Vehicle route planning method, device and equipment based on environmental perception

Abstract

The invention relates to the technical field of route planning, in particular to a vehicle route planning method, device and equipment based on environmental perception. According to the method, the physical element information and the emotion perception data of each road section are identified from the streetscape image of the road section, and then the road sections are comprehensively evaluated according to the physical element information and the emotion perception data of the road sections, so that the comprehensive score is obtained; and finally, according to the comprehensive score, the road section length and the number of passengers who need to take the vehicle on each road section, intelligently screening a connected target road section from each road section, and planning a driving route of the vehicle from the starting point to the ending point by using the target road section. According to the analysis, when the route is planned, not only are physical elements of the road section considered, but also the emotion notification data of the road section are fully considered, so that the planned route can meet the actual requirements of passengers.

Description

Technical Field

[0001] The present invention relates to the technical field of route planning, and in particular to a vehicle route planning method, device and equipment based on environment perception. Background Art

[0002] Vehicles travel along different routes from the starting point to the end point, and the street scenes on different routes will bring different experiences to passengers. The experience is related to the visual quality of the street, and the visual quality of the street is not only reflected in the physical conditions of the street, such as the vegetation coverage on both sides of the road, the breadth of vision, etc., but also includes the emotional perception brought by the street scene: environmental aesthetics, historical and cultural characteristics, etc. These factors together constitute the "perceived quality" of the street, which directly affects the perception and experience of tourists. However, when planning the driving route of the vehicle, the existing technology only considers the physical conditions of the street, but ignores the emotional perception brought to the passengers by the street scene, which makes it difficult for the planned route to meet the actual needs of the passengers.

[0003] In summary, the routes planned by the existing technology are difficult to meet the actual needs of passengers.

[0004] Therefore, the prior art still needs to be improved and enhanced. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a vehicle route planning method, device and equipment based on environmental perception, which solves the problem that the routes planned by the prior art are difficult to meet the actual needs of passengers.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] In a first aspect, the present invention provides a vehicle route planning method based on environment perception, which comprises:

[0008] Acquire each road section between the starting point and the end point, and acquire a street view image of each road section, and identify physical element information and emotional perception data on the street view image, wherein the physical element information is used to characterize object information contained in the environment where the road section is located, and the emotional perception data is used to characterize emotional responses that can be induced by the environment where the road section is located;

[0009] Obtaining a comprehensive score for each of the road sections based on the physical element information and the emotion perception data of each of the road sections;

[0010] The length of each of the road sections is obtained, the number of passengers that need to take the vehicle for each of the road sections is obtained, and based on the length of each of the road sections, the number of passengers for each of the road sections and the comprehensive score of each of the road sections, connected target road sections are screened out from each of the road sections, and the driving route of the vehicle from the starting point to the end point is planned based on the path where the target road sections are located.

[0011] In one implementation, sky pixels, vegetation pixels, and building pixels on the street view image are identified through a semantic segmentation model, and the number of sky pixels, the number of vegetation pixels, and the number of building pixels are used as the physical element information; the semantic segmentation model includes a backbone network, a hole convolution, an ASPP module, and a decoder connected in sequence.

[0012] In one implementation, the vitality score, beauty score, safety score, richness score, boredom score and depression score corresponding to the street view image are determined by an image classification model, and the vitality score, the beauty score, the safety score, the richness score, the boredom score and the depression score are used as the emotion perception data; the image classification model includes a convolutional layer, a residual module, a global average pooling layer and a fully connected layer connected in sequence.

[0013] In one implementation, obtaining a comprehensive score for each of the road sections based on the physical element information and the emotion perception data of each of the road sections includes:

[0014] Determining positive physical indicators and negative physical indicators in the physical element information, wherein the positive physical indicators are physical indicators that can enhance passenger experience, and the negative physical indicators are physical indicators that reduce passenger experience;

[0015] Determining a positive emotion perception index and a negative emotion perception index in the emotion perception data, wherein the positive emotion perception index is an emotion index that enhances the passenger experience, and the negative emotion perception index is an emotion index that reduces the passenger experience;

[0016] According to the positive emotion perception index, the negative emotion perception index, the positive physical index, and the negative physical index of each road section, a comprehensive score of each road section is obtained.

[0017] In an implementation manner, a distance between the shooting geographic coordinates of the street view image and the road geographic coordinates of the road segment is less than a threshold.

