A traffic satellite monitoring method and system based on artificial intelligence

Abstract

The present invention relates to the field of traffic monitoring technology, and in particular to a traffic satellite monitoring method and system based on artificial intelligence. First, an image filling network and a vehicle detection network are trained to obtain a trained image filling network and a trained vehicle detection network; then, the time position sequence distribution data of historical traffic vehicles are obtained based on a historical remote sensing image data set; finally, based on real-time satellite remote sensing images, real-time vehicle distribution data is obtained, and the vehicle distribution data at the next moment is predicted based on the real-time vehicle distribution data and the time position sequence distribution data of historical traffic vehicles. The present invention is based on feature extraction and deep learning, combined with real-time satellite remote sensing images to predict traffic distribution, has no usage restrictions, can be applied to any scene, and provides a good foundation for avoiding traffic congestion.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic monitoring, and in particular to a traffic satellite monitoring method and system based on artificial intelligence. Background Art

[0002] In recent years, with the continuous acceleration of my country's urbanization process and the gradual improvement of urbanization level, the contradiction between the growing population and limited land resources has led to a series of urban traffic problems, such as traffic congestion and exhaust pollution. In terms of traffic distribution, the problem of unbalanced distribution of traffic demand has become increasingly prominent, and the problem of urban congestion has become more serious. Simply relying on traffic management solutions and urban road network expansion cannot fundamentally solve the problem of unbalanced demand distribution. For this reason, it is necessary to monitor the traffic status to predict traffic distribution, so as to avoid traffic congestion as much as possible.

[0003] Although the current traffic distribution prediction methods have been widely used in the engineering field, they still have the following shortcomings: ① The existing traffic distribution prediction methods only model based on the single characteristics of traffic demand. However, as land use develops in the direction of integration and combination, traffic demand also presents the characteristics of diverse composition and spatiotemporal differences. The current traffic distribution prediction methods cannot realize refined traffic distribution prediction for multiple traffic demands; ② The existing traffic distribution prediction methods mainly focus on urban transportation systems, but there are few traffic distribution prediction methods for comprehensive transportation systems, and most of them directly use the traffic distribution prediction methods of urban transportation systems. In addition, the traffic travel of the comprehensive transportation system has the characteristics of large spatial span and diverse mode selection, and the traffic demand is more diverse, which leads to the limitations of the research results of the existing traffic distribution prediction methods in the direction of comprehensive transportation systems. Summary of the invention

[0004] In view of this, the present invention provides a traffic satellite monitoring method and system based on artificial intelligence, which predicts traffic distribution based on feature extraction and deep learning, combined with real-time satellite remote sensing images. It has no usage restrictions and can be applied to any scenario, providing a good foundation for avoiding traffic congestion.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A traffic satellite monitoring method based on artificial intelligence, comprising:

[0007] Acquire a historical remote sensing image data set, and preprocess the historical remote sensing image data set to obtain a historical preprocessed data set;

[0008] Training the image filling network based on the historical preprocessed data set to obtain the trained image filling network;

[0009] Based on the preprocessed data set and the trained image filling network, a filling data set is obtained, and based on the filling data set, a vehicle detection network is trained to obtain the trained vehicle detection network;

[0010] Obtaining historical vehicle recognition results based on the filled data set and the trained vehicle detection network, and obtaining vehicle time position series distribution data based on the historical vehicle recognition results;

[0011] Acquire real-time satellite remote sensing images, and obtain real-time vehicle distribution data based on the real-time satellite remote sensing images, the trained image filling network and the trained vehicle detection network;

[0012] The vehicle distribution data at the next moment is predicted based on the real-time vehicle distribution data and the vehicle time position series distribution data.

[0013] Preferably, the acquiring of the historical remote sensing image dataset and preprocessing the historical remote sensing image dataset to obtain the historical preprocessed dataset includes:

[0014] Obtain historical remote sensing image datasets;

[0015] Performing color interpolation on the historical remote sensing image dataset to obtain an interpolated dataset;

[0016] Performing color correction on the interpolation data set to obtain a corrected data set;

[0017] The correction data set is filtered to obtain the historical preprocessing data set.

[0018] Preferably, the image filling network is a generative adversarial network, which fills the remote sensing image to remove the occlusion of the vehicle in the remote sensing image.

[0019] Preferably, the acquiring of real-time satellite remote sensing images, and obtaining real-time vehicle distribution data based on the real-time satellite remote sensing images, the trained image filling network and the trained vehicle detection network, includes:

[0020] Acquiring the real-time satellite remote sensing image;

[0021] Preprocessing the real-time satellite remote sensing image to obtain a real-time preprocessed image;

[0022] Based on the real-time preprocessed image and the trained image filling network, a real-time filling image is obtained;

[0023] Based on the real-time filled image and the trained vehicle detection network, the real-time vehicle distribution data is obtained.

