CN113763392B - Model prediction method and system for road surface flatness detection and intelligent terminal - Google Patents
Model prediction method and system for road surface flatness detection and intelligent terminal Download PDFInfo
- Publication number
- CN113763392B CN113763392B CN202111324097.8A CN202111324097A CN113763392B CN 113763392 B CN113763392 B CN 113763392B CN 202111324097 A CN202111324097 A CN 202111324097A CN 113763392 B CN113763392 B CN 113763392B
- Authority
- CN
- China
- Prior art keywords
- convolution
- result
- loss
- prediction
- value
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30248—Vehicle exterior or interior
- G06T2207/30252—Vehicle exterior; Vicinity of vehicle
Abstract
The invention discloses a model prediction method, a system and an intelligent terminal for detecting the flatness of a road surface, wherein the method comprises the following steps: the method comprises the steps of obtaining a pavement image of a target scene, preprocessing the pavement image, and processing the pavement image through a semantic segmentation neural network model to obtain a prediction result and a real result of probability that each pixel point belongs to each pixel category; performing convolution on the prediction result by combining with a Gaussian Laplacian operator to obtain a prediction convolution value, and performing convolution on the real result by combining with the Gaussian Laplacian operator to obtain a real convolution value; and calculating loss values based on the predicted convolution values and the real convolution values, and reversely propagating the loss values to update the weights of the neural network model until the model training is finished. The technical problem that in the prior art, the detection efficiency of the road surface flatness is low due to the fact that isolated points exist in an existing semantic segmentation model is solved.
Description
Technical Field
The invention relates to the technical field of automatic driving assistance, in particular to a model prediction method and system for detecting road flatness and an intelligent terminal.
Background
With the development of automatic driving technology, people have increasingly higher requirements on safety and comfort of vehicles for assisting driving. In the semantic segmentation based on deep learning, the obtained result is the semantic category of each pixel point on the image, however, a strong constraint relation does not exist between each pixel point, which easily causes the network to predict isolated points, namely, the category of one or a plurality of pixel points is different from the categories of surrounding pixel points. In a real scene, the pixel distribution of the same category is continuous and closed, and isolated points do not exist. The existence of the isolated points in the category increases the number of objects in front of the wheels, and finally increases the post-processing time of the road flatness detection algorithm.
Disclosure of Invention
Therefore, the embodiment of the invention provides a model prediction method, a system and an intelligent terminal for detecting the road flatness, so as to at least partially solve the technical problem of low road flatness detection efficiency caused by isolated points existing in a semantic segmentation model in the prior art.
In order to achieve the above object, the embodiments of the present invention provide the following technical solutions:
a model prediction method for road flatness detection, the method comprising:
the method comprises the steps of obtaining a pavement image of a target scene, preprocessing the pavement image, and processing the pavement image through a semantic segmentation neural network model to obtain a prediction result and a real result of probability that each pixel point belongs to each pixel category;
performing convolution on the prediction result by combining with a Gaussian Laplacian operator to obtain a prediction convolution value, and performing convolution on the real result by combining with the Gaussian Laplacian operator to obtain a real convolution value;
and calculating loss values based on the predicted convolution values and the real convolution values, and reversely propagating the loss values to update the weights of the neural network model until the model training is finished.
Further, the preprocessing the road surface image specifically includes:
and acquiring a road surface image to be segmented, and cutting, normalizing and data enhancing the road surface image to obtain a target area in the road surface image.
Further, a smooth convolution kernel of the laplacian of gaussian is calculated using the following formula:
wherein x and y represent the coordinates of a certain pixel in the image, sigma represents the standard deviation, pi is the circumferential ratio, and e is a natural constant.
Further, outliers are detected using the laplacian based on the following formula:
wherein x and y represent the coordinates of a certain pixel in the image, sigma represents the standard deviation, pi is the circumferential ratio, and e is a natural constant.
Further, the loss value is calculated using the following formula:
loss=loss1+λ*loss2
wherein loss is the model total loss value;
loss1 is a predicted loss value calculated based on the predicted convolution value;
loss2 is the true loss value calculated based on the true convolution value;
λ is the weight.