[0018] In one implementation, the selecting the connected target road segments from the road segments according to the lengths of the road segments, the number of passengers of the road segments and the comprehensive scores of the road segments includes:

[0019] Dividing each of the road sections into a plurality of groups, each road section in each of the groups having connectivity;

[0020] According to the number of passengers on the road section in each group, a first function is established for each group; according to the comprehensive score of the road section in each group, a second function is established for each group;

[0021] Obtaining the capacity limit of the vehicle, obtaining the straight line length from the starting point to the end point, and establishing a constraint condition for each of the groups according to the capacity limit and the straight line length;

[0022] According to the constraint condition of each of the groups and the first function and the second function, a target group is screened out from each of the groups, and each of the road sections in the target group is used as a target road section.

[0023] In one implementation, the selecting the connected target road segments from the road segments according to the lengths of the road segments, the number of passengers of the road segments and the comprehensive scores of the road segments includes:

[0024] Constructing a route optimization model according to the length of each of the road sections, the number of passengers on each of the road sections, and the comprehensive score of each of the road sections;

[0025] A non-dominated sorting algorithm is applied to the route optimization model to select connected target road segments from each of the road segments.

[0026] In a second aspect, an embodiment of the present invention further provides a vehicle route planning device based on environment perception, wherein the device includes the following components:

[0027] An image analysis module, used to obtain each road section between the starting point and the end point, and obtain street view images of each road section, and identify physical element information and emotional perception data on the street view images, wherein the physical element information is used to characterize the object information contained in the environment where the road section is located, and the emotional perception data is used to characterize the emotional response that can be induced by the environment where the road section is located;

[0028] A scoring module, used for obtaining a comprehensive score of each road section according to the physical element information and the emotion perception data of each road section;

[0029] The planning module is used to obtain the length of each of the road sections, obtain the number of passengers that need to take the vehicle for each of the road sections, and screen out connected target road sections from each of the road sections based on the length of each of the road sections, the number of passengers for each of the road sections and the comprehensive score of each of the road sections, and plan the driving route of the vehicle from the starting point to the end point based on the path where the target road sections are located.

[0030] In a third aspect, an embodiment of the present invention further provides a terminal device, wherein the terminal device includes a memory, a processor, and a vehicle route planning program based on environment perception stored in the memory and executable on the processor, and when the processor executes the vehicle route planning program based on environment perception, the steps of the vehicle route planning method based on environment perception described above are implemented.

[0031] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, on which a vehicle route planning program based on environment perception is stored. When the vehicle route planning program based on environment perception is executed by a processor, the steps of the above-mentioned vehicle route planning method based on environment perception are implemented.

[0032] Beneficial effects: The present invention first identifies the physical element information and emotional perception data of each road section from the street view image of the road section, and then comprehensively evaluates each road section based on the physical element information and emotional perception data of each road section to obtain a comprehensive score. Finally, based on the comprehensive score, the length of the road section and the number of passengers who need to take the vehicle on each road section, the connected target road sections are intelligently screened from each road section, and the target road sections are used to plan the vehicle's driving route from the starting point to the end point. From the above analysis, it can be seen that when planning the route, the present invention not only considers the physical elements of the road section but also fully considers the emotional notification data of the road section, so that the route planned by the present invention can meet the actual needs of passengers. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is the overall flow chart of the present invention;

[0034] Figure 2 is a schematic diagram of a road section and a route in an embodiment of the present invention;

[0035] Figure 3 A structural diagram of a vehicle route planning device based on environment perception provided by the present invention;

[0036] Figure 4 This is a block diagram of the internal structure principle of the terminal device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0037] The following is a clear and complete description of the technical solution of the present invention in combination with the embodiments and the accompanying drawings. 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.

[0038] Research has found that when a vehicle travels from a starting point to a destination along different routes, the street scenes on different routes will bring different experiences to passengers. The experience is related to the visual quality of the street, and the visual quality of the street is not only reflected in the physical conditions of the street, such as the vegetation coverage on both sides of the road, the breadth of vision, etc., but also includes the emotional perception brought by the street scene: environmental aesthetics, historical and cultural characteristics, etc. These factors together constitute the "perceived quality" of the street, which directly affects the perception and experience of tourists. However, when planning the driving route of the vehicle, the existing technology only considers the physical conditions of the street, but ignores the emotional perception brought to the passengers by the street scene, which makes it difficult for the planned route to meet the actual needs of the passengers.

[0039] In order to solve the above technical problems, the present invention provides a vehicle route planning method, device and equipment based on environmental perception, which solves the problem that the routes planned by the prior art are difficult to meet the actual needs of passengers. In specific implementation, each road section between the starting point and the end point is obtained, and the street view image of each road section is obtained, and the physical element information and emotional perception data on the street view image are identified. The physical element information is used to characterize the object information contained in the environment where the road section is located, and the emotional perception data is used to characterize the emotional response that can be triggered by the environment where the road section is located; based on the physical element information and emotional perception data of each road section, a comprehensive score of each road section is obtained; the length of each road section is obtained, the number of passengers who need to take the vehicle for each road section is obtained, and based on the length of each road section, the number of passengers of each road section and the comprehensive score of each road section, the connected target road section is selected from each road section, and the vehicle's driving route from the starting point to the end point is planned based on the path where the target road section is located.

