CN114283343A

CN114283343A - Map updating method, training method and equipment based on remote sensing satellite image

Info

Publication number: CN114283343A
Application number: CN202111567400.7A
Authority: CN
Inventors: 王紫玉; 吴彬; 钟开; 杨建忠; 张通滨; 卢振
Original assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Current assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Priority date: 2021-12-20
Filing date: 2021-12-20
Publication date: 2022-04-05
Anticipated expiration: 2041-12-20
Also published as: CN114283343B

Abstract

The invention provides a map updating method, a map training method and map updating equipment based on a remote sensing satellite image, relates to the field of data processing, and particularly relates to the technical field of road networks. The specific implementation scheme is as follows: acquiring a remote sensing satellite image; performing semantic segmentation processing on the remote sensing satellite image to obtain a first topological graph, wherein the first topological graph comprises a road at a position corresponding to the remote sensing satellite image; determining a three-dimensional tensor map according to the remote sensing satellite image, wherein the three-dimensional tensor map represents the coding information of the road at the position corresponding to the remote sensing satellite image; decoding the three-dimensional tensor image to obtain a second topological image, wherein the second topological image comprises a road at a position corresponding to the remote sensing satellite image; and determining a road topological graph according to the first topological graph and the second topological graph, and updating the map according to the road topological graph.

Description

Map updating method, training method and equipment based on remote sensing satellite image

Technical Field

The disclosure relates to the technical field of road networks in the field of data processing, in particular to a map updating method, a map training method and a map updating device based on remote sensing satellite images.

Background

With the development of mobile internet and intelligent devices, maps have become an important basis for people going out. The roads in the road network change and the map needs to be updated.

In the prior art, data of roads can be manually collected according to collection equipment, and then a map is updated based on the data of the roads.

However, in the prior art, a large amount of manpower and material resources are required to be consumed for manually collecting the data of the road, so that the cost for updating the map is high; in addition, the above method has low operation efficiency and is easy to cause errors, which results in untimely map updating and map updating errors.

Disclosure of Invention

The disclosure provides a map updating method, a training method and equipment based on remote sensing satellite images.

According to a first aspect of the present disclosure, there is provided a map updating method based on a remote sensing satellite image, including:

acquiring a remote sensing satellite image;

performing semantic segmentation processing on the remote sensing satellite image to obtain a first topological graph, wherein the first topological graph comprises a road at a position corresponding to the remote sensing satellite image; determining a three-dimensional tensor map according to the remote sensing satellite image, wherein the three-dimensional tensor map characterizes the coding information of the road at the position corresponding to the remote sensing satellite image;

decoding the three-dimensional tensor image to obtain a second topological image, wherein the second topological image comprises a road in a position corresponding to the remote sensing satellite image;

and determining a road topological graph according to the first topological graph and the second topological graph, and updating a map according to the road topological graph.

According to a second aspect of the present disclosure, there is provided a training method applied to a map coding model for map update, including:

acquiring a plurality of remote sensing satellite images to be trained, wherein the remote sensing satellite images to be trained have standard three-dimensional tensor maps;

repeating the following steps until a preset condition is reached: inputting the remote sensing satellite image to be trained into a graph coding model to obtain a predicted three-dimensional tensor graph, wherein the predicted three-dimensional tensor graph represents coding information of a road at a position corresponding to the remote sensing satellite image to be trained; performing parameter adjustment on the graph coding model according to the predicted three-dimensional tensor graph and the standard three-dimensional tensor graph;

the map coding model obtained when the preset condition is reached is used for determining the three-dimensional tensor map of the remote sensing satellite image in the method of the first aspect of the disclosure.

According to a third aspect of the present disclosure, there is provided a map updating apparatus based on a remote sensing satellite image, comprising:

the acquisition unit is used for acquiring a remote sensing satellite image;

the first determining unit is used for performing semantic segmentation processing on the remote sensing satellite image to obtain a first topological graph, and the first topological graph comprises a road at a position corresponding to the remote sensing satellite image; determining a three-dimensional tensor map according to the remote sensing satellite image, wherein the three-dimensional tensor map characterizes the coding information of the road at the position corresponding to the remote sensing satellite image;

the second determining unit is used for decoding the three-dimensional tensor image to obtain a second topological image, wherein the second topological image comprises a road at a position corresponding to the remote sensing satellite image;

and the third determining unit is used for determining a road topological graph according to the first topological graph and the second topological graph and updating a map according to the road topological graph.

According to a fourth aspect of the present disclosure, there is provided a training apparatus applied to a map coding model for map updating, comprising:

the system comprises a first acquisition unit, a second acquisition unit and a third acquisition unit, wherein the first acquisition unit is used for acquiring a plurality of remote sensing satellite images to be trained, and the remote sensing satellite images to be trained have standard three-dimensional tensor maps;

a first determining unit, configured to repeat the following steps until a preset condition is reached: inputting the remote sensing satellite image to be trained into a graph coding model to obtain a predicted three-dimensional tensor graph, wherein the predicted three-dimensional tensor graph represents coding information of a road at a position corresponding to the remote sensing satellite image to be trained; performing parameter adjustment on the graph coding model according to the predicted three-dimensional tensor graph and the standard three-dimensional tensor graph;

wherein, the map coding model obtained when the preset condition is reached is used for determining the three-dimensional tensor map of the remote sensing satellite image in the device of the third aspect of the disclosure.

According to a fifth aspect of the present disclosure, there is provided a computer device comprising: at least one processor; and

a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of the first aspect or the second aspect.

According to a sixth aspect of the present disclosure, there is provided a non-transitory computer readable storage medium having stored thereon computer instructions for causing the computer to perform the method of the first or second aspect.

According to a seventh aspect of the present disclosure, there is provided a computer program product comprising: a computer program, stored in a readable storage medium, from which at least one processor of a computer device can read the computer program, execution of the computer program by the at least one processor causing the computer device to perform the method of the first aspect or the second aspect.

The technology solves the problems of high map updating cost and map updating errors caused by manual collection of road data.

It should be understood that the statements in this section do not necessarily identify key or critical features of the embodiments of the present disclosure, nor do they limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description.

Drawings

The drawings are included to provide a better understanding of the present solution and are not to be construed as limiting the present disclosure. Wherein:

FIG. 1 is a schematic diagram according to a first embodiment of the present disclosure;

FIG. 2 is a schematic illustration of a remote sensing satellite image according to a first embodiment of the disclosure;

FIG. 3 is a schematic diagram according to a second embodiment of the present disclosure;

FIG. 4 is a schematic diagram of a local feature map generation process according to a second embodiment of the present disclosure;

FIG. 5 is a schematic diagram according to a third embodiment of the present disclosure;

FIG. 6 is a schematic diagram according to a fourth embodiment of the present disclosure;

FIG. 7 is a schematic diagram of a user labeling a road key point in a remote sensing satellite image to be trained according to a fourth embodiment of the present disclosure;

fig. 8 is a schematic diagram illustrating statistics of the number of other neighboring road key points of a pixel point of a road key point according to a fourth embodiment of the disclosure;

fig. 9 is a graph of encoded data and three-dimensional tensor of each pixel point of a remote sensing satellite image to be trained according to a fourth embodiment of the present disclosure;

FIG. 10 is a schematic diagram according to a fifth embodiment of the present disclosure;

FIG. 11 is a schematic diagram according to a sixth embodiment of the present disclosure;

FIG. 12 is a schematic diagram according to a seventh embodiment of the present disclosure;

FIG. 13 is a schematic diagram according to an eighth embodiment of the present disclosure;

FIG. 14 is a schematic diagram according to a ninth embodiment of the present disclosure;

FIG. 15 is a schematic diagram according to a tenth embodiment of the present disclosure;

FIG. 16 is a block diagram of a computer device for implementing a method for remote sensing satellite image-based map updating according to an embodiment of the present disclosure;

fig. 17 is a block diagram of a computer device for implementing a training method of a graph coding model applied to map updating according to an embodiment of the present disclosure.

