CN115100650A - Expressway abnormal scene denoising and identifying method and device based on multiple Gaussian models - Google Patents
Expressway abnormal scene denoising and identifying method and device based on multiple Gaussian models Download PDFInfo
- Publication number
- CN115100650A CN115100650A CN202210734673.4A CN202210734673A CN115100650A CN 115100650 A CN115100650 A CN 115100650A CN 202210734673 A CN202210734673 A CN 202210734673A CN 115100650 A CN115100650 A CN 115100650A
- Authority
- CN
- China
- Prior art keywords
- model
- scene
- denoising
- recognition result
- 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
- G06V20/00—Scenes; Scene-specific elements
- G06V20/70—Labelling scene content, e.g. deriving syntactic or semantic representations
-
- 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
- G06V10/30—Noise filtering
-
- 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
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/778—Active pattern-learning, e.g. online learning of image or video features
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method and a device for denoising and identifying an abnormal scene of a highway based on a multiple Gaussian model, wherein the method comprises the steps of firstly obtaining an image of the scene of the highway; then, a multiple Gaussian model is built according to preset parameters, the multiple Gaussian model comprises two mixed Gaussian models with different background learning rates, the two models are used for learning the background in the image, and a first recognition result image and a second recognition result image are obtained; and performing differentiation comparison on the first recognition result graph and the second recognition result graph to realize the recognition of the abnormal scene. The method can detect abnormal scene changes such as sprinkled objects and the like which affect traffic, simultaneously eliminate other moving objects, further remove noise points generated by foreground target detection due to frequent and fine jitter of the camera, and improve the accuracy and reliability of Gaussian mixture model detection.
Description
Technical Field
The invention relates to the technical field of scene recognition, in particular to a method and a device for denoising and recognizing an abnormal scene of a highway based on a multiple Gaussian model.
Background
In the prior art, the following methods are generally adopted to identify an abnormal scene:
1. abnormal scene changes are identified using depth recognition and convolutional neural network models. The method includes the steps of firstly collecting scene data as sample images, simultaneously generating a part of the sample images in an image superposition mode, then training a model by using the sample images, finally taking video images as input, and judging which areas in the whole image are abnormal through the model.
2. And identifying by using a Gaussian mixture model. The method comprises the steps of establishing a Gaussian mixture model for each pixel point in an image, and separating a moving target from a background by using the Gaussian mixture model; removing noise of the moving target image, strengthening the moving target image, and highlighting the object to be detected; and inputting the enhanced moving target image into a trained YOLO v3 target detection network according to frames, and determining whether the moving object is a vehicle, a pedestrian or other sprinklers.
The inventor of the present application finds that the method of the prior art has at least the following technical problems in the process of implementing the present invention:
the biggest problem in identifying abnormal scenes through a neural network is that samples are too few and the identification speed is slow. The expressway abnormal scene belongs to an accidental event, samples which can be used for training and verification are too few, corresponding training samples can be manually generated only in a mode of matting and image superposition, but the samples have the problem that the scene is not rich enough, for example, identified sprinkled objects are basically limited to common articles such as carton tires and the like; the low recognition speed is because the camera assumes that the camera is on the gantry, the high resolution can accurately recognize the abnormal scene on the lane, but the high resolution can reduce the operation speed of the algorithm after entering the convolutional neural network, occupies large calculation power and cannot give an alarm in time.
The single Gaussian mixture model can accurately separate the moving object in the scene from the background, but due to the characteristic that the background of the Gaussian model can be continuously learned, the object can be learned into the background by the Gaussian model after being completely static, compared with the moving object, abnormal scenes staying on a highway are more consistent with the definition and higher in harm, and under the condition that only the moving object is detected, abnormal scenes such as moving vehicles, sprinklers and the like cannot be directly distinguished, and a neural network is further required to be further utilized.
Therefore, the method in the prior art has the technical problems of low identification accuracy and reliability.