[0040] For example, Figure 2 As shown, the vehicle needs to travel from the starting point S to the end point T. The route from the starting point S to the end point T includes a route SABT composed of a section SA, a section AB, and a section BT, a route SDEBT composed of a section SD, a section DE, a section EB, and a section BT, a route SEBT composed of a section SE, a section EB, and a section BT, and a route SEFGHT composed of a section SE, a section EF, a section FG, a section GH, and a section HT. Figure 2There are other routes, which are not listed here one by one. The present invention collects street view images of the above 11 road sections, a total of 11 street view images, and obtains physical element information and emotional perception data of each road section according to the street view image of each road section, and then obtains a comprehensive score of each road section according to the physical element information and emotional perception data, and then selects the connected target road section from the above 11 road sections according to the length of each road section, the number of passengers who need to take the vehicle for each road section, and the comprehensive score. Since the connected road section includes a road section group consisting of road section SA, road section AB, and road section BT, the connected road section also includes a road section group consisting of road section SD, road section DE, road section EB, and road section BT, the connected road section also includes a road section group consisting of road section SE, road section EB, and road section BT, and the connected road section also includes a road section group consisting of road section SE, road section EF, road section FG, road section GH, and road section HT, then selecting the connected target road section from the 11 road sections is to select one group of road sections from the above four groups of road sections as the target road section. If the screened target road section is a road section group consisting of road section SA, road section AB, and road section BT, the vehicle is controlled to travel along route SABT.

[0041] The vehicle route planning method based on environment perception of this embodiment can be applied to a terminal device, which can be a terminal product with image acquisition and image processing functions, such as a vehicle system. Figure 1 As shown in , the vehicle route planning method based on environment perception specifically includes the following steps:

[0042] S100, obtaining each road section between a starting point and an end point, and obtaining a street view image of each road section, identifying physical element information and emotion perception data on the street view image, wherein the physical element information is used to characterize object information contained in the environment where the road section is located, and the emotion perception data is used to characterize an emotional response that can be induced by the environment where the road section is located;

[0043] Among them, physical element information and emotion perception data both belong to environmental perception data.

[0044] S200, obtaining a comprehensive score of each of the road sections according to the physical element information and the emotion perception data of each of the road sections;

[0045] S300, obtaining the length of each of the road sections, obtaining the number of passengers that need to take the vehicle for each of the road sections, and selecting a connected target road section from each of the road sections based on the length of each of the road sections, the number of passengers for each of the road sections, and the comprehensive score of each of the road sections, and planning a driving route of the vehicle from the starting point to the end point based on the path where the target road section is located. The vehicle is a tourist bus.

[0046] In the first embodiment, step S100 identifies sky pixels S, vegetation pixels V, and building pixels B on the street view image through a semantic segmentation model (semantic segmentation model DeepLab v3), and counts the number of sky pixels S Num S and the number of vegetation pixels V Num V And the number of building pixels B Num B ; The semantic segmentation model includes a backbone network, a hole convolution, an ASPP module and a decoder connected in sequence.

[0047] Among them, the backbone network is ResNet-101, which is used to extract multi-scale features of street view images. The dilated convolution can expand the perception field of view while retaining high-resolution features, and is used to process street view physical elements of different scales. The ASPP module is Atrous Spatial Pyramid Pooling. The ASPP module uses convolution kernels with multiple dilation rates to perform multi-scale processing on the feature map output by the dilated convolution to generate a comprehensive feature map with rich contextual information. The decoder segments the street view image at the pixel level according to the comprehensive feature map to obtain sky pixels, vegetation pixels, and building pixels. The sky pixels are counted to obtain the number of sky pixels S, the vegetation pixels are counted to obtain the number of vegetation pixels V, and the building pixels are counted to obtain the number of building pixels B.

[0048] The vitality score corresponding to the street view image is determined by an image classification model (the image classification model adopts a ResNet architecture) E , Beauty Score b , Safety Score s , Rich Score W , Boredom score Score Bo and repression score D The image classification model includes a convolutional layer, a residual module, a global average pooling layer, and a fully connected layer connected in sequence.

[0049] Among them, the convolution layer is a 7*7 convolution kernel, and the convolution layer is used to extract the preliminary features of the street view image. The residual module is composed of multiple basic residual modules (Basic Block or Bottleneck Block) stacked together. Each module contains a convolution layer, batch normalization (BatchNorm) and ReLU activation function, and enhances gradient propagation through jump connections. The global average pooling layer (Global Average Pooling) performs global pooling on the feature map and compresses the spatial dimension H*W to 1 dimension. The fully connected layer (Fully Connected Layer) outputs the scores of the six emotional dimensions, that is, the fully connected layer outputs the vitality score ScoreE , Beauty Score b , Safety Score s , Rich Score W , Boredom score Score Bo and repression score D .