Detailed Description

Exemplary embodiments of the present disclosure are described below with reference to the accompanying drawings, in which various details of the embodiments of the disclosure are included to assist understanding, and which are to be considered as merely exemplary. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present disclosure. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.

The method and the device are suitable for being used in the scene of collecting the map information, and some road information, especially internal roads, namely some roads in the cell, may be lost in the current map information. If the remote sensing satellite image is processed by singly adopting semantic segmentation or the remote sensing satellite image is processed by singly adopting image coding, the problem that the recalled map information is incomplete can occur.

The invention provides a map updating method, a map training method and map updating equipment based on remote sensing satellite images, which are applied to the technical field of road networks in the field of data processing and are used for solving the problems of high map updating cost and map updating errors caused by manual collection of road data.

Fig. 1 is a schematic diagram according to a first embodiment of the present disclosure, and as shown in fig. 1, fig. 1 shows a map updating method based on a remote sensing satellite image, the method including:

and S101, acquiring a remote sensing satellite image.

Illustratively, a remote sensing satellite image refers to a film or a photo recording electromagnetic wave sizes of various ground objects, and has a high resolution, wherein the resolution includes a spatial resolution, a spectral resolution, a radiation resolution and a time resolution. The spatial resolution is the size or dimension of the smallest unit which can be distinguished in detail on the remote sensing satellite image, or refers to the measurement of the smallest angle or linear distance for distinguishing two targets by a remote sensor. Spectral resolution refers to the minimum wavelength separation that a remote sensor can resolve when receiving target radiation. The radiance resolution is the minimum difference in radiance that the remote sensor sensing element can resolve when receiving a spectral signal. The time resolution is a performance index related to the interval time of the remote sensing satellite images.

S102, performing semantic segmentation processing on the remote sensing satellite image to obtain a first topological graph, wherein the first topological graph comprises a road in a position corresponding to the remote sensing satellite image; and determining a three-dimensional tensor map according to the remote sensing satellite image, wherein the three-dimensional tensor map represents the coding information of the road at the position corresponding to the remote sensing satellite image.

Exemplarily, the semantic segmentation processing refers to segmenting the remote sensing satellite image according to regions, and each region is composed of pixel points with the same attribute. The semantic segmentation processing can be divided into an encoder network and a decoder network, wherein the encoder network is a classification network trained in advance, and the decoder network projects the recognition feature semantics learned by the encoder onto a pixel space to obtain dense classification. Further, the speech segmentation process requires not only differentiation at the pixel level, but also a mechanism to project the differentiated features learned at different stages of the encoder onto the pixel space. For example, FIG. 2 shows a schematic representation of a remotely sensed satellite image. The remote sensing satellite image of fig. 2 is analyzed to see the road structure of the ground.

The remote sensing satellite image shown in fig. 2 can obtain a first topological graph after semantic segmentation processing. The first topological graph can represent the boundary of the road and the background in the remote sensing satellite image.

In this embodiment, the three-dimensional tensor map represents the encoded information of the road at the position corresponding to the remote sensing satellite image. The three-dimensional tensor map is an image composed of three parts and can represent parameters of a remote sensing satellite image. The first part can represent whether the current pixel point is a road key point, the second part can represent whether the current pixel point has an adjacent road key point, and the third part can represent the position relationship between the current pixel point and other pixel points, wherein the position relationship can be the position offset of the current pixel point and other pixel points.

S103, decoding the three-dimensional tensor image to obtain a second topological image, wherein the second topological image comprises a road at a position corresponding to the remote sensing satellite image.

Illustratively, the process of decoding the three-dimensional tensor map is to reversely interpret the encoded data bits of the three parts of the three-dimensional tensor map according to the meaning of the encoded data bits of each part, so as to obtain the second topological map. Wherein the second topological graph can also represent the boundary of the road and the background in the remote sensing satellite image.

And S104, determining a road topological graph according to the first topological graph and the second topological graph, and updating the map according to the road topological graph.

Illustratively, the first topological graph and the second topological graph are compared, the difference between the first topological graph and the second topological graph is determined, and the road topological graph is obtained according to the difference between the first topological graph and the second topological graph. And after the road topological graph is obtained, updating the map by using the latest road topological graph.

The invention provides a map updating method based on a remote sensing satellite image, which comprises the following steps: the method comprises the steps of obtaining historical positioning information of a vehicle on a historical track, determining weight information of a frame according to the strength of a positioning signal corresponding to the frame and indicated by global positioning information of the frame, and determining optimized positioning information of the frame according to the global positioning information, interframe positioning information and weight information of the frame. By the technical scheme, the problems of high map updating cost and map updating errors caused by manual collection of road data can be solved.

Fig. 3 is a schematic diagram according to a second embodiment of the present disclosure, and as shown in fig. 3, fig. 3 shows a map updating method based on a remote sensing satellite image, the method including:

s301, obtaining a remote sensing satellite image.

For example, this step may refer to step S101, which is not described herein again.

S302, performing semantic segmentation processing on the remote sensing satellite image to obtain a first topological graph, wherein the first topological graph comprises a road in a position corresponding to the remote sensing satellite image.

For example, this step may refer to step S102, which is not described herein again.

S303, inputting the remote sensing satellite image into the image coding model to obtain a three-dimensional tensor map; the image coding model is obtained by training a remote sensing satellite image based on a standard three-dimensional tensor map.

Illustratively, the graph coding model is a model for outputting a three-dimensional tensor graph, and a plurality of remote sensing satellite images to be identified are input into the graph coding model, so that the three-dimensional tensor graph corresponding to each remote sensing satellite image to be identified can be obtained respectively. The graph coding model is obtained by training in advance. The advantage of this arrangement is that end-to-end input and output can be realized by using the graph coding model, and a three-dimensional tensor graph corresponding to the remote sensing satellite image can be rapidly output.

Illustratively, the remote sensing satellite image is input into the image coding model, and a three-dimensional tensor map is obtained, wherein the three-dimensional tensor map comprises the following steps:

carrying out feature extraction on the remote sensing satellite image based on the image coding model to obtain a global feature map and a local feature map; the global feature map represents global features of the remote sensing satellite image, and the local feature map represents road features of the remote sensing satellite image;

performing feature fusion on the global feature map and the local feature map based on the map coding model to obtain a fused feature map;

and generating a three-dimensional tensor map according to the fused feature map.