Disclosure of Invention
The invention provides a method and a device for denoising and identifying an abnormal scene of a highway based on a multiple Gaussian model, which are used for solving or at least partially solving the technical problems of low identification accuracy and reliability in the prior art.
In order to solve the above technical problems, a first aspect of the present invention provides a highway abnormal scene denoising and identifying method based on a multiple gaussian model, including:
acquiring an image of a highway scene;
constructing a multiple Gaussian model according to preset parameters, wherein the multiple Gaussian model comprises two Gaussian mixture models with different background learning rates, the Gaussian mixture model with the faster background learning rate is a model S, the Gaussian mixture model with the slower background learning rate is a model L, and the model S and the model L are used for learning the background in the image to obtain a first recognition result graph of the model S and a second recognition result graph of the model L;
and carrying out differentiation comparison on the first recognition result image and the second recognition result image, and simultaneously realizing scene denoising and abnormal scene recognition.
In one embodiment, the predetermined parameters include VarThreshold, lerningrate, and backgroudratio, where VarThreshold is a threshold for determining variance, lerningrate is a learning rate, and backgroudratio is a background ratio.
In one embodiment, the value of VarThreshold for model S and model L is set to 36, the learngrate for model S is set to-1, the backgroundRatio is set to 0.5, the learngrate for model L is set to 0.001, and the backgroundRatio is adjusted to 0.01.
In one embodiment, the differential comparison between the first recognition result image and the second recognition result image is performed, and the scene denoising and the abnormal scene recognition are simultaneously realized, including:
making a difference between the first recognition result image and the second recognition result image, if the color of a certain pixel in the first recognition result image is the same as that of a pixel at the same position in the second recognition result image, indicating that the two models at the same position are both moving targets or both noise points, and if the colors of the certain pixel and the pixel at the same position are different, indicating that the pixel at the position is a static object appearing in the default background;
and removing noise points appearing at the same position by differentiating, filtering the moving target, and identifying the abnormal scene by taking the static object separated from the differentiated comparison result as an abnormal object.
In one embodiment, after the first recognition result map and the second recognition result map are compared differentially, and recognition of an abnormal scene is achieved, the method further includes:
and marking the position of the abnormal object and giving an alarm.
In one embodiment, after performing differential comparison on the first recognition result map and the second recognition result map, and simultaneously performing scene denoising and abnormal scene recognition, the method further includes:
denoising by adopting a morphological method.
Based on the same inventive concept, the second aspect of the present invention provides a device for denoising and identifying an abnormal scene of a highway based on a multiple gaussian model, comprising:
the image acquisition module is used for acquiring an image of a highway scene;
the model construction module is used for constructing a multiple Gaussian model according to preset parameters, the multiple Gaussian model comprises two Gaussian mixture models with different background learning rates, the Gaussian mixture model with the faster background learning rate is a model S, the Gaussian mixture model with the slower background learning rate is a model L, and the model S and the model L are used for learning the background in the image to obtain a first recognition result graph of the model S and a second recognition result graph of the model L;
and the abnormal scene identification module is used for carrying out differentiation comparison on the first identification result graph and the second identification result graph to realize identification of the abnormal scene.
Based on the same inventive concept, a third aspect of the present invention provides a computer-readable storage medium having stored thereon a computer program which, when executed, performs the method of the first aspect
Based on the same inventive concept, a fourth aspect of the present invention provides a computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method according to the first aspect when executing the program.