[0050] The semantic segmentation model and image classification model of this embodiment are both trained models, wherein the model training steps include acquiring original street view images, preprocessing the original street view images to obtain sample street view images, constructing a physical perception sample data set, constructing an emotion perception sample data set, and training the model.

[0051] Wherein, obtaining the original street view image and preprocessing the original street view image include the following specific steps:

[0052] The road network in the sample area is divided into preset intervals d to ensure that the sampling points can fully cover the key areas and capture the typical characteristics of the road section. For each section of the road network, the starting point coordinates are set to (x0, y0) and the end point coordinates are set to (x n ,y n , then the coordinates of the sampling points along this section can be obtained by interpolation calculation:

[0053]

[0054] Where i is the sampling point number, and n is the total number of sampling points scheduled for the section, which is determined by the section length and d.

[0055] Apply for API usage permission. After the application, you can obtain an API key (API_KEY), that is, you are authorized. Then configure the API request parameters, which include but are not limited to the latitude and longitude coordinates of the target location, the viewing angle and resolution of the required image, etc. Then, use the API to obtain the original local street view images of different angles at each sampling point on the same road section that meet the request parameters. Then, stitch the original local street view images of different angles at each sampling point to obtain the original street view image of the road section. The original street view image is a panoramic image. Stitching requires the use of image processing technology to identify and match the overlapping areas between different images, and automatically stitch them through an algorithm to generate a seamless panoramic view.

[0056] After the original street view image is generated, the original street view image is preprocessed by verification and cleaning to obtain a sample street view image. Afterwards, the sample street view image needs to be stored. This embodiment uses a structured database model to store the sample street view image. When the database model is a non-relational database (such as MongoDB), a unique identifier (ID) needs to be assigned to the sample street view image when storing the sample street view image, and its location information, shooting time and other metadata are stored in related tables to facilitate subsequent query of the sample street view image.

[0057] This embodiment constructs a physical perception sample data set, including the following specific steps:

[0058] The sky pixels, vegetation pixels and building pixels on the sample street view images are marked, and the sky pixels are represented by label 1, the vegetation pixels are represented by label 2, the building pixels are represented by label 3, and the other pixels are represented by label 0 to form a label matrix M. The row and column positions in M ​​represent the positions of the pixels on the sample street view images, and the element values ​​on the row and column positions represent the physical feature types of the sample street view images. The sample street view images and the label matrix M are used as physical perception sample datasets.

[0059] This embodiment trains the semantic segmentation model, including the following specific steps:

[0060] The sample street view image I i And its label matrix M is input into the semantic segmentation model, and the semantic segmentation model outputs the segmentation probability matrix The element value in represents the corresponding I i The probability that a pixel belongs to category k (k∈{1,2,3}) is calculated according to and M to calculate the cross-extraction loss L of the semantic segmentation model seg :

[0061]

[0062] N is I i The total number of pixels is used to calculate the cross-extraction loss L seg Afterwards, according to the cross-extraction loss L seg Adjust the parameters of the semantic segmentation model and then perform the next iterative training until the cross-extraction loss L seg When the set conditions are met or the number of iterations reaches the set number, the training is terminated and the trained semantic segmentation model is obtained.

[0063] This embodiment constructs an emotion perception sample data set, including the following specific steps:

[0064] Mark the vitality score, beauty score, safety score, richness score, boredom score and depression score of the sample street view image to obtain the score label Score′ respectivelyE , Score′ b , Score′ s , Score′ W , Score′ Bo and Score′ D , E stands for vitality, b stands for beauty, s stands for safety, W stands for wealth, Bo stands for boredom, and D stands for depression. Each label is determined by the average score of multi-dimensional reviewers. To ensure the fairness and consistency of the scoring process, multiple rounds of scoring processes may be required, as well as regular review training and review of scoring standards to ensure that each reviewer has a high degree of consistency in understanding and application of the standards. E , Score′ b , Score′ s , Score′ W , Score′ Bo and Score′ D As a sample dataset for emotion perception.

[0065] This embodiment trains the image classification model, including the following specific steps:

[0066] The sample street view image is input into the image classification model, and the training score Score" output by the image classification model is i , according to the score label Score′ i and training score Score″ i Calculate the loss function L emo :

[0067]

[0068] According to the loss function L emo Adjust the parameters of the image classification model and then perform the next iterative training until the loss function L emo When the set conditions are met or the number of iterations reaches the set number, the training is terminated and the trained image classification model is obtained.