Illustratively, the global feature map is an overall feature used to describe color, texture, and/or shape features of the remotely sensed satellite image. The local feature map is used for describing local features of the remote sensing satellite image, and specifically, the local feature map can be features extracted from the remote sensing satellite image, including edges, corners, lines, curves, regions with special attributes, and the like. The local feature map has small correlation degree among features, and detection and matching of other local features cannot be influenced due to disappearance of partial local features under the shielding condition. And fusing the global feature map and the local feature map through a map coding model, wherein the fused feature map is composed of the global feature map and the local feature map, and the fused feature map does not modify the global feature map and the local feature map but only integrates the global feature map and the local feature map. And generating a corresponding three-dimensional tensor map from the remote sensing satellite image according to the fused feature map. The method has the advantages that the global features and the local features of the remote sensing satellite images can be fully combined, so that the information of the generated three-dimensional tensor map is more accurate and richer.

Illustratively, the feature extraction is performed on the remote sensing satellite image based on the graph coding model to obtain a global feature map and a local feature map, and the method comprises the following steps:

carrying out feature extraction on the remote sensing satellite image based on the image coding model to obtain a global feature image;

carrying out binarization processing on the global feature map to obtain a binarized feature map, wherein the binarized feature map comprises road features;

determining a road position area corresponding to the road point in the global feature map based on the road point in the binarized feature map; and generating a local feature map according to the road position area corresponding to the road point.

Exemplarily, after the global feature map is obtained, the global feature map is subjected to binarization processing, a road point in the global feature map is taken as 1, a non-road point in the global feature map is taken as 0, a binarized feature map is obtained, one road point is arbitrarily selected from the binarized feature map, a corresponding road point is found in the global feature map, and a road position area is determined by the road point, wherein the number and the shape of the road position area are not limited. For example, the road position area may be a circle defined by taking the road point as a center and a preset distance as a radius, and the circle may be used as the road position area of the road point. The road point can be used as a centroid, a preset range is used as 4 rectangles, and the 4 rectangles are used as the road position area of the road point. It should be noted that the road location area is a range including the road point, and the division criterion of the range is not limited. Specifically, see a schematic diagram of a local feature map generation process shown in fig. 4. It can be seen from the figure that after a global feature map a is output from a remote sensing satellite image a through a first-layer neural network model of a map coding model, binarization processing is performed on the global feature map a to obtain a binarized feature map B, a road point C is selected from the binarized feature map B, a road position area is determined in the global feature map a, and the road position area in the figure is 4 rectangles.

The method has the advantages that the global feature map and the local feature map are respectively obtained through the two-layer neural network model structure in the map coding model, and compared with the information obtained by a single-layer neural network model, the method is more comprehensive.

Exemplarily, the feature fusion is performed on the global feature map and the local feature map based on the map coding model to obtain a fused feature map, which includes:

carrying out up-sampling processing on the local feature map to obtain an up-sampled local feature map; the size of the up-sampled local feature map is the same as that of the global feature map;

and performing feature fusion on the global feature map and the up-sampled local feature map based on the map coding model to obtain a fused feature map.

For example, the upsampling process is based on the number of samples with a larger data size, and the number of samples with a smaller data size is generated to be the same as the number of samples with a larger data size. For example, in this embodiment, the local feature map may be 4 × N, the global feature map may be 8 × N, and the local feature map is subjected to upsampling using the global feature map as a standard, so that the upsampled local feature map is 8 × N. And (4) obtaining the fused feature map 8 x N (N + N) by the local feature map 8 x N and the global feature map 8 x N after the upsampling based on a map coding model. The advantage of this arrangement is that it is a more reasonable way to make the global feature map and the local feature map be compared in the same dimension.

S304, decoding the three-dimensional tensor image to obtain a second topological image, wherein the second topological image comprises a road in a position corresponding to the remote sensing satellite image.

For example, this step may refer to step S103, which is not described herein again.

S305, if the road pixel points in the first topological graph are determined not to exist in the second topological graph, adding the road pixel points in the first topological graph into the second topological graph to generate the road topological graph.

Illustratively, a road pixel point in the first topological graph is searched according to coordinate information in the first topological graph, if the coordinate information is a non-road pixel point in the second topological graph, the road pixel point in the first topological graph is added into the second topological graph, and the modified second topological graph is used as the road topological graph.

For example, if the coordinate information a (a, b) in the first topological graph is a road pixel, it is searched for whether the pixel of the coordinate information a (a, b) in the second topological graph is a road pixel, if so, no processing is performed on the second topological graph, and if not, the pixel of the coordinate information a (a, b) in the second topological graph is added as a road pixel. It should be noted that the division criteria of the coordinate information in the second topological graph and the first topological graph are the same, for example, the lower left corners of the first topological graph and the second topological graph are both the origin, the horizontal edge at the lowest of the first topological graph and the second topological graph is taken as the x-axis, and the vertical edge at the leftmost of the first topological graph and the second topological graph is taken as the y-axis. The method has the advantages that the road topological graph can be comprehensively determined by combining the two topological graphs, and the problem of insufficient road recall of one topological graph can be solved.

Further, the first topological graph is a binary graph, and the second topological graph is a binary graph. The advantage of this arrangement is that the efficiency of the comparison between the first topology map and the second topology map can be improved.

And S306, carrying out image enhancement processing on the road topological graph to obtain the road topological graph after enhancement processing.

For example, the image enhancement processing can be classified into spatial domain and frequency domain based methods according to the space where the process is located. Directly processing the road topological graph based on an airspace method; the method based on the frequency domain is to modify the transformation coefficient of the road topological graph in a certain transformation domain of the road topological graph and then inversely transform the road topological graph to the original space domain to obtain the road topological graph after enhancement processing. The road topological graph is arranged in the road surface, so that the visual effect of the road topological graph is improved, and the definition of the road topological graph is improved; or the characteristics of interest are highlighted and the characteristics of non-interest are suppressed aiming at the application occasion of the road topological graph, so that the difference between the characteristics of different objects in the road topological graph is enlarged.

Fig. 5 is a schematic diagram according to a third embodiment of the present disclosure, and as shown in fig. 5, fig. 5 illustrates a training method applied to a map coding model for map update, the method including:

s501, obtaining a plurality of remote sensing satellite images to be trained, wherein the remote sensing satellite images to be trained have standard three-dimensional tensor maps.

Illustratively, each remote sensing satellite image to be trained has a unique standard three-dimensional tensor map, and a plurality of remote sensing satellite images to be trained correspond to a plurality of standard three-dimensional tensor maps.

S502, repeating the following steps until a preset condition is reached: inputting the remote sensing satellite image to be trained into a graph coding model to obtain a predicted three-dimensional tensor map, wherein the predicted three-dimensional tensor map represents coding information of a road at a position corresponding to the remote sensing satellite image to be trained; adjusting parameters of the graph coding model according to the predicted three-dimensional tensor graph and the standard three-dimensional tensor graph; the image coding model obtained when the preset condition is reached is used for determining the three-dimensional tensor map of the remote sensing satellite image in the method of the embodiment.

Illustratively, the graph coding model is a model formed by two layers of neural networks, a remote sensing satellite image to be trained is input into the graph coding model, a predicted three-dimensional tensor map is output by the graph coding model, then the predicted three-dimensional tensor map is compared with a standard three-dimensional tensor map through a loss function to obtain parameters of each layer of neural network in the graph coding model, until the remote sensing satellite image to be trained is input into the graph coding model, the standard three-dimensional tensor map can be output, and the graph coding model reaching preset conditions at this time determines the three-dimensional tensor map of the remote sensing satellite image to be identified.