[0024] The present invention also provides a traffic satellite monitoring system based on artificial intelligence, comprising:

[0025] A data acquisition module is used to acquire a historical remote sensing image data set and preprocess the historical remote sensing image data set to obtain a historical preprocessed data set;

[0026] A first training module, used for training the image filling network based on the historical preprocessing data set to obtain the trained image filling network;

[0027] A second training module is used to fill the network with images based on the preprocessed data set and the trained image to obtain a filled data set, and train the vehicle detection network based on the filled data set to obtain the trained vehicle detection network;

[0028] A historical distribution module, used to obtain historical vehicle recognition results based on the filled data set and the trained vehicle detection network, and to obtain vehicle time position series distribution data based on the historical vehicle recognition results;

[0029] A real-time distribution module, used to obtain real-time satellite remote sensing images, and obtain real-time vehicle distribution data based on the real-time satellite remote sensing images, the trained image filling network and the trained vehicle detection network;

[0030] The distribution prediction module is used to predict the vehicle distribution data at the next moment based on the real-time vehicle distribution data and the vehicle time position series distribution data.

[0031] Preferably, the data acquisition module includes:

[0032] A data acquisition unit, used to acquire historical remote sensing image data sets;

[0033] A color interpolation unit, used for performing color interpolation on the historical remote sensing image data set to obtain an interpolated data set;

[0034] A color correction unit, used for performing color correction on the interpolation data set to obtain a corrected data set;

[0035] A filtering unit is used to filter the correction data set to obtain the historical preprocessing data set.

[0036] Preferably, the image filling network is a generative adversarial network, which fills the remote sensing image to remove the occlusion of the vehicle in the remote sensing image.

[0037] Preferably, the real-time distribution module includes:

[0038] A data acquisition unit, used to acquire the real-time satellite remote sensing image;

[0039] A preprocessing unit, used for preprocessing the real-time satellite remote sensing image to obtain a real-time preprocessed image;

[0040] A real-time filling unit, used to obtain a real-time filled image based on the real-time pre-processed image and the trained image filling network;

[0041] A real-time distribution unit is used to obtain the real-time vehicle distribution data based on the real-time filling image and the trained vehicle detection network.

[0042] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0043] The present invention relates to a traffic satellite monitoring method and system based on artificial intelligence, the method comprising: obtaining a historical remote sensing image data set, and preprocessing the historical remote sensing image data set to obtain a historical preprocessing data set; training an image filling network based on the historical preprocessing data set to obtain the trained image filling network; obtaining a filling data set based on the preprocessing data set and the trained image filling network, and training a vehicle detection network based on the filling data set to obtain the trained vehicle detection network; obtaining a historical vehicle recognition result based on the filling data set and the trained vehicle detection network, and obtaining vehicle time position sequence distribution data based on the historical vehicle recognition result; obtaining real-time satellite remote sensing images, and obtaining real-time vehicle distribution data based on the real-time satellite remote sensing images, the trained image filling network and the trained vehicle detection network; predicting vehicle distribution data at the next moment based on the real-time vehicle distribution data and the vehicle time position sequence distribution data. The present invention predicts traffic distribution based on feature extraction and deep learning, combined with real-time satellite remote sensing images, has no usage restrictions, can be applied to any scene, and provides a good foundation for avoiding traffic congestion. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0045] Figure 1 This is a flow chart of the traffic satellite monitoring method based on artificial intelligence of the present invention;

[0046] Figure 2 This is a structural diagram of the traffic satellite monitoring system based on artificial intelligence of the present invention;

[0047] Figure 3 This is a comparison chart of the effects before and after removing occlusion by the image filling network of the present invention.

[0048] Explanation of symbols: 1. Data acquisition module; 2. First training module; 3. Second training module; 4. Historical distribution module; 5. Real-time distribution module; 6. Distribution prediction module. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] The purpose of the present invention is to provide a traffic satellite monitoring method and system based on artificial intelligence, which predicts traffic distribution based on feature extraction and deep learning, combined with real-time satellite remote sensing images. It has no usage restrictions and can be applied to any scenario, providing a good foundation for avoiding traffic congestion.

[0051] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0052] Figure 1 FIG. 1 is a flow chart of the traffic satellite monitoring method based on artificial intelligence of the present invention. Figure 1 As shown, the present invention provides a traffic satellite monitoring method based on artificial intelligence, comprising:

[0053] Step S1, obtaining a historical remote sensing image data set, and preprocessing the historical remote sensing image data set to obtain a historical preprocessed data set.

[0054] Specifically, the step S1 includes:

[0055] Step S11, obtaining a historical remote sensing image dataset.

