm, setting the maximum iteration times, calculating a network error by adopting a minimum batch mode (mini-batch) in each iteration, and updating parameters in the Faster RCNN target detection network by using the network error;
wherein the network error is an error between the classification label output by the Faster RCNN target detection network and the sample label.
And 2.2, terminating the training when a preset maximum iteration number is reached or the error rate on the verification set is not reduced any more, and obtaining the initial pavement toss detector.
In the present embodiment, the maximum number of iterations is set to 1300, the learning rate is set to 0.003, and 64 samples are input per iteration. The effect graph of the training sample is shown in fig. 3.
The step 3 comprises the following steps:
step 3.1, acquiring a detection image from a camera on the sweeper as an original detection image;
and 3.2, preprocessing the original detection image, checking the original detection image by using a mean value with the size of 5x5, filtering the mean value to remove noise in the image, dividing all pixel values of the image by 255 to enable the value range to be limited between 0 and 1 and to include 0 and 1, obtaining a detection image, and forming a detection sample set.
Step 3.3, extracting image features, and constructing a feature pyramid: extracting the characteristics of a detected image by using a backbone network in an initial tossing object detector, and extracting all levels of characteristics of the network to form a characteristic pyramid;
step 3.4, feature classification: the classification network in the initial missile detector is used to classify the features, and a threshold is set, in this embodiment, the threshold is set to 0.7. If the output value is larger than the threshold value, judging that the image contains the to-be-detected projectile, otherwise, judging that the image is a background;
and 3.5, drawing a detection result of the projectile in the detection image, such as a detection effect image shown in fig. 4. And the detection result is the coordinates and confidence of the target output by the initial pavement sprinkle detector, and the result is output to finish the detection of the sprinkle on the current picture.
And 3.6, storing the original detection image and the detection result.
Step 4 comprises the following steps:
and 4.1, setting a confidence threshold, wherein in the embodiment, the confidence threshold is set to be 0.8. Traversing the confidence, directly adding the samples with the confidence greater than the confidence threshold into the training sample set, and adding the samples with the confidence less than the confidence threshold into the post-processing sample set;
step 4.2, re-labeling the targets in the post-processing sample set, and adding the targets into the training sample set to obtain a target sample set;
and 4.3, repeating the step 2.1 and the step 2.2 to the pavement tossing detector by using the target sample set for fine adjustment. After the preset iteration times are completed, a relatively stable target road surface projectile detector with self-adaptive capacity to the target road section and the projectile is obtained. In this embodiment, 2 iterations can result in a projectile detector that performs better on a particular road segment.
The foregoing detailed description of the preferred embodiments of the invention has been presented. It should be understood that numerous modifications and variations could be devised by those skilled in the art in light of the present teachings without departing from the inventive concepts. Therefore, the technical solutions available to those skilled in the art through logic analysis, reasoning and limited experiments based on the prior art according to the concept of the present invention should be within the scope of protection defined by the claims.
Claims (10)
1. An adaptive pavement shed intelligent detection method based on fast RCNN is characterized by comprising the following steps:
step 1, collecting road surface throwing object images to form a source sample set, and preprocessing the source sample set to obtain a training sample set;
step 2, training a fast RCNN target detection network by using the training sample set to obtain an initial road surface projectile detector;
step 3, deploying the initial road surface throwing object detector on the target road section to detect the road surface throwing objects, and recording the detection result;
and 4, clustering and manually intervening the detection results and then labeling to obtain a target sample set, and carrying out online fine adjustment on the initial road surface projectile detector by using the target sample set, so that the accuracy of the target road section is improved, and the target road surface projectile detector is obtained.
2. The adaptive intelligent pavement projectile detection method based on fast RCNN according to claim 1, wherein said step 1 further comprises:
and counting the occurrence frequency of various sprinklers in the source sample set, sequencing, and selecting the first N categories with the highest occurrence frequency as the N sprinklers to be detected.
And labeling each sample in the training sample set to obtain a sample label corresponding to the training sample set.
3. The adaptive pavement shed intelligent detection method based on fast RCNN according to claim 1, wherein the preprocessing comprises:
horizontally turning the image to obtain an image to be further processed;
randomly carrying out translation transformation, rotation transformation and color transformation on the image to be further processed to obtain an image to be normalized;
and carrying out normalization processing on the image to be normalized.