The invention provides a method for training a map coding model applied to map updating, which is characterized in that the map coding model is trained through a plurality of standard three-dimensional tensor maps corresponding to a plurality of remote sensing satellite images to be trained and a plurality of remote sensing satellite images to be trained, and the remote sensing satellite images to be identified are input into the obtained map coding model to determine the three-dimensional tensor maps of the remote sensing satellite images to be identified. By adopting the technical means, a more accurate image coding model can be obtained, and then the remote sensing satellite image to be identified can be input into the accurate image coding model to obtain a more accurate three-dimensional tensor map of the remote sensing satellite image to be identified.

Fig. 6 is a schematic diagram according to a fourth embodiment of the present disclosure, and as shown in fig. 6, fig. 6 illustrates a training method applied to a map coding model for map update, the method including:

s601, obtaining a plurality of remote sensing satellite images to be trained, wherein the remote sensing satellite images to be trained have standard three-dimensional tensor maps.

For example, this step may be referred to as step S501, and is not described herein again.

S602, extracting features of a remote sensing satellite image to be trained based on a graph coding model to obtain a global feature graph and a local feature graph; the global feature map represents global features of the remote sensing satellite image to be trained, and the local feature map represents road features of the remote sensing satellite image to be trained.

Exemplarily, the feature extraction is performed on the remote sensing satellite image to be trained based on the graph coding model to obtain a global feature map and a local feature map, and the method comprises the following steps:

carrying out feature extraction on a remote sensing satellite image to be trained based on a graph coding model to obtain a global feature graph;

For example, this step may refer to step S303, which is not described herein again.

And S603, performing feature fusion on the global feature map and the local feature map based on the map coding model to obtain a fused feature map.

In this embodiment, the feature fusion is performed on the global feature map and the local feature map based on the map coding model to obtain a fused feature map, which includes:

And S604, generating a three-dimensional tensor map according to the fused feature map.

S605, adjusting parameters of the graph coding model according to the predicted three-dimensional tensor diagram and the standard three-dimensional tensor diagram; the graph coding model obtained when the preset condition is reached is used for determining the three-dimensional tensor graph of the remote sensing satellite image in the method.

Illustratively, responding to the marking operation of a user, and acquiring a road key point in a remote sensing satellite image to be trained; generating road coding information of the remote sensing satellite image to be trained according to the road key points in the remote sensing satellite image to be trained and other road key points adjacent to the road key points; the road coding information comprises coding data of each pixel point of a remote sensing satellite image to be trained; and generating a standard three-dimensional tensor map of the remote sensing satellite image to be trained according to the road coding information of the remote sensing satellite image to be trained.

In this embodiment, a user marks road key points in a remote sensing satellite image to be trained, where the road key points in the remote sensing satellite image to be trained are different road junctions and end points of different roads. For example, fig. 7 is a schematic diagram of a user labeling road key points in a remote sensing satellite image to be trained, and it can be seen from fig. 7 that the road key points are A, B, C, D, E and F, the road key point a is connected with the road key point B, the road key point B is connected with the road key point C, and the road key point C is simultaneously connected with the road key point D and the road key point E. The road key point E connects the road key points F.

Further, in this embodiment, after determining a road key point in the remote sensing satellite image to be trained, other road key points adjacent to the road key point can be determined. For example, if the preliminarily determined road key point in the remote sensing satellite image to be trained is the road key point C, other road key points adjacent to the road key point are the road key point B, the road key point D and the road key point E. And generating the coded data of the road key point C for the road key point B, the road key point D and the road key point E according to the road key point C and other road key points adjacent to the road key point. The remote sensing satellite image marked by the user cannot be directly used for training the image coding model, so that the remote sensing satellite image marked by the user needs to be converted into a data form capable of training the image coding model.

The coded data of each pixel point of the remote sensing satellite image to be trained comprises one or more of the following information:

whether each pixel point in the remote sensing satellite image to be trained is a key point of a road or not;

the number of other adjacent road key points of the pixel points as the road key points;

and distance information between the pixel point serving as the road key point and other adjacent road key points of the pixel point serving as the road key point.

In this embodiment, the encoded data includes information about whether each pixel point in the remote sensing satellite image to be trained is a road key point, or the number of other adjacent road key points to the pixel point serving as the road key point, or the distance between the pixel point serving as the road key point and other adjacent road key points to the pixel point serving as the road key point.

The encoded data further includes: whether each pixel point in the remote sensing satellite image to be trained is a road key point or not and the number of other adjacent road key points of the pixel point serving as the road key point; the number of other adjacent road key points of the pixel points as the road key points and the distance information between the pixel points as the road key points and other adjacent road key points of the pixel points as the road key points; and whether each pixel point in the remote sensing satellite image to be trained is the distance information between the road key point and the pixel point serving as the road key point and other adjacent road key points serving as the pixel points of the road key point.

The encoded data further includes: and whether each pixel point in the remote sensing satellite image to be trained is a road key point, the number of other adjacent road key points of the pixel point serving as the road key point, and the distance information between the pixel point serving as the road key point and other adjacent road key points of the pixel point serving as the road key point.

In this embodiment, whether each pixel point in the remote sensing satellite image to be trained is a key point of a road can be represented by two coded data bits. For example, the encoded data bit is 10 when the pixel is a road key point, and the encoded data bit is 01 when the pixel is not a road key point. Two coded data bits of the road key point a, the road key point B, the road key point C, the road key point D, the road key point E and the road key point F in fig. 7 may be 10.

In this embodiment, the encoded data bits of the number of other road key points adjacent to the pixel point serving as the road key point are 12 bits, and the reason for this is that the number of other road key points adjacent to the pixel point serving as the road key point is generally 6 at most. In this embodiment, the coded data bit of the number of other adjacent road key points to the pixel point as the road key point is 12 bits only for illustration, and the coded data bit may be set by itself, for example, may be 20 bits. As seen from fig. 7, the number of other road key points adjacent to the pixel point of the road key point a is 1, the number of other road key points adjacent to the pixel point of the road key point B is 2, the number of other road key points adjacent to the pixel point of the road key point C is 3, the number of other road key points adjacent to the pixel point of the road key point D is 1, the number of other road key points adjacent to the pixel point of the road key point E is 2, and the number of other road key points adjacent to the pixel point of the road key point F is 1. After determining the number of other adjacent road key points of the pixel points of the road key points, six regions are divided by taking the pixel points of the road key points as the center, and the specific dividing manner can refer to a statistical schematic diagram of the number of other adjacent road key points of the pixel points of one road key point shown in fig. 8. It can be confirmed in the six regions in fig. 8 in turn, if there is a road key point in the first region, the first two bits of 12 bits of the encoded data bit are confirmed as 10, if there is a road key point in the second region, the third four bits of 12 bits of the encoded data bit are confirmed as 10, and so on.

The distance information between the pixel point serving as the road key point and other adjacent road key points serving as the pixel point of the road key point is specifically divided according to coordinate information of the remote sensing satellite image to be trained, the coordinate information of the pixel point of each road key point is confirmed, and the distance information between the pixel point of each road key point and the coordinate information of other adjacent road key points is confirmed. The coded data bits are 12 bits, wherein each two bits are distance information of other adjacent road key points. The distance information of the coded data bits corresponds to the coded data bits of the number of the road key points.