[0056] Step S12, performing color interpolation on the historical remote sensing image dataset to obtain an interpolated dataset.

[0057] Step S13, performing color correction on the interpolation data set to obtain a corrected data set.

[0058] Step S14, filtering the correction data set to obtain the historical preprocessing data set.

[0059] Step S2, training the image filling network based on the historical preprocessing data set to obtain the trained image filling network.

[0060] In this embodiment, the image filling network is a generative adversarial network, which fills the remote sensing image to remove the occlusion caused by the vehicle in the remote sensing image. The present invention removes the obstruction caused by clouds and trees in the remote sensing image through the image filling network, which provides a good foundation for subsequent vehicle recognition. The effect before and after deocclusion is as follows: Figure 3 As shown, Figure 3 The middle rectangular frame is de-occluded to eliminate the occlusion caused by trees.

[0061] The specific principle of generating adversarial networks is as follows:

[0062] In the training data, paired correspondences between input conditions and output images are available. The generator takes an image y and a latent vector z as input and produces an output x; the discriminator takes a pair (x, y) as input and tries to distinguish fake generated pairs from the true distribution. Image filling can be viewed as a constrained image conditional generation problem.

[0063] Step S3, filling the network with images based on the preprocessed data set and the trained image to obtain a filling data set, and training the vehicle detection network based on the filling data set to obtain the trained vehicle detection network.

[0064] In this embodiment, the vehicle detection network uses an improved YOLOv3 convolutional neural network, replaces Darknet-53 with Darknet-33 as the backbone network to extract features, and generates a priori boxes on the constructed feature pyramid network. Then, BBox regression and multi-label classification loss calculation are performed on the prior boxes whose IOU values ​​between the real boxes and the prior boxes are greater than 0.5. Finally, the non-maximum suppression method is used to remove redundant detection boxes and output the optimal detection results.

[0065] Step S4, obtaining historical vehicle recognition results based on the filled data set and the trained vehicle detection network, and obtaining vehicle time position series distribution data based on the historical vehicle recognition results.

[0066] Step S5, acquiring real-time satellite remote sensing images, and obtaining real-time vehicle distribution data based on the real-time satellite remote sensing images, the trained image filling network and the trained vehicle detection network.

[0067] Furthermore, the step S5 comprises:

[0068] Step S51, acquiring the real-time satellite remote sensing image.

[0069] Step S52, preprocessing the real-time satellite remote sensing image to obtain a real-time preprocessed image. For the specific preprocessing process, refer to "step S12-step S14".

[0070] Step S53, obtaining a real-time filled image based on the real-time preprocessed image and the trained image filling network.

[0071] Step S54, obtaining the real-time vehicle distribution data based on the real-time filled image and the trained vehicle detection network.

[0072] Step S6: predicting the vehicle distribution data at the next moment based on the real-time vehicle distribution data and the vehicle time position series distribution data.

[0073] Assume that the total number of vehicles at location B at time A in the vehicle time position series distribution data is 50, and the total number of vehicles at location B at time A+1 in the vehicle time position series distribution data is 60.

[0074] Assuming that the time of the real-time vehicle distribution data is time A, and the total number of vehicles at location B in the real-time vehicle distribution data is 60, then the total number of vehicles at location B at time A+1 corresponding to the real-time vehicle distribution data is 60*60÷50=70.

[0075] Figure 2 Figure 1 is a structural diagram of the traffic satellite monitoring system based on artificial intelligence of the present invention. Figure 2 As shown, the present invention provides a traffic satellite monitoring system based on artificial intelligence, including: a data acquisition module 1, a first training module 2, a second training module 3, a historical distribution module 4, a real-time distribution module 5 and a distribution prediction module 6.

[0076] The data acquisition module 1 is used to acquire a historical remote sensing image data set, and preprocess the historical remote sensing image data set to obtain a historical preprocessed data set.

[0077] The first training module 2 is used to train the image filling network based on the historical preprocessing data set to obtain the trained image filling network.

[0078] The second training module 3 is used to fill the network based on the preprocessed data set and the trained image to obtain a filled data set, and train the vehicle detection network based on the filled data set to obtain the trained vehicle detection network.

[0079] The historical distribution module 4 is used to obtain historical vehicle recognition results based on the filled data set and the trained vehicle detection network, and to obtain vehicle time position series distribution data based on the historical vehicle recognition results.

[0080] The real-time distribution module 5 is used to obtain real-time satellite remote sensing images, and obtain real-time vehicle distribution data based on the real-time satellite remote sensing images, the trained image filling network and the trained vehicle detection network.

[0081] The distribution prediction module 6 is used to predict the vehicle distribution data at the next moment based on the real-time vehicle distribution data and the vehicle time position series distribution data.