4. The adaptive pavement shed intelligent detection method based on fast RCNN according to claim 1, wherein the step 2 further comprises:
step 2.1, training the Faster RCNN target detection network by adopting an SGD random gradient descent algorithm, setting the maximum iteration times, calculating a network error in a minimum batch mode (mini-batch) in each iteration, and updating parameters in the Faster RCNN target detection network by using the network error;
and 2.2, terminating the training when the preset maximum iteration times are reached or the error rate on the verification set is not reduced any more, and obtaining the initial pavement tossing object detector.
5. The method according to claim 4, wherein the network error is an error between the classification label and the sample label output by the fast RCNN target detection network.
6. The adaptive pavement shed intelligent detection method based on fast RCNN according to claim 1, wherein the step 3 comprises:
step 3.1, collecting the road surface throwing object image again as an original detection image;
and 3.2, preprocessing the original detection image to obtain a detection image, and forming a detection sample set.
3.3, extracting the characteristics of the detection image by using a backbone network in the initial tossing object detector, and extracting all levels of characteristics of the network to form a characteristic pyramid;
step 3.4, feature classification;
step 3.5, drawing a detection result of the sprinkled object in the detection image, and outputting the result to finish the detection of the sprinkled object in the current picture;
and 3.6, storing the original detection image and the detection result.
7. The method according to claim 6, wherein the feature classification is to classify the features by using a classification network in the initial projectile detector, and set a threshold, if an output value is greater than the threshold, the image is determined to contain the projectile to be detected, otherwise, the image is determined to be the background.
8. The method for intelligent adaptive pavement spray detection by fast RCNN according to claim 6, wherein the detection result is the coordinates and confidence of the target output by the initial pavement spray detector.
9. The adaptive pavement shed intelligent detection method based on fast RCNN according to claim 1, wherein the step 4 comprises:
step 4.1, setting a confidence threshold, traversing the confidence, directly adding the sample with the confidence greater than the confidence threshold into the training sample set, and adding the sample with the confidence less than the confidence threshold into a post-processing sample set;
step 4.2, re-labeling the targets in the post-processing sample set, and adding the targets into the training sample set to obtain a target sample set;
and 4.3, repeating the step 2.1 and the step 2.2 on the initial pavement tossing detector by using the target sample set for fine adjustment. And obtaining the target pavement tossing object detector after the preset iteration times are finished.
10. The adaptive intelligent detection method for road surface sprinklers based on fast RCNN according to claim 1 or 6, wherein the collecting of the road surface sprinklers images is performed by intercepting N images containing road surface sprinklers samples from videos shot by the sweeper truck of each road section, wherein the image mode is RGB.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010279716.5A CN111523536A (en) | 2020-04-10 | 2020-04-10 | Self-adaptive road surface projectile intelligent detection method based on fast RCNN |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010279716.5A CN111523536A (en) | 2020-04-10 | 2020-04-10 | Self-adaptive road surface projectile intelligent detection method based on fast RCNN |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111523536A true CN111523536A (en) | 2020-08-11 |
Family
ID=71902434
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010279716.5A Pending CN111523536A (en) | 2020-04-10 | 2020-04-10 | Self-adaptive road surface projectile intelligent detection method based on fast RCNN |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111523536A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112115982A (en) * | 2020-08-27 | 2020-12-22 | 中犹(南京)智慧城市创新研究院有限公司 | Yoov 3-based automatic detection method for road surface leakage diseases |
| CN112446335A (en) * | 2020-12-02 | 2021-03-05 | 电子科技大学中山学院 | Terahertz contraband detection method based on deep learning |
| CN113221724A (en) * | 2021-05-08 | 2021-08-06 | 杭州鸿泉物联网技术股份有限公司 | Vehicle spray detection method and system |