The present invention also provides a model prediction system for road flatness detection, the system comprising:
the class probability prediction unit is used for acquiring a pavement image of a target scene, preprocessing the pavement image and then obtaining a prediction result and a real result of each pixel point belonging to each pixel class probability through a semantic segmentation neural network model;
a convolution value obtaining unit, configured to perform convolution on the prediction result in combination with a laplacian of gaussian operator to obtain a prediction convolution value, and perform convolution on the real result in combination with the laplacian of gaussian operator to obtain a real convolution value;
and the loss value acquisition unit is used for calculating a loss value based on the predicted convolution value and the real convolution value and reversely propagating the loss value to update the weight of the neural network model until the model training is finished.
The present invention also provides an intelligent terminal, including: the device comprises a data acquisition device, a processor and a memory;
the data acquisition device is used for acquiring data; the memory is to store one or more program instructions; the processor is configured to execute one or more program instructions to perform the method as described above.
The present invention also provides a computer readable storage medium having embodied therein one or more program instructions for executing the method as described above.
According to the model prediction method for detecting the road flatness, provided by the invention, the loss function is added in the deep learning-based semantic segmentation training process, so that the prediction of semantic segmentation isolated points can be reduced, the model prediction effect is improved, the effect of reducing the isolated points can be played in a road flatness detection task, the post-processing time is reduced, and the detection algorithm efficiency is improved. The technical problem that in the prior art, the detection efficiency of the road surface flatness is low due to the isolated points in the semantic segmentation model is solved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below. It should be apparent that the drawings in the following description are merely exemplary, and that other embodiments can be derived from the drawings provided by those of ordinary skill in the art without inventive effort.
The structures, ratios, sizes, and the like shown in the present specification are only used for matching with the contents disclosed in the specification, so as to be understood and read by those skilled in the art, and are not used to limit the conditions that the present invention can be implemented, so that the present invention has no technical significance, and any structural modifications, changes in the ratio relationship, or adjustments of the sizes, without affecting the effects and the achievable by the present invention, should still fall within the range that the technical contents disclosed in the present invention can cover.
Fig. 1 is a flowchart of a binocular stereo vision-based road flatness grade detection method according to an embodiment of the present invention;
FIG. 2 is a diagram of a convolution template with a convolution kernel size of 3 x 3 for a particular usage scenario;
FIG. 3 is a diagram of a convolution template with a convolution kernel size of 5 x 5 for the particular use scenario shown in FIG. 2;
FIG. 4 is a prediction result before applying the Gaussian Laplacian to add local constraint loss addition in the specific usage scenario shown in FIG. 2;
FIG. 5 is a prediction result after local constraint loss addition is added by applying the Gaussian Laplace operator in the specific usage scenario shown in FIG. 2;
fig. 6 is a block diagram of a model prediction system for detecting road flatness according to an embodiment of the present invention.
Detailed Description
The present invention is described in terms of particular embodiments, other advantages and features of the invention will become apparent to those skilled in the art from the following disclosure, and it is to be understood that the described embodiments are merely exemplary of the invention and that it is not intended to limit the invention to the particular embodiments disclosed. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
When the road surface flatness is detected, if more isolated points exist on the pixel point category, the number of objects right in front of the wheels becomes larger, and finally the time for the road surface flatness post-processing is increased. The reason for the appearance of the isolated points is analyzed, that the adjacent pixels are not constrained during prediction, and the relationship between each pixel point is only constrained by the receptive fields between the neural networks, so that a large number of isolated points are predicted. Therefore, the model prediction method for detecting the road flatness can establish local association constraint between adjacent pixels, so that a network inhibits generation of isolated points, the semantic segmentation effect of the road flatness is finally improved, and the efficiency of post-processing of the flatness detection is improved.
In one embodiment, the model prediction method for detecting the flatness of the road surface provided by the invention comprises the following steps:
s1: the method comprises the steps of obtaining a pavement image of a target scene, preprocessing the pavement image, and obtaining a prediction result and a real result of probability that each pixel point belongs to each pixel category through a semantic segmentation neural network model.
Specifically, after a road surface image to be segmented is obtained, the road surface image is cut, normalized and data enhanced to obtain a target area in the road surface image.