[0069] Embodiment 2, based on embodiment 1, when collecting street view images of each road section, the street view images are all accompanied by the geographic coordinates of the shooting location, that is, each street view image has corresponding shooting geographic coordinates. Use GIS data (such as road centerline shapefile or GeoJSON format) to obtain the spatial geometric information of the road section (spatial geometric information includes road geographic coordinates). Use the nearest distance method to project the shooting geographic coordinates of each street view image onto the corresponding road section. If the distance between the shooting geographic coordinates and the road geographic coordinates is less than a threshold, it is considered that the street view image belongs to the road section, where the threshold is 10 meters.

[0070] After the street view images of each road section are collected, the semantic segmentation model and the image classification model trained in Example 1 are applied to the street view images to obtain the physical element information and emotion perception data of each road section.

[0071] Embodiment 3, based on embodiment 1 or embodiment 2, in this embodiment, step S200 includes the following specific steps S201, S202, S203:

[0072] S201, determining positive physical indicators and negative physical indicators in the physical element information, wherein the positive physical indicators are physical indicators that can enhance passenger experience, and the negative physical indicators are physical indicators that reduce passenger experience.

[0073] Physical element information includes the number of sky pixels S Num S and the number of vegetation pixels V V And the number of building pixels B Num B .

[0074] Since buildings and vegetation can increase visual beauty and improve the overall perceived quality, the number of building pixels B is Num B , the number of vegetation pixels V Num V It is a positive physical indicator. Too much sky may mean that the street is empty or lacks coverage, reflecting a low-quality street experience, so the lower the value, the better. Therefore, the number of sky pixels S is Num S It is a negative physical indicator.

[0075] S202, determining a positive emotion perception index and a negative emotion perception index in the emotion perception data, wherein the positive emotion perception index is an emotion index that enhances the passenger experience, and the negative emotion perception index is an emotion index that reduces the passenger experience.

[0076] Emotional perception data includes vitality score Score E , Beauty Score b , Safety Score s , Rich Score W , Boredom score Score Bo and repression score D , Score E 、Score b 、Score s 、Score W 、Score Bo 、Score Bo The value range is 1-10.

[0077] Since the emotional dimensions of vitality, beauty, safety, and richness directly represent the attractiveness, safety, and beauty of the street, and are also the core of tourists' preferences, the higher the value, the better. Therefore, the vitality score is E , Beauty Score b , Safety Score s , Rich Score W It is an indicator of positive emotion perception. Boredom is a negative emotion dimension, the lower the value, the better. Depression is also a negative emotion dimension, the lower the value, the better. Therefore, the boredom score Score is Bo and repression score D An indicator of negative emotion perception.

[0078] S203, obtaining a comprehensive score of each road section according to the positive emotion perception index, the negative emotion perception index, the positive physical index, and the negative physical index of each road section.

[0079] For the same road section, collect several street view images and calculate the number of sky pixels S in each street view image. S , the number of vegetation pixels V Num V 、Num of building pixels B B , Vitality Score E , Beauty Score b , Safety Score s , Rich Score W , Boredom score Score Bo , Suppression score Score D These nine indicators are then used to calculate the mean of these nine indicators using several street view images to obtain the mean number of sky pixel points S on the same road section. The mean number of vegetation pixels V The average number of building pixels B Mean Vitality Beautiful Mean Safety mean Rich in Mean Boredom score average Suppressed Mean Then the above nine means are normalized respectively, and the normalized number of sky pixel points S is χ norm,S , the normalized number of vegetation pixels V χ norm,V , the normalized number of building pixels B norm,B , vitality normalized value χ norm,E , beauty normalization value χ norm,b , safety normalized value χ norm,s, rich normalized value χ norm,W , boring normalized value χ norm,Bo , suppressed normalized value χ norm,D .

[0080] in

[0081] include χ min is the minimum value of the jth index, χ max is the maximum value of the jth index.

[0082] The comprehensive score Q of the kth road segment k :

[0083]

[0084] w j′ is the weight of the j′th positive indicator, χ norm,j′ is the normalized value of the j′th positive indicator, χ norm,j′ The normalized number of vegetation pixels V is x norm,V , the normalized number of building pixels B norm,B , vitality normalized value χ norm,E , beauty normalization value χ norm,b , safety normalized value χ norm,s , rich normalized value χ norm,W ;w j″ is the weight of the j″th negative indicator, χ norm,j″ is the normalized value of the j″th negative indicator, χ norm,j″ Including boring normalized value χ norm,Bo , suppressed normalized value χ norm,D , the normalized number of sky pixels S χ norm,S .