Specifically, fig. 9 shows encoded data and a three-dimensional tensor map of each pixel point of a remote sensing satellite image to be trained. As can be seen from fig. 9, it is assumed that encoded data of a pixel point is determined from the three-dimensional tensor map, where the pixel point is a pixel point of a road key point C, the encoded data is composed of three parts, a first part is whether the pixel point in the remote sensing satellite image to be trained is an encoded data bit of the road key point, and the encoded data bit of the first part is 10 because the pixel point is the pixel point of the road key point C; the second part is the number of other adjacent road key points which are the pixel points of the road key point, and since the pixel point is the pixel point of the road key point C and the number of other adjacent road key points is 3, as can be seen from fig. 8, the road key point C exists in one region, the third region and the fourth region, the encoded data bit of the second part is 100110100101; the third part is distance information between the pixel point as the road key point and other adjacent road key points as the pixel point of the road key point, and the distance information between the road key point B, the road key point D and the road key point E and the road key point C is calculated as the pixel point of the road key point C.

The advantage of this setting is that can accurately characterize the road key point information of the remote sensing satellite image to be trained through accurate data format.

The present disclosure provides a training method of a graph coding model applied to map updating, including: acquiring a plurality of remote sensing satellite images to be trained, wherein the remote sensing satellite images to be trained have standard three-dimensional tensor maps, and acquiring road key points in the remote sensing satellite images to be trained in response to marking operation of a user; generating road coding information of the remote sensing satellite image to be trained according to the road key points in the remote sensing satellite image to be trained and other road key points adjacent to the road key points; the road coding information comprises coding data of each pixel point of a remote sensing satellite image to be trained; and generating a standard three-dimensional tensor map of the remote sensing satellite image to be trained according to the road coding information of the remote sensing satellite image to be trained. By adopting the technical scheme, an accurate image coding model can be obtained, and then a more accurate three-dimensional tensor map is output, so that the recall rate of the obtained second topological map is higher, the updating speed of the map is ensured, and the operation process is optimized.

Fig. 10 is a schematic diagram according to a fifth embodiment of the present disclosure, and as shown in fig. 10, fig. 10 shows a map updating apparatus based on a remote sensing satellite image, the apparatus 10 includes:

an obtaining unit 1001 is configured to obtain a remote sensing satellite image.

The first determining unit 1002 is configured to perform semantic segmentation processing on a remote sensing satellite image to obtain a first topological graph, where the first topological graph includes a road in a position corresponding to the remote sensing satellite image; and determining a three-dimensional tensor map according to the remote sensing satellite image, wherein the three-dimensional tensor map represents the coding information of the road at the position corresponding to the remote sensing satellite image.

The second determining unit 1003 is configured to decode the three-dimensional tensor image to obtain a second topological image, where the second topological image includes a road at a position corresponding to the remote sensing satellite image;

and a third determining unit 1004, configured to determine a road topology map according to the first topology map and the second topology map, and update the map according to the road topology map.

It can be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working process of the above-described apparatus may refer to the corresponding process in the foregoing method embodiment, and is not described herein again.

Fig. 11 is a schematic diagram according to a sixth embodiment of the present disclosure, and as shown in fig. 11, fig. 11 shows a map updating apparatus based on a remote sensing satellite image, where the apparatus 11 includes:

an acquisition unit 1101 for acquiring a remote sensing satellite image;

the first determining unit 1102 is configured to perform semantic segmentation processing on the remote sensing satellite image to obtain a first topological graph, where the first topological graph includes a road in a position corresponding to the remote sensing satellite image; and determining a three-dimensional tensor map according to the remote sensing satellite image, wherein the three-dimensional tensor map represents the coding information of the road at the position corresponding to the remote sensing satellite image.

The second determining unit 1103 is configured to decode the three-dimensional tensor image to obtain a second topological graph, where the second topological graph includes a road at a position corresponding to the remote sensing satellite image.

And a third determining unit 1104, configured to determine a road topology map according to the first topology map and the second topology map, and update the map according to the road topology map.

Exemplarily, among others, the first determining unit 1102 includes:

the first determining module 11021 is used for inputting the remote sensing satellite image into the image coding model to obtain a three-dimensional tensor map; the image coding model is obtained by training a remote sensing satellite image based on a standard three-dimensional tensor map.

Illustratively, among other things, the first determining module 11021 includes:

the extraction submodule 110211 is used for extracting the characteristics of the remote sensing satellite image based on the image coding model to obtain a global characteristic image and a local characteristic image; the global characteristic graph represents global characteristics of the remote sensing satellite image, and the local characteristic graph represents road characteristics of the remote sensing satellite image.

And the fusion sub-module 110212 is used for performing feature fusion on the global feature map and the local feature map based on the map coding model to obtain a fused feature map.

And a generation submodule 110213 for generating a three-dimensional tensor map according to the fused feature map.

Illustratively, among other things, the extraction sub-module 110211 includes:

and carrying out binarization processing on the global feature map to obtain a binarized feature map, wherein the binarized feature map comprises the features of the road.

Illustratively, among other things, the fusion submodule 110212 includes:

carrying out up-sampling processing on the local feature map to obtain an up-sampled local feature map; the size of the up-sampled local feature map is the same as the size of the global feature map.

Exemplarily, the third determining unit 1104 includes:

the adding module 11041 is configured to add the road pixel in the first topological graph to the second topological graph to generate the road topological graph if it is determined that the road pixel in the first topological graph does not exist in the second topological graph.

Exemplarily, wherein, the apparatus further comprises:

the processing unit 1105 is configured to perform image enhancement processing on the road topology map to obtain an enhanced road topology map.

Illustratively, the first topological graph is a binary graph, and the second topological graph is a binary graph.

Fig. 12 is a schematic diagram according to a seventh embodiment of the present disclosure, and as shown in fig. 12, fig. 12 shows a training apparatus for a graph coding model applied to map updating, where the apparatus 12 includes:

the first obtaining unit 1201 is used for obtaining a plurality of remote sensing satellite images to be trained, wherein the remote sensing satellite images to be trained have standard three-dimensional tensor maps;

a first determining unit 1202, configured to repeat the following steps until a preset condition is reached: inputting the remote sensing satellite image to be trained into a graph coding model to obtain a predicted three-dimensional tensor map, wherein the predicted three-dimensional tensor map represents coding information of a road at a position corresponding to the remote sensing satellite image to be trained; adjusting parameters of the graph coding model according to the predicted three-dimensional tensor graph and the standard three-dimensional tensor graph;

wherein the map coding model obtained when the preset condition is reached is used for determining the three-dimensional tensor map of the remote sensing satellite image in the device of any one of claims 15-22.

Fig. 13 is a schematic diagram according to an eighth embodiment of the present disclosure, and as shown in fig. 13, fig. 13 shows a training apparatus for a graph coding model applied to map updating, where the apparatus 13 includes:

the first obtaining unit 1301 is configured to obtain a plurality of remote sensing satellite images to be trained, where the remote sensing satellite images to be trained have a standard three-dimensional tensor map.