Compared with the prior art, the invention has the advantages and beneficial technical effects as follows:
the invention provides a highway abnormal scene denoising and identifying method based on multiple Gaussian models, which comprises the steps of firstly obtaining an image of a highway scene; then, a multiple Gaussian model is built according to preset parameters, the model S and the model L are used for learning the background in the image, and a first recognition result graph of the model S and a second recognition result graph of the model L are obtained; and performing differentiation comparison on the first recognition result graph and the second recognition result graph to realize the recognition of the abnormal scene. Because two Gaussian models S and model L in the multiple Gaussian models have different background learning rates, when an object stays in the model S, the object can be learned into a background quickly, and when the object stays in the model L, the object basically cannot be learned into the background, namely, for a moving object, the recognition results of the two models are the same, the moving object and the static object can be marked as a foreground at the same time, and the learning difference can occur, so that after the recognition result graphs of the two models are subjected to difference comparison, the same part (moving object) can be filtered, and different parts (static object) can be highlighted, so that the accurate recognition of an abnormal object can be realized, and the accuracy and reliability of the recognition are improved. Meanwhile, the influence of noise points with the same mode appearing in the two Gaussian models can be greatly reduced through a differentiation comparison mode, and the recognition effect is further improved.
Drawings
In order to more clearly illustrate the embodiments or technical solutions of the present invention, the drawings used in the embodiments or technical solutions in the prior art are briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a flow chart of an implementation of a method for denoising and identifying an abnormal scene of a highway based on a multiple Gaussian model according to an embodiment of the present invention;
FIG. 2 is a diagram illustrating foreground noise in an embodiment of the present invention;
FIG. 3 is a schematic diagram of morphological processing of foreground noise in an embodiment of the present invention;
FIG. 4 is a diagram illustrating noise generated by the recognition result of the model S according to an embodiment of the present invention;
FIG. 5 is a diagram illustrating the noise generated by the recognition result of the model L according to the embodiment of the present invention;
FIG. 6 is a schematic diagram illustrating the filtered noise points according to the embodiment of the present invention;
FIG. 7 is a schematic diagram of a noise result after morphological denoising again based on FIG. 6 in the embodiment of the present invention;
FIG. 8 is a schematic diagram of a model S identifying a vehicle drop projectile in an embodiment of the present invention;
FIG. 9 is a schematic illustration of a model L identifying a vehicle drop projectile in an embodiment of the present invention;
FIG. 10 is a schematic diagram of the embodiment of the present invention in which the projectile is separated in a denoised environment.
Detailed Description
The inventor of the application finds out through a great deal of research and practice that:
in the prior art, a Gaussian mixture model identification method is used for establishing a Gaussian mixture model for each pixel point in an image, and the Gaussian mixture model is used for separating a moving target from a background; removing noise of the moving target image, strengthening the moving target image, and highlighting the object to be detected; and inputting the enhanced moving target image into a trained YOLO v3 target detection network according to frames, determining whether a moving object is a vehicle, a pedestrian or other sprinklers, and processing noise points. Most of images processed by the Gaussian mixture model have noise points with different degrees, and the foreground images are subjected to denoising processing through morphological filtering; carrying out corrosion and expansion treatment on the denoised image in sequence by using open operation; and finally, smoothing the image.
Based on the above consideration, the invention provides a method for detecting an abnormal scene of a highway by using multiple Gaussian models, which can detect abnormal scene changes affecting traffic such as spilled objects and the like, simultaneously eliminate other moving objects, further perform denoising processing on noise points generated by foreground target detection due to frequent and fine jitter of a camera, and improve the accuracy and reliability of mixed Gaussian model detection.
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example one
The embodiment of the invention provides a method for denoising and identifying an abnormal scene of a highway based on a multiple Gaussian model, which comprises the following steps:
acquiring an image of a highway scene;
constructing a multiple Gaussian model according to preset parameters, wherein the multiple Gaussian model comprises two Gaussian mixture models with different background learning rates, the Gaussian mixture model with the faster background learning rate is a model S, the Gaussian mixture model with the slower background learning rate is a model L, and the model S and the model L are used for learning the background in the image to obtain a first recognition result graph of the model S and a second recognition result graph of the model L;
and carrying out differentiation comparison on the first recognition result image and the second recognition result image, and simultaneously realizing scene denoising and abnormal scene recognition.