[0085] Embodiment 4, this embodiment is based on embodiment 1 or embodiment 2 or embodiment 3. In this embodiment, step S300 includes the following specific steps S301 to S304:

[0086] S301, dividing each of the road sections into a plurality of groups, each road section in each of the groups having connectivity.

[0087] All road sections covered from the starting point to the end point are divided into several groups, and the path composed of all road sections in each group allows the vehicle to travel from the starting point to the end point.

[0088] S302, establishing a first function for each group according to the number of passengers on the road section in each group; and establishing a first function for each group according to the comprehensive score Q of the road section in each group.k , establishing a second function for each of the groups.

[0089] The first function DS of the mth group m : where d mk is the number of passengers corresponding to the kth road section in the mth group, that is, the number of passengers who need to take the vehicle on the kth road section, S m is the total number of road segments in the mth group. The second function VQ of the mth group m :

[0090] S303, obtaining the capacity limit C of the vehicle (the capacity limit C is the maximum number of people that the vehicle can accommodate within a safe range), and obtaining the straight line length L from the starting point to the end point sp , and based on the capacity limit C and the straight line length L sp , establish constraints for each of the groups.

[0091] S304: Filter out a target group from each of the groups according to the constraint condition of each of the groups and the first function and the second function, and use each of the road sections in the target group as a target road section.

[0092] Filter out the pre-selected groups from all groups. The pre-selected groups meet the constraint condition, that is, the sum of the lengths of all sections in the pre-selected groups must be less than or equal to αL sp , and the sum of the number of passengers corresponding to all sections in the pre-selected group must be less than or equal to the capacity limit C, where α is a constant greater than or equal to 1.

[0093] Then filter out the target group from the pre-selected groups, where the first function corresponding to the target group The second function VQ corresponding to the target group is the maximum value among all groups. m Also obtain the maximum value.

[0094] Embodiment 5. This embodiment is based on embodiment 1 or embodiment 2 or embodiment 3. In this embodiment, step S300 includes the following specific steps: constructing a route optimization model based on the length of each of the road sections, the number of passengers on each of the road sections and the comprehensive score of each of the road sections; applying a non-dominated sorting algorithm to the route optimization model to screen out connected target road sections from each of the road sections.

[0095] The objective function of this embodiment: Max DS = ∑ (i,j)∈A d ij x ij

[0096]

[0097] Constraints:

[0098]

[0099] Where DS is the passenger flow satisfaction, VQI is the visual index value, d ij is the number of passengers who need to take the vehicle on the road section between the i-th station and the j-th station, C is the capacity limit of the vehicle, A is the set of all stations, Q ij x ij is the comprehensive score of the road section between the i-th station and the j-th station, l ij is the length of the road section between the i-th station and the j-th station, α is a constant greater than or equal to 1, L sp is the length of the straight line from the starting point to the end point, x ij is the binary decision variable for the section between the i-th station and the j-th station, x ij The value of is 0 and 1.

[0100] In this embodiment, under the constraint conditions, the corresponding road section is calculated when DS and VQI reach the maximum value, and this road section is the target road section, that is, which x is calculated under the constraint conditions. ij The value 1 can make DS and VQI reach the maximum value. ij The corresponding section is the target section, so solving the target section becomes solving the x that maximizes DS and VQI under the constraints. ij The value of .

[0101] This example uses NSGA-II to solve x ij The value of includes the following specific steps:

[0102] Initialization: Randomly generate a population, where individuals represent vehicle route designs. The stations that the route passes through are represented as a sequence of station numbers (such as [A, B, C, D]). Randomly generate a population with a population size of N (usually 50-200).

[0103] Fitness calculation: Calculate the target value of each individual based on DS and VQI.

[0104] Non-dominated sorting: stratify by dominance relationship and calculate crowding distance. Individual A dominates individual B if A is not worse than B in all objectives (DS A ≥DS B ,VQI A ≥VQI B ). A is better than B in at least one objective. Crowding distance calculation, for each individual in each layer, the crowding is calculated according to the distribution of the target value:

[0105] Crowding Distance=∑ k (f k+1 -f k-1 ) / (f max -f min )

[0106] f k+1 is the target value of the next individual of the current individual i on target k. k-1 is the target value of the previous individual of the current individual i on target k. max and f min They represent the maximum and minimum values ​​of all individuals in the population on target k. Individuals with higher crowding degree are retained first to maintain the diversity of the population.

[0107] Genetic operations: Generate the next generation through selection, crossover and mutation. Selection through tournament: Compare the non-dominated layers of two individuals and give priority to the individual with lower layer (better solution). Randomly select two parents and swap part of the site sequence to ensure the generated offspring sequence is legal. Mutation site sequence mutation: Randomly swap two sites in the site sequence.

[0108] Stop condition: reaching the maximum number of generations or Pareto frontier convergence, then x is obtained ij The value of .