A first determining unit 1302, configured to repeat the following steps until a preset condition is reached: inputting the remote sensing satellite image to be trained into a graph coding model to obtain a predicted three-dimensional tensor map, wherein the predicted three-dimensional tensor map represents coding information of a road at a position corresponding to the remote sensing satellite image to be trained; and carrying out parameter adjustment on the graph coding model according to the predicted three-dimensional tensor graph and the standard three-dimensional tensor graph.

The map coding model obtained when the preset condition is reached is used for determining the three-dimensional tensor map of the remote sensing satellite image in the device in the embodiment.

Exemplarily, the first determining unit 1302 includes:

an extraction module 13021, configured to perform feature extraction on a remote sensing satellite image to be trained based on a graph coding model to obtain a global feature map and a local feature map; the global feature map represents global features of the remote sensing satellite image to be trained, and the local feature map represents road features of the remote sensing satellite image to be trained.

A determining module 13022, configured to perform feature fusion on the global feature map and the local feature map based on the map coding model to obtain a fused feature map.

A generating module 13023, configured to generate a three-dimensional tensor map according to the fused feature map.

Illustratively, among other things, the extraction module 13021 includes:

and the extraction submodule 130211 is used for extracting the features of the remote sensing satellite image to be trained on the basis of the image coding model to obtain a global feature map.

The processing submodule 130212 is configured to perform binarization processing on the global feature map to obtain a binarized feature map, where the binarized feature map includes features of a road.

A generation submodule 130213, configured to determine, based on the road point in the binarized feature map, a road position area corresponding to the road point in the global feature map; and generating a local feature map according to the road position area corresponding to the road point.

Exemplarily, the determining module 13022 includes, among others:

the processing submodule 130221 is configured to perform upsampling processing on the local feature map to obtain an upsampled local feature map; the size of the up-sampled local feature map is the same as the size of the global feature map.

And the fusion sub-module 130222 is configured to perform feature fusion on the global feature map and the up-sampled local feature map based on the map coding model to obtain a fused feature map.

Exemplarily, the method further comprises the following steps:

and the second obtaining unit 1303, configured to obtain a road key point in the remote sensing satellite image to be trained in response to a labeling operation of the user.

A first generating unit 1304, configured to generate road coding information of the remote sensing satellite image to be trained according to a road key point in the remote sensing satellite image to be trained and other road key points adjacent to the road key point; the road coding information comprises coding data of each pixel point of the remote sensing satellite image to be trained.

The second generating unit 1305 is configured to generate a standard three-dimensional tensor map of the remote sensing satellite image to be trained according to the road coding information of the remote sensing satellite image to be trained.

Exemplarily, the method further comprises the following steps: the coded data of each pixel point of the remote sensing satellite image to be trained comprises one or more of the following information: whether each pixel point in the remote sensing satellite image to be trained is a key point of a road or not; the number of other adjacent road key points of the pixel points as the road key points; and distance information between the pixel point serving as the road key point and other adjacent road key points of the pixel point serving as the road key point.

The present disclosure also provides a computer device, a readable storage medium, and a computer program product according to embodiments of the present disclosure.

According to an embodiment of the present disclosure, the present disclosure also provides a computer program product comprising: a computer program, the computer program being stored in a readable storage medium, from which the computer program can be read by at least one processor of a computer device, execution of the computer program by the at least one processor causing the computer device to carry out the solution provided by any of the embodiments described above.

Fig. 14 is a schematic diagram according to a ninth embodiment of the present disclosure, and as shown in fig. 14, a server 1400 in the present disclosure may include: a processor 1401, and a memory 1402.

A memory 1402 for storing programs; the Memory 1402 may include a volatile Memory (RAM), such as a Static Random Access Memory (SRAM), a Double Data Rate Synchronous Dynamic Random Access Memory (DDR SDRAM), and the like; the memory may also comprise a non-volatile memory, such as a flash memory. The memory 1402 is used to store computer programs (e.g., applications, functional modules, etc. that implement the above-described methods), computer instructions, etc., which can be stored in partitions in the one or more memories 1402. And the above-described computer programs, computer instructions, data, and the like, can be called by the processor 1401.

The computer programs, computer instructions, etc. described above may be stored in one or more memories 1402 in a partitioned manner. And the above-described computer program, computer data, or the like can be called by the processor 1401.

A processor 1401 for executing the computer program stored in the memory 1402 to implement the steps of the method for updating a map based on remote sensing satellite images according to the above embodiments.

Reference may be made in particular to the description relating to the preceding method embodiment.

The processor 1401 and the memory 1402 may be separate structures or may be an integrated structure integrated together. When the processor 1401 and the memory 1402 are separate structures, the memory 1402 and the processor 1401 may be coupled by a bus 1403.

The server of this embodiment may execute the technical solution in the method, and the specific implementation process and the technical principle are the same, which are not described herein again.

According to an embodiment of the present disclosure, there is also provided a non-transitory computer readable storage medium storing computer instructions for causing a computer to execute the solution provided by the above respective embodiments.

According to an embodiment of the present disclosure, the present disclosure also provides a computer program product comprising: a computer program, stored in a readable storage medium, from which at least one processor of the server can read the computer program, the at least one processor executing the computer program to cause the server to perform the solution provided by the respective embodiments described above.

According to an embodiment of the present disclosure, the present disclosure also provides a computer program product comprising: a computer program, stored in a readable storage medium, from which at least one processor of a control device of a vehicle can read the computer program, the execution of which by the at least one processor causes the control device of the vehicle to carry out the solution provided by the respective embodiments described above.

Fig. 15 is a schematic diagram according to a tenth embodiment of the present disclosure, and as shown in fig. 15, a server 1500 in the present disclosure may include: a processor 1501 and memory 1502.

A memory 1502 for storing programs; the Memory 1502 may include a volatile Memory (RAM), such as a Random Access Memory (SRAM), a Double Data Rate Synchronous Dynamic Random Access Memory (DDR SDRAM), and the like; the memory may also comprise a non-volatile memory, such as a flash memory. The memory 1502 is used to store computer programs (e.g., applications, functional modules, etc. that implement the above-described methods), computer instructions, etc., which can be stored in one or more of the memories 1502 in a partitioned manner. And the above-described computer programs, computer instructions, data, and the like, can be called by the processor 1501.

The computer programs, computer instructions, and the like described above can be stored in one or more memories 1502 in partitions. And the above-mentioned computer program, computer data, or the like can be called by the processor 1501.

The processor 1501 is configured to execute the computer program stored in the memory 1502 to implement the steps of the training method applied to the map coding model for map update according to the embodiment.

The processor 1501 and the memory 1502 may be separate structures or may be integrated structures that are integrated together. When the processor 1501 and the memory 1502 are separate structures, the memory 1502 and the processor 1501 may be coupled by a bus 1503.

FIG. 16 shows a schematic block diagram of an example computer device 1600 that can be used to implement embodiments of the present disclosure. Computer apparatus is intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, computer apparatus, blade computer apparatus, mainframes, and other appropriate computers. The computer device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be examples only, and are not meant to limit implementations of the disclosure described and/or claimed herein.