Specifically, the detection of foreground and background by a gaussian mixture model is a technique often used for extracting a background, which can detect a moving object of a foreground, and selects a proper amount of gaussian distribution for each pixel, and if the difference between a pixel point and the mean value of any gaussian model is greater than a standard deviation of a certain multiple, the pixel point can be judged as the foreground, and the characteristic of continuous learning can better adapt to illumination changes and the like of different scenes.
The application of the existing Gaussian model is that a background of a mixed Gaussian model is constructed, a foreground, namely a moving object in a scene, is detected and separated, then the moving object is identified and classified, and the remaining moving object except for targets such as vehicles and the like can be classified into abnormal objects which do not exist in a highway. The judgment is simple and easy to use, and the Gaussian mixture model is also widely applied as one of the means for foreground detection, but the method has a plurality of limitations:
firstly, the gaussian mixture model is a model which is constructed according to the values of each pixel at different time points, so that the resistance to high-frequency movement is insufficient, because the application scene provided by the invention is an expressway scene, and the camera is fixed on a portal frame, when a large vehicle passes by the portal frame in windy weather, the camera can generate high-frequency vibration of different degrees, the vibration is not obvious to naked eyes, and the observation of an event can not be influenced, but for the gaussian mixture model, the whole screen moves rapidly in a short time, so that a large number of noise points can be directly generated, as shown in fig. 2, the moving vehicle is a foreground and is marked as a clear white area, but leaves and object edges can be displayed on the foreground due to the shake of the camera in windy weather.
The result of using only the conventional morphological denoising method is shown in fig. 3, the effect is not obvious, and the edge condition cannot be processed.
Therefore, in addition to the traditional denoising method, further denoising processing is required, otherwise, the recognition result cannot be used as a subsequent reference at all, and the whole screen is covered by noise (i.e., the recognition accuracy and reliability are not high).
Secondly, regarding the definition problem of the abnormal scene, the inventor of the present application finds out through a great deal of research and practice that: moving objects on the highway except vehicles should not be directly recognized as abnormal scenes such as sprinklers or pedestrians, but the abnormal scenes which affect the vehicles should be static obstacles which do not belong to the background but stay in the area of the road surface, so that the moving objects should be tracked and alarmed rather than recognized only after the obstacles are static, so that the moving objects are far from enough tracked only by a single Gaussian model background, the objects (static objects) staying in the video are required to be not learned as the background and be highlighted, and the moving objects (vehicles and the like) are excluded.
In one embodiment, the predetermined parameters include VarThreshold, lerningrate, and BackgroundRatio, where VarThreshold is a threshold for determining variance, lerningrate is a learning rate, and BackgroundRatio is a background ratio.
In one embodiment, the value of VarThreshold for model S and model L is set to 36, the learngrate for model S is set to-1, the backgroundRatio is set to 0.5, the learngrate for model L is set to 0.001, and the backgroundRatio is adjusted to 0.01.
Specifically, the method of the multiple Gaussian model is used for filtering moving objects on the basis of identifying abnormal scenes, and meanwhile, noise reduction processing is further performed on noise generated by the camera. The method firstly needs to construct two Gaussian mixture models with different background learning rates, the Gaussian mixture model with the faster background learning rate is called a model S, and the Gaussian mixture model with the slower background learning rate is called a model L.
In general, a Gaussian mixture model has the following important parameters that affect the performance of the model: leringmate: the learning rate is between 0 and 1, the closer to 1 represents that a new object is more easily learned as a background, and the closer to 0 represents that the background is more stable and less easily learned as the background; history: the number of frames for constructing the background model is 500 as a default value; BackgroundRatio: background proportion, which is learned as background and added to the model if the value of a foreground pixel stabilizes background ratio history frames; NMixtures: the number of models is used for constructing the number of Gaussian models of the background; VarThreshold: the threshold value of the variance is determined, and can be properly adjusted to be high under a scene with obvious light change (such as a reflective water surface), and the default value is 16.