[0109] This embodiment also uses historical passenger flow data to identify high-demand routes, and combines the distribution of tourist hotspots to select routes that can cover important attractions and carry large passenger flows. After the routes to be optimized are identified, the above multi-objective optimization algorithm is used to optimize these routes. Finally, the results need to be evaluated to verify whether the effect meets expectations. The optimization effect can be quantified by comparing the changes in passenger flow satisfaction and passenger experience (such as visual quality) before and after optimization. In addition, if the optimization effect of certain routes does not meet the standards, it is necessary to further analyze the reasons, and adjust and iterate the optimization for the problem links.

[0110] In summary, the present invention comprehensively analyzes the quality of street environment by integrating a two-layer evaluation framework of physical perception and emotional perception. At the physical perception level, large-scale street view images are used to extract key physical features of the street, such as vegetation coverage, sky index, etc., based on computer vision technology. At the emotional perception level, an emotional perception system is constructed based on a deep learning model. By classifying and scoring the emotional features of street view images, the potential impact of the street environment on the user's psychology and behavior is explored. The street environment includes relaxed and depressing, thereby realizing a comprehensive evaluation of the physical and emotional dimensions. At the same time, the present invention proposes a solution based on a multi-objective heuristic optimization algorithm in vehicle route optimization, which effectively takes into account the multi-dimensional requirements in street visual quality evaluation. By introducing advanced optimization methods such as non-dominated sorting genetic algorithm (NSGA-II), joint optimization is performed for bus efficiency and street visual indicators to balance the requirements of each dimension and ensure the robustness and applicability of the model in practical applications. The present invention has achieved breakthroughs in theory and application in the fields of street visual quality evaluation, urban traffic optimization, and public space design.

[0111] This embodiment also provides a vehicle route planning device based on environment perception, such as Figure 3 As shown, the device comprises the following components:

[0112] The image analysis module 01 is used to obtain each road section between the starting point and the end point, and obtain the street view image of each road section, and identify the physical element information and emotional perception data on the street view image, wherein the physical element information is used to characterize the object information contained in the environment where the road section is located, and the emotional perception data is used to characterize the emotional response that can be induced by the environment where the road section is located;

[0113] Scoring module 02, used to obtain a comprehensive score of each road section based on the physical element information and the emotion perception data of each road section;

[0114] The planning module 03 is used to obtain the length of each of the road sections, obtain the number of passengers that need to take the vehicle for each of the road sections, and screen out connected target sections from each of the road sections based on the length of each of the road sections, the number of passengers for each of the road sections and the comprehensive score of each of the road sections, and plan the driving route of the vehicle from the starting point to the end point based on the path where the target road sections are located.

[0115] Based on the above embodiments, the present invention further provides a terminal device, whose principle block diagram can be shown as follows: Figure 4As shown. The terminal device includes a processor, a memory, a network interface, and a display screen connected via a system bus. Among them, the processor of the terminal device is used to provide computing and control capabilities. The memory of the terminal device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the terminal device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a vehicle route planning method based on environmental perception is implemented. The display screen of the terminal device can be a liquid crystal display screen or an electronic ink display screen.

[0116] Those skilled in the art will understand that Figure 4 The principle block diagram shown in the figure is only a block diagram of a partial structure related to the scheme of the present invention, and does not constitute a limitation on the terminal device to which the scheme of the present invention is applied. The specific terminal device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0117] In one embodiment, a terminal device is provided, the terminal device comprising a memory, a processor, and a vehicle route planning program based on environment perception stored in the memory and executable on the processor, and when the processor executes the vehicle route planning program based on environment perception, the following operation instructions are implemented:

[0118] Acquire each road section between the starting point and the end point, and acquire a street view image of each road section, and identify physical element information and emotional perception data on the street view image, wherein the physical element information is used to characterize object information contained in the environment where the road section is located, and the emotional perception data is used to characterize emotional responses that can be induced by the environment where the road section is located;

[0119] Obtaining a comprehensive score for each of the road sections based on the physical element information and the emotion perception data of each of the road sections;

[0120] The length of each of the road sections is obtained, the number of passengers that need to take the vehicle for each of the road sections is obtained, and based on the length of each of the road sections, the number of passengers for each of the road sections and the comprehensive score of each of the road sections, connected target road sections are screened out from each of the road sections, and the driving route of the vehicle from the starting point to the end point is planned based on the path where the target road sections are located.