As shown in fig. 16, the computer apparatus 1600 includes a computing unit 1601, which can perform various appropriate actions and processes according to a computer program stored in a Read Only Memory (ROM)1602 or a computer program loaded from a storage unit 1608 into a Random Access Memory (RAM) 1603. In the RAM 1603, various programs and data required for the operation of the device 1600 can also be stored. The computing unit 1601, ROM 1602 and RAM 1603 are connected to each other via a bus 1604. An input/output (I/O) interface 1605 is also connected to the bus 1604.

A number of components in computer device 1600 are connected to I/O interface 1605, including: an input unit 1606 such as a keyboard, a mouse, and the like; an output unit 1607 such as various types of displays, speakers, and the like; a storage unit 1608, such as a magnetic disk, optical disk, or the like; and a communication unit 1609 such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 1609 allows the computer device 1600 to exchange information/data with other devices via a computer network, such as the internet, and/or various telecommunications networks.

Computing unit 1601 may be a variety of general purpose and/or special purpose processing components with processing and computing capabilities. Some examples of computing unit 1601 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various specialized Artificial Intelligence (AI) computing chips, various computing units running machine learning model algorithms, a Digital Signal Processor (DSP), and any suitable processor, controller, microcontroller, and so forth. The calculation unit 1601 executes the respective methods and processes described above, such as a processing method for generating positioning information of a high-precision map. For example, in some embodiments, the method model training for image processing may be implemented as a computer software program tangibly embodied in a machine-readable medium, such as storage unit 1608. In some embodiments, part or all of the computer program can be loaded and/or installed onto computer device 1600 via ROM 1602 and/or communications unit 1609. When loaded into RAM 1603 and executed by computing unit 1601, the computer program may perform one or more steps of the remotely sensed satellite image based map update method described above. Alternatively, in other embodiments, the computing unit 1601 may be configured by any other suitable means (e.g., by means of firmware) to perform the method for model training of image processing.

Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuitry, Field Programmable Gate Arrays (FPGAs), Application Specific Integrated Circuits (ASICs), Application Specific Standard Products (ASSPs), system on a chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, receiving data and instructions from, and transmitting data and instructions to, a storage system, at least one input device, and at least one output device.

Program code for implementing the methods of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the program codes, when executed by the processor or controller, cause the functions/operations specified in the flowchart and/or block diagram to be performed. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine or entirely on the remote machine or computer device.

In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which a user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic, speech, or tactile input.

The systems and techniques described here can be implemented in a computing system that includes a back-end component (e.g., as a data computer device), or that includes a middleware component (e.g., an application computer device), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), Wide Area Networks (WANs), and the Internet.

The computer system may include a client and a computer device. The client and computer devices are generally remote from each other and typically interact through a communication network. The relationship of client and computer devices arises by virtue of computer programs running on the respective computers and having a client-computer device relationship to each other. The computer device may be a cloud computer device, which is also called a cloud computer device or a cloud host, and is a host product in a cloud computing service system, so as to solve the defects of high management difficulty and weak service extensibility in the conventional physical host and VPS service ("Virtual Private Server", or "VPS" for short). The computer device may also be a computer device of a distributed system or a computer device incorporating a blockchain.

FIG. 17 shows a schematic block diagram of an example computer device 1700 that can be used to implement embodiments of the present disclosure. Computer apparatus is intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, computer apparatus, blade computer apparatus, mainframes, and other appropriate computers. The computer device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be examples only, and are not meant to limit implementations of the disclosure described and/or claimed herein.

As shown in fig. 17, the computer apparatus 1700 includes a computing unit 1701 that can perform various appropriate actions and processes according to a computer program stored in a Read Only Memory (ROM)1702 or a computer program loaded from a storage unit 1708 into a Random Access Memory (RAM) 1703. In the RAM 1703, various programs and data required for the operation of the device 1700 can also be stored. The computing unit 1701, the ROM 1702, and the RAM 1703 are connected to each other through a bus 1704. An input/output (I/O) interface 1705 is also connected to bus 1704.

Various components within computer device 1700 connect to I/O interface 1705, including: an input unit 1706 such as a keyboard, a mouse, and the like; an output unit 1707 such as various types of displays, speakers, and the like; a storage unit 1708 such as a magnetic disk, optical disk, or the like; and a communication unit 1709 such as a network card, modem, wireless communication transceiver, etc. A communication unit 1709 allows the computer device 1700 to exchange information/data with other devices via a computer network such as the internet and/or various telecommunication networks.

The computing unit 1701 may be a variety of general purpose and/or special purpose processing components with processing and computing capabilities. Some examples of the computing unit 1701 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various dedicated Artificial Intelligence (AI) computing chips, various computing units running machine learning model algorithms, a Digital Signal Processor (DSP), and any suitable processor, controller, microcontroller, and the like. The calculation unit 1701 executes the respective methods and processes described above, such as a processing method for generating positioning information of a high-precision map. For example, in some embodiments, the method model training for image processing may be implemented as a computer software program tangibly embodied in a machine-readable medium, such as storage unit 1708. In some embodiments, part or all of the computer program may be loaded and/or installed onto computer device 1700 via ROM 1702 and/or communications unit 1709. When the computer program is loaded into RAM 1703 and executed by the computing unit 1701, one or more steps of the above-described training method applied to the map coding model for map updating may be performed. Alternatively, in other embodiments, the computing unit 1701 may be configured in any other suitable manner (e.g., by way of firmware) to perform a method for model training for image processing.

Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuitry, Field Programmable Gate Arrays (FPGAs), Application Specific Integrated Circuits (ASICs), Application Specific Standard Products (ASSPs), system on a chip (SOCs), Complex Programmable Logic Devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, receiving data and instructions from, and transmitting data and instructions to, a storage system, at least one input device, and at least one output device.

It should be understood that various forms of the flows shown above may be used, with steps reordered, added, or deleted. For example, the steps described in the present disclosure may be executed in parallel, sequentially, or in different orders, as long as the desired results of the technical solutions disclosed in the present disclosure can be achieved, and the present disclosure is not limited herein.

The above detailed description should not be construed as limiting the scope of the disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made in accordance with design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present disclosure should be included in the scope of protection of the present disclosure.

Claims

1. A map updating method based on remote sensing satellite images comprises the following steps:

acquiring a remote sensing satellite image;

2. The method of claim 1, wherein determining a three-dimensional tensor map from the remote sensing satellite image comprises:

inputting the remote sensing satellite image into an image coding model to obtain a three-dimensional tensor map; the image coding model is obtained by training a remote sensing satellite image based on a standard three-dimensional tensor map.

3. The method of claim 2, wherein inputting the remote sensing satellite image into a graph coding model, resulting in a three-dimensional tensor map, comprises:

carrying out feature extraction on the remote sensing satellite image based on the image coding model to obtain a global feature image and a local feature image; the global feature map represents global features of the remote sensing satellite image, and the local feature map represents road features of the remote sensing satellite image;

and generating the three-dimensional tensor map according to the fused feature map.

4. The method of claim 3, wherein performing feature extraction on the remote sensing satellite image based on the graph coding model to obtain a global feature map and a local feature map comprises:

carrying out feature extraction on the remote sensing satellite image based on the image coding model to obtain the global feature image;

determining a road position area corresponding to the road point in the global feature map based on the road point in the binarized feature map; and generating the local feature map according to the road position area corresponding to the road point.