By repeatedly adjusting and testing the performance of these parameters in an actual scene, embodiments of the present invention determine the parameter values needed in a highway detection scenario for a spill: firstly, VarThreshold is used, and for a model S and a model L, the value needs to be properly adjusted to 36 so as to adapt to sufficient illumination in most cases of a highway, and obvious light reflection can be generated on smooth surfaces such as vehicle windows and vehicle roofs; then, regarding the background learning speed of the model, two values of the learngrate and the backsgroupratio need to be adjusted, the learngrate is adjusted to be-1 and the backsgroupratio is adjusted to be 0.5 for the model S, the learngrate is adjusted to be 1e-3 and the backsgroupratio is adjusted to be 1e-2 for the model L, and the settings can ensure that the model S and the model L have similar detection results for short-term high-frequency camera shake and can separate out objects existing in the image for a long time.
In one embodiment, the differential comparison of the first recognition result image and the second recognition result image is performed, and the scene denoising and the abnormal scene recognition are simultaneously realized, and the method comprises the following steps:
making a difference between the first recognition result image and the second recognition result image, if the color of a certain pixel in the first recognition result image is the same as that of a pixel at the same position in the second recognition result image, indicating that the two models at the same position are both moving targets or both noise points, and if the colors of the certain pixel and the pixel at the same position are different, indicating that the pixel at the position is a static object appearing in the default background;
and removing noise points appearing at the same position by differentiating, filtering the moving target, and taking the separated static object in the differentiated comparison result as an abnormal object, thereby simultaneously realizing scene denoising and abnormal scene identification.
Specifically, although different gaussian mixture models are different in learning rate of the background, according to basic properties of the gaussian mixture models, the gaussian mixture models have almost the same reflection effect on a rapidly changing scene, namely, under the condition that a camera has high-speed vibration, noise points with the same mode can appear in a plurality of gaussian mixture models, under the condition that the noise points are the same, the influence of the noise points can be greatly reduced through a basic differentiation comparison mode, the noise points of a large piece of a full screen are weakened to fine noise points, and then the traditional morphological mode corrosion and other denoising methods are adopted, so that the noise point filtering with a better denoising effect and stronger adaptability can be achieved.
The position is the same corresponding position in the two models, and the default background refers to the background obtained by learning the models S and L.
Please refer to fig. 1, which is a flowchart illustrating an implementation of a method for denoising and identifying an abnormal scene of a highway based on a multiple gaussian model according to an embodiment of the present invention.
The following description is made by way of a specific scenario, and specifically refers to fig. 4-7.
A great amount of noise which cannot be eliminated by the traditional method appears in both the model S and the model L in the same scene, and specifically, see fig. 4 and 5, which are recognition result diagrams of the model S and the model L, respectively.
After the difference comparison, noise due to camera movement is filtered, as shown in fig. 6. Then, the conventional morphological denoising is performed to obtain a result picture with a small number of noise points, as shown in fig. 7.
When foreground denoising is completed through differentiation comparison, abnormal scenes can be identified and detected through differentiation comparison. This embodiment is illustrated by a large projectile dropped by a vehicle, and as mentioned above, the two gaussian models (S and L) have different background learning rates, so that the object is learned to the background faster when staying in the model S (fig. 8), and is not learned to the background substantially when staying in the model L (fig. 9). Therefore, the model S and the model L are differentiated and compared, the moving target is labeled as a foreground, and the stationary object has a learning difference, so that after the differentiation and comparison, the moving target is filtered out, and the stationary object is displayed, as shown in fig. 10.