[0082] Optionally, the data acquisition module 1 includes:

[0083] The data acquisition unit is used to acquire historical remote sensing image data sets.

[0084] The color interpolation unit is used to perform color interpolation on the historical remote sensing image dataset to obtain an interpolated dataset.

[0085] The color correction unit is used to perform color correction on the interpolation data set to obtain a corrected data set.

[0086] A filtering unit is used to filter the correction data set to obtain the historical preprocessing data set.

[0087] Optionally, the image filling network is a generative adversarial network, which fills the remote sensing image to remove occlusion of the vehicle in the remote sensing image.

[0088] Optionally, the real-time distribution module 5 includes:

[0089] The data acquisition unit is used to obtain the real-time satellite remote sensing image.

[0090] The preprocessing unit is used to preprocess the real-time satellite remote sensing image to obtain a real-time preprocessed image.

[0091] A real-time filling unit is used to obtain a real-time filled image based on the real-time pre-processed image and the trained image filling network.

[0092] A real-time distribution unit is used to obtain the real-time vehicle distribution data based on the real-time filling image and the trained vehicle detection network.

[0093] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0094] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A traffic satellite monitoring method based on artificial intelligence, It is characterized in that include: Acquiring a historical remote sensing image dataset and preprocessing the historical remote sensing image dataset to obtain a historical preprocessed dataset, specifically including: Obtain historical remote sensing imagery datasets; Performing color interpolation on the historical remote sensing image dataset to obtain an interpolated dataset; Performing color correction on the interpolation data set to obtain a corrected data set; Filtering the correction data set to obtain the historical preprocessing data set; Training the image filling network based on the historical preprocessed data set to obtain the trained image filling network specifically includes: The image filling network is a generative adversarial network, which fills the remote sensing image to remove the occlusion caused to the vehicle in the remote sensing image; Based on the preprocessed data set and the trained image filling network, a filling data set is obtained, and based on the filling data set, a vehicle detection network is trained to obtain the trained vehicle detection network; Obtaining historical vehicle recognition results based on the filled data set and the trained vehicle detection network, and obtaining vehicle time position series distribution data based on the historical vehicle recognition results; Acquiring real-time satellite remote sensing images, and obtaining real-time vehicle distribution data based on the real-time satellite remote sensing images, the trained image filling network, and the trained vehicle detection network, specifically includes: Acquiring the real-time satellite remote sensing image; Preprocessing the real-time satellite remote sensing image to obtain a real-time preprocessed image; Based on the real-time preprocessed image and the trained image filling network, a real-time filling image is obtained; Based on the real-time filled image and the trained vehicle detection network, obtaining the real-time vehicle distribution data; The vehicle distribution data at the next moment is predicted based on the real-time vehicle distribution data and the vehicle time position series distribution data.

2. A traffic satellite monitoring system based on artificial intelligence, It is characterized in that include: The data acquisition module is used to acquire a historical remote sensing image data set and preprocess the historical remote sensing image data set to obtain a historical preprocessed data set, specifically including: A data acquisition unit, used to acquire historical remote sensing image data sets; A color interpolation unit, used for performing color interpolation on the historical remote sensing image data set to obtain an interpolated data set; A color correction unit, used for performing color correction on the interpolation data set to obtain a corrected data set; A filtering unit, used for filtering the correction data set to obtain the historical preprocessing data set; The first training module is used to train the image filling network based on the historical preprocessing data set to obtain the trained image filling network, specifically including: The image filling network is a generative adversarial network, which fills the remote sensing image to remove the occlusion caused to the vehicle in the remote sensing image; A second training module is used to fill the network with images based on the preprocessed data set and the trained image to obtain a filled data set, and train the vehicle detection network based on the filled data set to obtain the trained vehicle detection network; A historical distribution module, used to obtain historical vehicle recognition results based on the filled data set and the trained vehicle detection network, and to obtain vehicle time position series distribution data based on the historical vehicle recognition results; The real-time distribution module is used to obtain real-time satellite remote sensing images, and obtain real-time vehicle distribution data based on the real-time satellite remote sensing images, the trained image filling network and the trained vehicle detection network, specifically including: A data acquisition unit, used to acquire the real-time satellite remote sensing image; A preprocessing unit, used for preprocessing the real-time satellite remote sensing image to obtain a real-time preprocessed image; A real-time filling unit, used to obtain a real-time filled image based on the real-time pre-processed image and the trained image filling network; A real-time distribution unit, configured to obtain the real-time vehicle distribution data based on the real-time filled image and the trained vehicle detection network; The distribution prediction module is used to predict the vehicle distribution data at the next moment based on the real-time vehicle distribution data and the vehicle time position series distribution data.

Images