In one usage scenario, assume that the total number of pixel classes is K. Firstly, a road surface image to be segmented is processed by calculation, cutting, normalization, data enhancement and the like of an interested region, and then is input into a semantic segmentation neural network, so that probability values of each pixel point belonging to K categories are obtained, and the range of the probability values is 0 to 1.
S2: and performing convolution on the prediction result by combining with a Gaussian Laplacian operator to obtain a prediction convolution value, and performing convolution on the real result by combining with the Gaussian Laplacian operator to obtain a real convolution value.
Because it is desirable that the isolated point can refer to the surrounding information when adding the constraint, and it is desirable that the more adjacent pixel point classes are more relevant to the class prediction of the central point, therefore, a two-dimensional gaussian smooth convolution kernel is adopted, and the formula is as follows:
where x and y represent coordinates of a certain pixel in an image, σ represents a standard deviation, π is a circumferential ratio, and e represents a natural constant. The laplacian can highlight the regions with rapidly changing intensity in the image, so that the laplacian can be reused to detect isolated points. Specifically, a Laplacian of Gaussian function (LoG) formula with a Gaussian standard deviation of σ centered at 0 is as follows:
where x and y represent coordinates of a certain pixel in an image, σ represents a standard deviation, π is a circumferential ratio, and e represents a natural constant.
In practical applications, in order not to change the expectation value of the gray value of the overall prediction tag after convolution, it is required that the sum of the elements of the overall convolution template is 0, and the values of the symmetric elements around the center or the axisymmetric elements are the same. Adjusting the value of the gaussian standard deviation sigma, and when the size of the convolution kernel is 3 x 3, generating a convolution template in a form shown in fig. 2; if the convolution kernel size is 5 x 5, the resulting convolution template is of the form shown in fig. 3. For comparison, the prediction result before adding the local constraint loss by applying the laplacian of gaussian is shown in fig. 4, and more holes are visible; the prediction result after adding the local constraint loss by applying the laplacian gaussian is shown in fig. 5, and it is obvious that the hole is removed.
S3: and calculating loss values based on the predicted convolution values and the real convolution values, and reversely propagating the loss values to update the weights of the neural network model until the model training is finished.
Specifically, the loss value is calculated using the following formula:
loss=loss1+λ*loss2
wherein loss is the model total loss value;
loss1 is a predicted loss value calculated based on the predicted convolution value;
loss2 is the true loss value calculated based on the true convolution value;
λ is the weight.
Still taking the above scenario as an example, for the predicted value and the true value of the neural network, a loss is calculated as loss 1. And (3) performing convolution on the real label by using a convolution template to obtain a result A, performing convolution on the prediction result of the neural network by using the convolution template to obtain a result B, and calculating loss values of the result A and the result B, wherein the loss values are recorded as loss 2. Final loss = loss1+ λ loss2, where λ is the weight that adjusts the importance of the two losses. Because the punishment of the prediction type error and the punishment of the predicted isolated point are considered during the error calculation, high-precision semantic segmentation can be realized and the generation of the isolated point is reduced. And after the error is calculated, performing back propagation, and updating the weight of the neural network until the network training is finished. After the model is trained, no real value supervision exists in the forward reasoning process, so that the prediction result of the model is directly used as output, and convolution is not required to be carried out by using a convolution template.
In the above specific embodiment, the model prediction method for detecting road flatness provided by the present invention adds a loss function in the semantic segmentation training process based on deep learning, so that the prediction of semantic segmentation isolated points can be reduced, the effect of model prediction can be improved, the model prediction method can play a role in reducing isolated points in the road flatness detection task, reduce the post-processing time, and improve the detection algorithm efficiency. The technical problem that in the prior art, the detection efficiency of the road surface flatness is low due to the isolated points in the semantic segmentation model is solved.
In addition to the above method, the present invention also provides a model prediction system for road surface flatness detection, as shown in fig. 6, and in a specific embodiment, the system includes:
the class probability prediction unit 100 is configured to obtain a road surface image of a target scene, preprocess the road surface image, and obtain a prediction result and a real result of each pixel point belonging to each pixel class probability through a semantic segmentation neural network model;
a convolution value obtaining unit 200, configured to perform convolution on the predicted result in combination with a laplacian of gaussian operator to obtain a predicted convolution value, and perform convolution on the real result in combination with the laplacian of gaussian operator to obtain a real convolution value;
and a loss value obtaining unit 300, configured to calculate a loss value based on the predicted convolution value and the real convolution value, and back-propagate the loss value to update the weight of the neural network model until the model training is finished.