[0121] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0122] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A vehicle route planning method based on environment perception, characterized in that: include: Acquire each road section between the starting point and the end point, and acquire a street view image of each road section, and identify physical element information and emotional perception data on the street view image, wherein the physical element information is used to characterize object information contained in the environment where the road section is located, and the emotional perception data is used to characterize emotional responses that can be induced by the environment where the road section is located; Obtaining a comprehensive score for each of the road sections based on the physical element information and the emotion perception data of each of the road sections; The length of each of the road sections is obtained, the number of passengers that need to take the vehicle for each of the road sections is obtained, and based on the length of each of the road sections, the number of passengers for each of the road sections and the comprehensive score of each of the road sections, connected target road sections are screened out from each of the road sections, and the driving route of the vehicle from the starting point to the end point is planned based on the path where the target road sections are located.

2. The vehicle route planning method based on environment perception as claimed in claim 1, characterized in that: The sky pixels, vegetation pixels and building pixels on the street view image are identified by a semantic segmentation model, and the number of the sky pixels, the number of the vegetation pixels and the number of the building pixels are used as the physical element information; the semantic segmentation model includes a backbone network, a hole convolution, an ASPP module and a decoder connected in sequence.

3. The vehicle route planning method based on environment perception as claimed in claim 1, characterized in that: The vitality score, beauty score, safety score, richness score, boredom score and depression score corresponding to the street view image are determined by an image classification model, and the vitality score, beauty score, safety score, richness score, boredom score and depression score are used as the emotion perception data; the image classification model includes a convolutional layer, a residual module, a global average pooling layer and a fully connected layer connected in sequence.

4. The vehicle route planning method based on environment perception as claimed in claim 1, characterized in that: The step of obtaining a comprehensive score of each road section based on the physical element information and the emotion perception data of each road section includes: Determining positive physical indicators and negative physical indicators in the physical element information, wherein the positive physical indicators are physical indicators that can enhance passenger experience, and the negative physical indicators are physical indicators that reduce passenger experience; Determining a positive emotion perception index and a negative emotion perception index in the emotion perception data, wherein the positive emotion perception index is an emotion index that enhances the passenger experience, and the negative emotion perception index is an emotion index that reduces the passenger experience; A comprehensive score of each road section is obtained according to the positive emotion perception index, the negative emotion perception index, the positive physical index, and the negative physical index of each road section.

5. The vehicle route planning method based on environment perception according to any one of claims 1 to 4, characterized in that: The distance between the shooting geographic coordinates of the street view image and the road geographic coordinates of the road segment is less than a threshold.

6. The vehicle route planning method based on environment perception as claimed in claim 1, characterized in that: The step of selecting a connected target road section from each road section according to the length of each road section, the number of passengers in each road section and the comprehensive score of each road section comprises: Dividing each of the road sections into a plurality of groups, each road section in each of the groups having connectivity; According to the number of passengers on the road section in each group, a first function is established for each group; according to the comprehensive score of the road section in each group, a second function is established for each group; Obtaining the capacity limit of the vehicle, obtaining the straight line length from the starting point to the end point, and establishing a constraint condition for each of the groups according to the capacity limit and the straight line length; According to the constraint condition of each of the groups and the first function and the second function, a target group is screened out from each of the groups, and each of the road sections in the target group is used as a target road section.

7. The vehicle route planning method based on environment perception as claimed in claim 1, characterized in that: The step of selecting a connected target road section from each road section according to the length of each road section, the number of passengers in each road section and the comprehensive score of each road section comprises: Constructing a route optimization model according to the length of each of the road sections, the number of passengers on each of the road sections, and the comprehensive score of each of the road sections; A non-dominated sorting algorithm is applied to the route optimization model to select connected target road segments from each of the road segments.

8. A vehicle route planning device based on environment perception, characterized in that: The device comprises the following components: An image analysis module, used to obtain each road section between the starting point and the end point, and obtain street view images of each road section, and identify physical element information and emotional perception data on the street view images, wherein the physical element information is used to characterize the object information contained in the environment where the road section is located, and the emotional perception data is used to characterize the emotional response that can be induced by the environment where the road section is located; A scoring module, used for obtaining a comprehensive score of each road section according to the physical element information and the emotion perception data of each road section; The planning module is used to obtain the length of each of the road sections, obtain the number of passengers that need to take the vehicle for each of the road sections, and screen out connected target road sections from each of the road sections based on the length of each of the road sections, the number of passengers for each of the road sections and the comprehensive score of each of the road sections, and plan the driving route of the vehicle from the starting point to the end point based on the path where the target road sections are located.

9. A terminal device, characterized in that: The terminal device includes a memory, a processor, and a vehicle route planning program based on environment perception stored in the memory and executable on the processor. When the processor executes the vehicle route planning program based on environment perception, the steps of the vehicle route planning method based on environment perception as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a vehicle route planning program based on environment perception. When the vehicle route planning program based on environment perception is executed by the processor, the steps of the vehicle route planning method based on environment perception as described in any one of claims 1 to 7 are implemented.

Images