| CN113688825A (en) * | 2021-05-17 | 2021-11-23 | 海南师范大学 | AI intelligent garbage recognition and classification system and method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109165582A (en) * | 2018-08-09 | 2019-01-08 | 河海大学 | A kind of detection of avenue rubbish and cleannes appraisal procedure |
| CN109753890A (en) * | 2018-12-18 | 2019-05-14 | 吉林大学 | A kind of pavement garbage object intelligent recognition and cognitive method and its realization device |
| CN110659622A (en) * | 2019-09-27 | 2020-01-07 | 北京文安智能技术股份有限公司 | Detection method, device and system for garbage dumping |
| CN110991466A (en) * | 2019-11-21 | 2020-04-10 | 同济大学 | Highway road surface condition detecting system based on novel vision sensing equipment |
-
2020
- 2020-04-10 CN CN202010279716.5A patent/CN111523536A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109165582A (en) * | 2018-08-09 | 2019-01-08 | 河海大学 | A kind of detection of avenue rubbish and cleannes appraisal procedure |
| CN109753890A (en) * | 2018-12-18 | 2019-05-14 | 吉林大学 | A kind of pavement garbage object intelligent recognition and cognitive method and its realization device |
| CN110659622A (en) * | 2019-09-27 | 2020-01-07 | 北京文安智能技术股份有限公司 | Detection method, device and system for garbage dumping |
| CN110991466A (en) * | 2019-11-21 | 2020-04-10 | 同济大学 | Highway road surface condition detecting system based on novel vision sensing equipment |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112115982A (en) * | 2020-08-27 | 2020-12-22 | 中犹(南京)智慧城市创新研究院有限公司 | Yoov 3-based automatic detection method for road surface leakage diseases |
| CN112446335A (en) * | 2020-12-02 | 2021-03-05 | 电子科技大学中山学院 | Terahertz contraband detection method based on deep learning |
| CN113221724A (en) * | 2021-05-08 | 2021-08-06 | 杭州鸿泉物联网技术股份有限公司 | Vehicle spray detection method and system |
| CN113221724B (en) * | 2021-05-08 | 2022-11-04 | 杭州鸿泉物联网技术股份有限公司 | Vehicle spray detection method and system |
| CN113688825A (en) * | 2021-05-17 | 2021-11-23 | 海南师范大学 | AI intelligent garbage recognition and classification system and method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111523536A (en) | Self-adaptive road surface projectile intelligent detection method based on fast RCNN | |
| Nguyen et al. | Robust minutiae extractor: Integrating deep networks and fingerprint domain knowledge | |
| Rachmadi et al. | Vehicle color recognition using convolutional neural network | |
| CN110414414B (en) | SAR image ship target identification method based on multilevel feature depth fusion | |
| CN101859382B (en) | License plate detection and identification method based on maximum stable extremal region | |
| Abedin et al. | License plate recognition system based on contour properties and deep learning model | |
| CN107103317A (en) | Fuzzy license plate image recognition algorithm based on image co-registration and blind deconvolution | |
| Bianco et al. | Logo recognition using cnn features | |
| CN106845408A (en) | A kind of street refuse recognition methods under complex environment | |
| CN102998316B (en) | Transparent liquid impurity detection system and detection method thereof | |
| Fahmy et al. | A fingerprint segmentation technique based on morphological processing | |
| CN106778742B (en) | Car logo detection method based on Gabor filter background texture suppression | |
| CN107944354B (en) | Vehicle detection method based on deep learning | |
| CN109145742A (en) | A kind of pedestrian recognition method and system | |
| CN107092876A (en) | The low-light (level) model recognizing method combined based on Retinex with S SIFT features | |
| CN103345631A (en) | Image characteristic extraction, training, detection method, module, device and system | |
| CN114937179B (en) | Junk image classification method and device, electronic equipment and storage medium | |
| CN113221881B (en) | Multi-level smart phone screen defect detection method | |
| Ming et al. | Defect detection of LGP based on combined classifier with dynamic weights | |
| Chen et al. | Automatic sorting of fresh tea leaves using vision‐based recognition method | |
| Zhang et al. | A novel pavement crack detection approach using pre-selection based on transfer learning | |
| Sui et al. | ECGAN: An improved conditional generative adversarial network with edge detection to augment limited training data for the classification of remote sensing images with high spatial resolution | |
| CN111738237B (en) | Heterogeneous convolution-based target detection method for multi-core iteration RPN | |
| CN108288276A (en) | A kind of interference filtering method in projection interactive system under touch mode | |
| CN108765448A (en) | A kind of shrimp seedling analysis of accounts method based on improvement TV-L1 models |
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 |