In the above specific embodiment, the model prediction system for detecting road flatness provided by the present invention adds a loss function in the semantic segmentation training process based on deep learning, so that the prediction of semantic segmentation isolated points can be reduced, the effect of model prediction can be improved, the model prediction system can play a role in reducing isolated points in the road flatness detection task, reduce the post-processing time, and improve the detection algorithm efficiency. The technical problem that in the prior art, the detection efficiency of the road surface flatness is low due to the isolated points in the semantic segmentation model is solved.
The present invention also provides an intelligent terminal, including: the device comprises a data acquisition device, a processor and a memory;
the data acquisition device is used for acquiring data; the memory is to store one or more program instructions; the processor is configured to execute one or more program instructions to perform the method as described above.
In correspondence with the above embodiments, embodiments of the present invention also provide a computer storage medium containing one or more program instructions therein. Wherein the one or more program instructions are for executing the method as described above by a binocular camera depth calibration system.
In an embodiment of the invention, the processor may be an integrated circuit chip having signal processing capability. The Processor may be a general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, discrete Gate or transistor logic device, discrete hardware component.
The various methods, steps and logic blocks disclosed in the embodiments of the present invention may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of the method disclosed in connection with the embodiments of the present invention may be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software module may be located in ram, flash memory, rom, prom, or eprom, registers, etc. storage media as is well known in the art. The processor reads the information in the storage medium and completes the steps of the method in combination with the hardware.
The storage medium may be a memory, for example, which may be volatile memory or nonvolatile memory, or which may include both volatile and nonvolatile memory.
The nonvolatile Memory may be a Read-Only Memory (ROM), a Programmable ROM (PROM), an Erasable PROM (EPROM), an Electrically Erasable PROM (EEPROM), or a flash Memory.
The volatile Memory may be a Random Access Memory (RAM) which serves as an external cache. By way of example and not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), SLDRAM (SLDRAM), and Direct Rambus RAM (DRRAM).
The storage media described in connection with the embodiments of the invention are intended to comprise, without being limited to, these and any other suitable types of memory.
Those skilled in the art will appreciate that the functionality described in the present invention may be implemented in a combination of hardware and software in one or more of the examples described above. When software is applied, the corresponding functionality may be stored on or transmitted over as one or more instructions or code on a computer-readable medium. Computer-readable media includes both computer storage media and communication media including any medium that facilitates transfer of a computer program from one place to another. A storage media may be any available media that can be accessed by a general purpose or special purpose computer.
The above embodiments are only for illustrating the embodiments of the present invention and are not to be construed as limiting the scope of the present invention, and any modifications, equivalent substitutions, improvements and the like made on the basis of the embodiments of the present invention shall be included in the scope of the present invention.
Claims (7)
1. A model prediction method for road flatness detection, the method comprising:
obtaining a pavement image of a target scene, preprocessing the pavement image, and obtaining a prediction result and a real result of probability that each pixel point belongs to each pixel category through a semantic segmentation neural network model;
performing convolution on the prediction result by combining with a Gaussian Laplacian operator to obtain a prediction convolution value, and performing convolution on the real result by combining with the Gaussian Laplacian operator to obtain a real convolution value;
calculating a loss value based on the predicted convolution value and the real convolution value, and reversely propagating the loss value to update the weight of the neural network model until the model training is finished;
calculating a loss value based on the predicted convolution value and the true convolution value, specifically comprising: calculating a loss for a prediction result and a real result of the semantic segmentation neural network model, recording the loss as loss1, performing convolution on the real result by using a convolution template to obtain a result A, performing convolution on the prediction result of the neural network by using the convolution template to obtain a result B, and calculating loss values of the result A and the result B as loss 2; final loss = loss1+ λ loss2, where λ is the weight that adjusts the importance of the two losses.