5. The method according to claim 3 or 4, wherein performing feature fusion on the global feature map and the local feature map based on the map coding model to obtain a fused feature map comprises:

performing upsampling processing on the local feature map to obtain an upsampled local feature map; the size of the up-sampled local feature map is the same as that of the global feature map;

and performing feature fusion on the global feature map and the up-sampled local feature map based on the map coding model to obtain the fused feature map.

6. The method of any of claims 1-5, wherein determining a road topology map from the first topology map and the second topology map comprises:

and if the road pixel points in the first topological graph are determined not to exist in the second topological graph, adding the road pixel points in the first topological graph into the second topological graph to generate the road topological graph.

7. The method according to any of claims 1-6, wherein after determining a road topology map from the first topology map and the second topology map, further comprising:

and carrying out image enhancement processing on the road topological graph to obtain the road topological graph after enhancement processing.

8. The method according to any of claims 1-7, wherein the first topology map is a binary map and the second topology map is a binary map.

9. A training method of a graph coding model applied to map updating comprises the following steps:

wherein, the map coding model obtained by reaching the preset condition is used for determining the three-dimensional tensor map of the remote sensing satellite image in the method of any one of claims 1 to 8.

10. The method of claim 9, wherein inputting the remote sensing satellite image to be trained into a graph coding model to obtain a predicted three-dimensional tensor map comprises:

extracting the features of the remote sensing satellite image to be trained based on the image coding model to obtain a global feature map and a local feature map; the global characteristic chart is used for characterizing global characteristics of the remote sensing satellite image to be trained, and the local characteristic chart is used for characterizing road characteristics of the remote sensing satellite image to be trained;

11. The method of claim 10, wherein the extracting features of the remote sensing satellite image to be trained based on the graph coding model to obtain a global feature map and a local feature map comprises:

extracting the features of the remote sensing satellite image to be trained based on the image coding model to obtain the global feature image;

12. The method according to claim 10 or 11, wherein performing feature fusion on the global feature map and the local feature map based on the map coding model to obtain a fused feature map comprises:

13. The method according to any one of claims 9-12, further comprising:

responding to the marking operation of a user, and acquiring the road key points in the remote sensing satellite image to be trained;

generating road coding information of the remote sensing satellite image to be trained according to the road key points in the remote sensing satellite image to be trained and other road key points adjacent to the road key points; the road coding information comprises coding data of each pixel point of a remote sensing satellite image to be trained;

and generating a standard three-dimensional tensor map of the remote sensing satellite image to be trained according to the road coding information of the remote sensing satellite image to be trained.

14. The method of claim 13, wherein the encoded data for each pixel point of the remote sensing satellite image to be trained comprises one or more of the following information:

whether each pixel point in the remote sensing satellite image to be trained is a key point of the road or not;

15. A map updating device based on remote sensing satellite images comprises:

the acquisition unit is used for acquiring a remote sensing satellite image;

16. The apparatus of claim 15, wherein the first determining unit comprises:

the first determining module is used for inputting the remote sensing satellite image into an image coding model to obtain a three-dimensional tensor map; the image coding model is obtained by training a remote sensing satellite image based on a standard three-dimensional tensor map.

17. The apparatus of claim 16, wherein the first determining means comprises:

the extraction submodule is used for extracting the features of the remote sensing satellite image based on the image coding model to obtain a global feature map and a local feature map; the global feature map represents global features of the remote sensing satellite image, and the local feature map represents road features of the remote sensing satellite image;

the fusion submodule is used for carrying out feature fusion on the global feature map and the local feature map based on the map coding model to obtain a fused feature map;

and the generation submodule is used for generating the three-dimensional tensor map according to the fused feature map.

18. The apparatus of claim 17, wherein the extraction submodule comprises:

19. The apparatus of claim 17 or 18, wherein the fusion submodule comprises:

20. The apparatus according to any of claims 15-19, wherein the third determining unit comprises:

and the adding module is used for adding the road pixel points in the first topological graph into the second topological graph to generate the road topological graph if the road pixel points in the first topological graph are determined not to exist in the second topological graph.

21. The apparatus of any one of claims 15-20, wherein the apparatus further comprises:

and the processing unit is used for carrying out image enhancement processing on the road topological graph to obtain the road topological graph after enhancement processing.

22. The apparatus according to any of claims 15-21, wherein the first topology map is a binary map and the second topology map is a binary map.

23. A training apparatus for a graph coding model applied to map updating, comprising:

wherein the map coding model obtained when the preset condition is reached is used for determining a three-dimensional tensor map of the remote sensing satellite image in the device of any one of claims 15-22.

24. The apparatus of claim 23, wherein the first determining unit comprises:

the extraction module is used for extracting the features of the remote sensing satellite image to be trained based on the image coding model to obtain a global feature image and a local feature image; the global characteristic chart is used for characterizing global characteristics of the remote sensing satellite image to be trained, and the local characteristic chart is used for characterizing road characteristics of the remote sensing satellite image to be trained;

the determining module is used for carrying out feature fusion on the global feature map and the local feature map based on the map coding model to obtain a fused feature map;

and the generating module is used for generating the three-dimensional tensor map according to the fused feature map.

25. The apparatus of claim 24, wherein the extraction module comprises:

the extraction submodule is used for extracting the features of the remote sensing satellite image to be trained on the basis of the image coding model to obtain the global feature image;

the processing submodule is used for carrying out binarization processing on the global feature map to obtain a binarized feature map, and the binarized feature map comprises road features;

a generation submodule, configured to determine, based on a road point in the binarized feature map, a road location area corresponding to the road point in the global feature map; and generating the local feature map according to the road position area corresponding to the road point.

26. The apparatus of claim 24 or 25, wherein the determining means comprises:

the processing submodule is used for carrying out up-sampling processing on the local characteristic diagram to obtain an up-sampled local characteristic diagram; the size of the up-sampled local feature map is the same as that of the global feature map;

and the fusion submodule is used for carrying out feature fusion on the global feature map and the up-sampled local feature map based on the map coding model to obtain the fused feature map.

27. The apparatus of any of claims 23-26, further comprising:

the second acquisition unit is used for responding to marking operation of a user and acquiring the road key points in the remote sensing satellite image to be trained;

the first generating unit is used for generating road coding information of the remote sensing satellite image to be trained according to the road key points in the remote sensing satellite image to be trained and other road key points adjacent to the road key points; the road coding information comprises coding data of each pixel point of a remote sensing satellite image to be trained;

and the second generation unit is used for generating a standard three-dimensional tensor map of the remote sensing satellite image to be trained according to the road coding information of the remote sensing satellite image to be trained.

28. The apparatus of claim 27, wherein the encoded data for each pixel point of the remote sensing satellite image to be trained comprises one or more of:

29. A computer device, comprising:

at least one processor; and

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-8.

30. A computer device, comprising:

at least one processor; and

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 9-14.

31. A non-transitory computer readable storage medium having stored thereon computer instructions for causing the computer to perform the method of any one of claims 1-8.

32. A non-transitory computer readable storage medium having stored thereon computer instructions for causing the computer to perform the method of any of claims 9-14.

33. A computer program product comprising a computer program which, when executed by a processor, carries out the steps of the method of any one of claims 1 to 8.

34. A computer program product comprising a computer program which, when executed by a processor, carries out the steps of the method of any one of claims 9 to 14.