It should be noted that, in the drawings related to the present invention, the background portion (for example, the black portion of all the images in fig. 2-10) is a normal highway scene, i.e., a scene without any abnormality when the program starts to operate; then moving objects, namely vehicles, trees and the like, can appear in the background, and normal objects in scenes which do not have the influence on driving (as shown in 2); however, if there is an object that does not belong to the background and is still in the scene, such as a vehicle temporarily stops, a vehicle throws, a rockfall, a landslide, etc., the object that does not belong to the normal background is added in the scene, that is, the object is an abnormal object, that is, there is an abnormal situation different from the normal scene in the scene (as shown in fig. 10).
The method can realize the identification of the abnormal scenes, namely, the work of identifying the abnormal scenes by using a multiple Gaussian model is completed, and simultaneously, the image with a large amount of motion noise is subjected to denoising treatment, so that the identification result is more accurate and reliable.
In one embodiment, after performing differential comparison on the first recognition result map and the second recognition result map, and simultaneously performing scene denoising and abnormal scene recognition, the method further includes:
and marking the position of the abnormal object and giving an alarm.
In one embodiment, after the first recognition result map and the second recognition result map are compared differentially to realize the recognition of the abnormal scene, the method further includes:
and denoising by adopting a morphological method.
Specifically, in order to further improve the recognition effect, the invention further adopts a morphological method to denoise on the basis of the previous recognition.
For the abnormal scene identification scheme, a Gaussian model algorithm for extracting background separation foreground by using a traditional method is available, but only moving objects are detected, and the influence of interference factors needs to be eliminated by combining other methods subsequently. There are also methods that mainly use neural network recognition, which require a large amount of collected data samples to train the model, and also require preprocessing including road segmentation. For a video denoising scheme, the conventional method uses information of adjacent frames, performs fusion processing on similar pixel blocks in the adjacent frames, and then denoises in a morphological way, but the method has a general effect on the condition that a camera is greatly jittered.
Generally speaking, the abnormal scene identification method provided by the invention has higher accuracy and reliability in the aspect of abnormal scene detection compared with other traditional schemes (methods of mixing Gaussian models), can greatly improve the identification efficiency compared with a neural network algorithm, and does not need to collect a large amount of data. The model of the invention can synchronously solve the noise problem caused by camera shake, can save the computational resources of the server and can achieve better denoising effect.
Example two
Based on the same inventive concept, the embodiment provides a device for denoising and identifying an abnormal scene of a highway based on a multiple gaussian model, which comprises:
the image acquisition module is used for acquiring an image of a highway scene;
the model construction module is used for constructing a multiple Gaussian model according to preset parameters, the multiple Gaussian model comprises two Gaussian mixture models with different background learning rates, the Gaussian mixture model with the faster background learning rate is a model S, the Gaussian mixture model with the slower background learning rate is a model L, and the model S and the model L are used for learning the background in the image to obtain a first recognition result graph of the model S and a second recognition result graph of the model L;
and the abnormal scene identification module is used for carrying out differentiation comparison on the first identification result image and the second identification result image and simultaneously realizing scene denoising and abnormal scene identification.
Since the device introduced in the fourth embodiment of the present invention is a device used for implementing the method for denoising and identifying the abnormal scene of the highway based on the multiple gaussian model in the first embodiment of the present invention, a person skilled in the art can understand the specific structure and deformation of the device based on the method introduced in the first embodiment of the present invention, and thus details are not described herein. All the devices adopted in the method in the first embodiment of the invention belong to the protection scope of the invention.
EXAMPLE III
Based on the same inventive concept, the present invention also provides a computer-readable storage medium, on which a computer program is stored, which when executed performs the method as described in the first embodiment.
Since the computer-readable storage medium introduced in the third embodiment of the present invention is a computer-readable storage medium used for implementing the method for denoising and identifying an abnormal scene of a highway based on a multiple gaussian model in the first embodiment of the present invention, based on the method introduced in the first embodiment of the present invention, a person skilled in the art can understand the specific structure and deformation of the computer-readable storage medium, and thus, details are not described here. Any computer readable storage medium used in the method of the first embodiment of the present invention is within the protection scope of the present invention.