2. The model prediction method according to claim 1, wherein the preprocessing of the road surface image specifically comprises:
and acquiring a road surface image to be segmented, and cutting, normalizing and data enhancing the road surface image to obtain a target area in the road surface image.
3. The model prediction method of claim 1, wherein the smooth convolution kernel of the laplacian of gaussian is calculated using the following formula:
Wherein x and y represent the coordinates of a certain pixel in the image, sigma represents the standard deviation, pi is the circumferential ratio, and e is a natural constant.
4. The model prediction method of claim 3, characterized in that outliers are detected using the Laplace operator based on the following formula:
Wherein x and y represent the coordinates of a certain pixel in the image, sigma represents the standard deviation, pi is the circumferential ratio, and e is a natural constant.
5. A model prediction system for road flatness detection, the system comprising:
the class probability prediction unit is used for preprocessing the road surface image after acquiring the road surface image of the target scene, and obtaining a prediction result and a real result of each pixel point belonging to each pixel class probability through a semantic segmentation neural network model;
a convolution value obtaining unit, configured to perform convolution on the prediction result in combination with a laplacian of gaussian operator to obtain a prediction convolution value, and perform convolution on the real result in combination with the laplacian of gaussian operator to obtain a real convolution value;
a loss value obtaining unit, configured to calculate a loss value based on the predicted convolution value and the real convolution value, and perform back propagation on the loss value to update the weight of the neural network model until model training is completed;
calculating a loss value based on the predicted convolution value and the true convolution value, specifically comprising: calculating a loss for a prediction result and a real result of the semantic segmentation neural network model, recording the loss as loss1, performing convolution on the real result by using a convolution template to obtain a result A, performing convolution on the prediction result of the neural network by using the convolution template to obtain a result B, and calculating loss values of the result A and the result B as loss 2; final loss = loss1+ λ loss2, where λ is the weight that adjusts the importance of the two losses.
6. An intelligent terminal, characterized in that, intelligent terminal includes: the device comprises a data acquisition device, a processor and a memory;
the data acquisition device is used for acquiring data; the memory is to store one or more program instructions; the processor, configured to execute one or more program instructions to perform the method of any of claims 1-4.
7. A computer-readable storage medium having one or more program instructions embodied therein for performing the method of any of claims 1-4.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111324097.8A CN113763392B (en) | 2021-11-10 | 2021-11-10 | Model prediction method and system for road surface flatness detection and intelligent terminal |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111324097.8A CN113763392B (en) | 2021-11-10 | 2021-11-10 | Model prediction method and system for road surface flatness detection and intelligent terminal |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113763392A CN113763392A (en) | 2021-12-07 |
CN113763392B true CN113763392B (en) | 2022-03-18 |
Family
ID=78784883
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111324097.8A Active CN113763392B (en) | 2021-11-10 | 2021-11-10 | Model prediction method and system for road surface flatness detection and intelligent terminal |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113763392B (en) |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104063680A (en) * | 2014-02-18 | 2014-09-24 | 中国科学院遥感与数字地球研究所 | Remote sensing recognition method for snow disturbance traces based on Gaussian differential model |
CN109215046A (en) * | 2018-08-06 | 2019-01-15 | 浙江工贸职业技术学院 | A kind of Laplace operator edge detection method based on image interpolation arithmetic |
CN110717886A (en) * | 2019-09-03 | 2020-01-21 | 南京理工大学 | Pavement pool detection method based on machine vision in complex environment |
CN111179272A (en) * | 2019-12-10 | 2020-05-19 | 中国科学院深圳先进技术研究院 | Rapid semantic segmentation method for road scene |
CN111242037A (en) * | 2020-01-15 | 2020-06-05 | 华南理工大学 | Lane line detection method based on structural information |
WO2020244108A1 (en) * | 2019-06-05 | 2020-12-10 | Boe Technology Group Co., Ltd. | Methods and apparatuses for semantically segmenting input image, and computer-program product |