Example four
Based on the same inventive concept, the application also provides a computer device comprising a memory. A processor and a computer program stored on the memory and operable on the processor, the processor implementing the method of the first embodiment when executing the program.
Since the computer device introduced in the fourth embodiment of the present invention is a computer device used for implementing the method for denoising and identifying an abnormal scene of a highway based on a multiple gaussian model in the first embodiment of the present invention, based on the method introduced in the first embodiment of the present invention, a person skilled in the art can understand the specific structure and deformation of the computer device, and thus, details are not described herein. All the computer devices used in the method in the first embodiment of the present invention are within the scope of the present invention.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
While preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all such alterations and modifications as fall within the scope of the invention.
It will be apparent to those skilled in the art that various modifications and variations can be made in the embodiments of the present invention without departing from the spirit or scope of the embodiments of the invention. Thus, if such modifications and variations of the embodiments of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to encompass such modifications and variations.
Claims (9)
1. The highway abnormal scene denoising and identifying method based on the multiple Gaussian models is characterized by comprising the following steps:
acquiring an image of a highway scene;
constructing a multiple Gaussian model according to preset parameters, wherein the multiple Gaussian model comprises two Gaussian mixture models with different background learning rates, the Gaussian mixture model with the faster background learning rate is a model S, the Gaussian mixture model with the slower background learning rate is a model L, and the model S and the model L are used for learning the background in the image to obtain a first recognition result graph of the model S and a second recognition result graph of the model L;
and carrying out differentiation comparison on the first recognition result image and the second recognition result image, and simultaneously realizing scene denoising and abnormal scene recognition.
2. The method for denoising and identifying the abnormal scene of the expressway according to claim 1, wherein the preset parameters comprise VarThreshold, learngrate and backgroudratio, wherein VarThreshold is a threshold for determining variance, learngrate is a learning rate, and backgroudratio is a background proportion.
3. The method for denoising and identifying the abnormal scene of the expressway based on the multiple Gaussian models as claimed in claim 2, wherein the value of VarThreshold of the model S and the model L is set to 36, the learngrate of the model S is set to-1, the backgroudRate is set to 0.5, the learngrate of the model L is set to 0.001, and the backgroudRate is adjusted to 0.01.
4. The method for denoising and identifying the abnormal scene of the highway based on the multiple Gaussian models as claimed in claim 1, wherein the step of comparing the first identification result graph and the second identification result graph in a differentiation manner and simultaneously realizing scene denoising and abnormal scene identification comprises the steps of:
making a difference between the first recognition result image and the second recognition result image, if the color of a certain pixel in the first recognition result image is the same as that of a pixel at the same position in the second recognition result image, indicating that the two models at the same position are both moving targets or both noise points, and if the colors of the certain pixel and the pixel at the same position are different, indicating that the pixel at the position is a static object appearing in the default background;
and removing noise points appearing at the same position by differentiating, filtering the moving target, and identifying the abnormal scene by taking the static object separated from the differentiated comparison result as an abnormal object.
5. The method for denoising and identifying the abnormal scene of the highway based on the multiple gaussian models as claimed in claim 1, wherein the first identification result graph and the second identification result graph are compared differentially, and after the abnormal scene is identified, the method further comprises:
and marking the position of the abnormal object and giving an alarm.
6. The method for denoising and identifying the abnormal scene of the highway based on the multiple gaussian models as claimed in claim 1, wherein after the first identification result graph and the second identification result graph are compared differentially, and the scene denoising and the abnormal scene identification are realized simultaneously, the method further comprises:
and denoising by adopting a morphological method.
7. Highway abnormal scene denoising and recognition device based on multiple Gaussian models is characterized by comprising the following steps:
the image acquisition module is used for acquiring an image of a highway scene;
the model construction module is used for constructing a multiple Gaussian model according to preset parameters, the multiple Gaussian model comprises two Gaussian mixture models with different background learning rates, the Gaussian mixture model with the faster background learning rate is a model S, the Gaussian mixture model with the slower background learning rate is a model L, and the model S and the model L are used for learning the background in the image to obtain a first recognition result graph of the model S and a second recognition result graph of the model L;
and the abnormal scene identification module is used for carrying out differentiation comparison on the first identification result graph and the second identification result graph to realize identification of the abnormal scene.
8. A computer-readable storage medium, on which a computer program is stored, characterized in that the program, when executed, implements the method of any one of claims 1 to 6.
9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the method according to any of claims 1 to 6 when executing the program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210734673.4A CN115100650A (en) | 2022-06-27 | 2022-06-27 | Expressway abnormal scene denoising and identifying method and device based on multiple Gaussian models |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210734673.4A CN115100650A (en) | 2022-06-27 | 2022-06-27 | Expressway abnormal scene denoising and identifying method and device based on multiple Gaussian models |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115100650A true CN115100650A (en) | 2022-09-23 |
Family
ID=83292486
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210734673.4A Pending CN115100650A (en) | 2022-06-27 | 2022-06-27 | Expressway abnormal scene denoising and identifying method and device based on multiple Gaussian models |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115100650A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116259028A (en) * | 2023-05-06 | 2023-06-13 | 杭州宏景智驾科技有限公司 | Abnormal scene detection method for laser radar, electronic device and storage medium |
-
2022
- 2022-06-27 CN CN202210734673.4A patent/CN115100650A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116259028A (en) * | 2023-05-06 | 2023-06-13 | 杭州宏景智驾科技有限公司 | Abnormal scene detection method for laser radar, electronic device and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110178167B (en) | Intersection violation video identification method based on cooperative relay of cameras | |
| Alefs et al. | Road sign detection from edge orientation histograms | |
| CN100495438C (en) | Method for detecting and identifying moving target based on video monitoring | |
| US20180181817A1 (en) | Vehicular lane line data processing method, apparatus, storage medium, and device | |
| CN102479327B (en) | Target identification system, and the obstacle identification system and method for automobile | |
| CN107180230B (en) | Universal license plate recognition method | |
| CN103093198A (en) | Crowd density monitoring method and device | |
| KR101224027B1 (en) | Method for dectecting front vehicle using scene information of image | |
| CN107704833A (en) | A kind of front vehicles detection and tracking based on machine learning | |
| CN101369312B (en) | Method and equipment for detecting intersection in image | |
| Su et al. | A new local-main-gradient-orientation HOG and contour differences based algorithm for object classification | |
| CN110276318A (en) | Nighttime road rains recognition methods, device, computer equipment and storage medium | |
| CN115376108A (en) | Obstacle detection method and device in complex weather | |
| Ghahremannezhad et al. | Automatic road detection in traffic videos | |
| Helala et al. | Road boundary detection in challenging scenarios | |
| CN115100650A (en) | Expressway abnormal scene denoising and identifying method and device based on multiple Gaussian models | |
| Danilescu et al. | Road anomalies detection using basic morphological algorithms | |
| CN110782485A (en) | Vehicle lane change detection method and device | |
| CN113221603A (en) | Method and device for detecting shielding of monitoring equipment by foreign matters | |
| CN108288041B (en) | Preprocessing method for removing false detection of pedestrian target | |
| CN105069410A (en) | Unstructured road recognition method and device | |
| CN113822149A (en) | Emergency lane visual detection method and system based on view angle of unmanned aerial vehicle | |
| CN103400135A (en) | Video signal pre-processing method for clearly detecting traffic accidents under severe weather condition | |
| CN114913488A (en) | Sprinkler detection method, device, electronic device, and storage medium | |
| Rao et al. | Convolutional Neural Network Model for Traffic Sign Recognition |
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 |