CN112150410A (en) * | 2020-08-24 | 2020-12-29 | 浙江工商大学 | Automatic detection method and system for weld defects |
CN112949508A (en) * | 2021-03-08 | 2021-06-11 | 咪咕文化科技有限公司 | Model training method, pedestrian detection method, electronic device and readable storage medium |
CN113011305A (en) * | 2021-03-12 | 2021-06-22 | 中国人民解放军国防科技大学 | SAR image road extraction method and device based on semantic segmentation and conditional random field |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113011255B (en) * | 2021-02-05 | 2024-01-16 | 北京中科慧眼科技有限公司 | Road surface detection method and system based on RGB image and intelligent terminal |
-
2021
- 2021-11-10 CN CN202111324097.8A patent/CN113763392B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104063680A (en) * | 2014-02-18 | 2014-09-24 | 中国科学院遥感与数字地球研究所 | Remote sensing recognition method for snow disturbance traces based on Gaussian differential model |
CN109215046A (en) * | 2018-08-06 | 2019-01-15 | 浙江工贸职业技术学院 | A kind of Laplace operator edge detection method based on image interpolation arithmetic |
WO2020244108A1 (en) * | 2019-06-05 | 2020-12-10 | Boe Technology Group Co., Ltd. | Methods and apparatuses for semantically segmenting input image, and computer-program product |
CN110717886A (en) * | 2019-09-03 | 2020-01-21 | 南京理工大学 | Pavement pool detection method based on machine vision in complex environment |
CN111179272A (en) * | 2019-12-10 | 2020-05-19 | 中国科学院深圳先进技术研究院 | Rapid semantic segmentation method for road scene |
CN111242037A (en) * | 2020-01-15 | 2020-06-05 | 华南理工大学 | Lane line detection method based on structural information |
CN112150410A (en) * | 2020-08-24 | 2020-12-29 | 浙江工商大学 | Automatic detection method and system for weld defects |
CN112949508A (en) * | 2021-03-08 | 2021-06-11 | 咪咕文化科技有限公司 | Model training method, pedestrian detection method, electronic device and readable storage medium |
CN113011305A (en) * | 2021-03-12 | 2021-06-22 | 中国人民解放军国防科技大学 | SAR image road extraction method and device based on semantic segmentation and conditional random field |
Non-Patent Citations (1)
Title |
---|
数字图像特征点检测算法的研究;朱少何;《中国优秀博硕士学位论文全文数据库(硕士) 信息科技辑》;20160515(第05期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN113763392A (en) | 2021-12-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111062413A (en) | Road target detection method and device, electronic equipment and storage medium | |
CN111814794B (en) | Text detection method and device, electronic equipment and storage medium | |
WO2022105125A1 (en) | Image segmentation method and apparatus, computer device, and storage medium | |
CN109961107B (en) | Training method and device for target detection model, electronic equipment and storage medium | |
CN105930402A (en) | Convolutional neural network based video retrieval method and system | |
CN111160140B (en) | Image detection method and device | |
WO2020143316A1 (en) | Certificate image extraction method and terminal device | |
CN110852327A (en) | Image processing method, image processing device, electronic equipment and storage medium | |
CN111179270A (en) | Image co-segmentation method and device based on attention mechanism | |
CN111783797A (en) | Target detection method, device and storage medium | |
CN113763392B (en) | Model prediction method and system for road surface flatness detection and intelligent terminal | |
WO2024011859A1 (en) | Neural network-based face detection method and device | |
CN111914809A (en) | Target object positioning method, image processing method, device and computer equipment | |
CN113205102B (en) | Vehicle mark identification method based on memristor neural network | |
CN115170662A (en) | Multi-target positioning method based on yolov3 and convolutional neural network | |
Fu et al. | Pedestrian detection by feature selected self-similarity features | |
CN114565639A (en) | Target tracking method and system based on composite convolutional network | |
CN111598906A (en) | Vehicle detection method, system, device and storage medium | |
CN116071625B (en) | Training method of deep learning model, target detection method and device | |
TWI766560B (en) | Object recognition and ranging system using image semantic segmentation and lidar point cloud | |
CN112862719B (en) | Laser radar point cloud cell feature enhancement method based on graph convolution | |
CN117292304B (en) | Multimedia data transmission control method and system | |
CN109949335A (en) | A kind of image processing method and device | |
CN112949656B (en) | Underwater terrain matching positioning method, device and computer storage medium | |
CN116958749A (en) | Weak and small target detection model optimization method and device based on self-adaptive loss function |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |