Attribute map sketch search method based on spatial locality principle

Abstract

The invention discloses an attribute map sketch search method based on a spatial locality principle, which is used for generating a map mainly comprising point elements by processing basic data. The target of the method is that after a user inputs a sketch, an area similar to the sketch is searched on a map according to the content of the sketch and the characteristic that the ground object distance is close when the sketch is drawn. And finally, sorting the search results according to the space size of the returned region, and returning the results to the user.

Description

Technical Field

[0001] The present invention belongs to the field of data search, and particularly relates to a map search method for sketch search. Background Art

[0002] Maps are of indispensable importance in people's lives. It is not only a tool for us to explore the world, but also the key to helping us understand and plan space. From navigation and travel to urban planning, maps enable us to accurately locate, avoid getting lost, and grasp the overall space. Whether it is finding a destination during a trip or choosing the best route in daily life, maps invisibly assist our decision-making. With the rapid development of artificial intelligence, the Internet of Things, and big data technologies, map applications will further integrate virtual reality, real-time dynamic data, and intelligent recommendation functions to provide users with more accurate and convenient services.

[0003] In the future, map applications will not only be limited to helping users "find the way", but also play a crucial role in more fields. For example, by combining with autonomous driving technology, maps will become the core support for the intelligent operation of vehicles; in urban planning, map data will provide important basis for reasonable layout and resource allocation; in the era of the metaverse, maps may also become the bridge connecting the virtual and the real, creating new interactive experiences.

[0004] It can be foreseen that map applications will not only be tools, but also become the key driving force for the digital transformation and intelligent development of society, profoundly changing the way of human life.

[0005] Currently, mainstream map software such as Google Maps, Amap, and Apple Maps has become an indispensable tool in people's daily lives. They not only provide detailed global geographical information, but also integrate functions such as real-time traffic conditions, navigation guidance, public transportation routes, and point-of-interest recommendations. Users can easily find the optimal path, avoid congestion, and obtain detailed information about restaurants, hotels, scenic spots, etc. through these software. Some map software also supports offline use and voice navigation, further enhancing convenience and improving the travel experience of users. Whether it is self-driving travel, walking navigation, or taking public transportation, these map software provide people with efficient, accurate, and convenient services.

[0006] However, these software mainly rely on keyword search to help users find their destinations. This approach may cause certain difficulties when users have unclear or vague descriptions of their destinations. When users are unable to accurately provide the location name or address, the search results of map software are often not precise enough, and multiple attempts may be required to find the correct destination. In addition, in the face of vague location descriptions, such as only remembering a certain landmark or the surrounding environment, map software may not be able to provide intuitive matching results, limiting the user experience. This makes it difficult for users to quickly and effectively find the places they want in some cases.

[0007] In the face of the above problems, the present invention proposes a spatial search method based on map sketches, aiming to solve the deficiencies of traditional map software in dealing with fuzzy destination searches. This map sketch search method allows users to find areas that meet specific requirements by drawing simple schematic diagrams, thus effectively avoiding the dependence on precise input in traditional keyword searches. This method not only improves the flexibility and fault tolerance of the search, but also provides a more intuitive interaction method for users, enabling them to quickly find the required location even when the destination is uncertain. Summary of the Invention

[0008] The present invention proposes an attribute map sketch search method based on the principle of spatial locality. By processing the basic data, a map mainly composed of point features is generated. The goal of this method is that after the user inputs a sketch, according to the content of the sketch and the characteristics of the proximity of features when the sketch is drawn, similar areas are searched on the map. Finally, the search results are sorted according to the spatial size of the returned areas and the results are returned to the user.

[0009] The basic principle of the method of the present invention is as follows: The traditional map is simply processed into an attribute map containing only point features, and each point feature contains type, coordinates and detailed information. Then the input sketch of the user is processed to obtain the relative position relationship between the features in the sketch. Subsequently, the principle of spatial locality is used to determine the corresponding sets of any two points in the map for the sketch, and then all areas that meet the requirements are obtained under the condition of determining the two points. Finally, the returned set is sorted according to the area of the returned area according to the locality principle.

[0010] An attribute map sketch search method based on the principle of spatial locality includes the following steps:

[0011] Step 1: Map data abstraction and preprocessing

[0012] Step 11, abstraction of point features;

[0013] The features on the traditional map are abstracted into point features, and each point feature contains the following attributes:

[0014] Type t: The category of ground features (such as buildings, road nodes, etc.).

[0015] Coordinate (x, y): The spatial position of point features.

[0016] Detailed information: Other ground feature attributes (such as names).

[0017] Step 12, ground feature statistics; Count the number N of ground features of each type t , and record the relative position relationship (angle and distance) between any two points for subsequent search calculations.

[0018] Step 13, two-point relationship calculation; For any two points p i =(x i ,y i ) and p j =(x j ,y j ), calculate the angle and distance between them:

[0019]

[0020] Step 2: Sketch parsing and edge set construction;

[0021] Step 21, sketch point set extraction;

[0022] The user inputs a sketch and extracts the type t and coordinates (x, y) of each point feature from it. According to the number N of ground features counted in the map t , sort the points in the sketch in ascending order according to the number of types to generate a point set P. The main purpose of sorting the points is to preferentially determine the points with fewer types during the subsequent search process, so as to first lock two initial matching points and then gradually determine the matching of the remaining points based on these two points. In this process, selecting the points with a smaller candidate set as the preferred matching objects can significantly improve the search efficiency. From a probabilistic perspective, preferentially matching the points with a smaller candidate set will reduce the number of point pair combinations that meet the matching conditions, thereby effectively narrowing the search space. In addition, this search strategy is similar to depth-first search. By preferentially matching the nodes with a smaller candidate set, the search space shows the characteristics of an inverted triangle. This characteristic not only helps to reduce unnecessary search paths but also better adapts to various pruning strategies, further improving the performance and efficiency of the algorithm.

[0023] Step 22, edge set construction;

[0024] Calculate the angle and distance relationships between the first point p 1 in the sorting and the remaining points p i to construct an edge set E:

[0025] E = {e 1 ,e2 ,…,e k},e i =(p 1 ,p i ,d(p 1 ,p i ),θ(p 1 ,p i ))

[0026] To reduce the influence of scale error magnification, the longest side e max in the edge set is placed at the first position, and the remaining edges keep their order unchanged.

[0027] Step 23, error magnification effect;

[0028] Assume that the length of edge e 1 is d(e 1 ), and the length of another edge e 2 is d(e 2 ), and the two satisfy d(e 2 ) = N·d(e 1 ). If the drawing error of e 1 is ∈, then the error of e 2 is:

[0029]

[0030] When N > 1, the error will be magnified, thus affecting the accuracy of the matching result. Therefore, placing the longest edge e max at the first position, the ratio of the remaining edges to it will be less than 1, which can reduce the influence of the error on subsequent matching.

[0031] Step 3: Search and matching based on the principle of spatial locality;

[0032] Step 31, determination of the initial matching point. Enumerate all points p' 1 in the map that have the same type as the first point p 1 in the point set P. Based on the principle of spatial locality, with p' 1 as the center, find the point p' 1 that is the closest to it and has the same type as the other endpoint of the first edge e 2 in the edge set.

[0033] Step 32, derivation of the new edge set; Calculate the actual angles and distances in the map according to the positional relationship between p' 1 and p' 2 to generate a new edge set E'.

[0034] Step 33, gradually expand the matching points; Use E' and the already matched point pair (p' 1 ,p' 2), continue to search for other point features in the map until the matching of the entire point set is completed. During the matching process, the conditions are appropriately relaxed to tolerate sketch errors.

[0035] Step 34, generate a result set; each complete match generates a result r, and finally the result set R = {r 1 , r 2 , …, r m} is obtained.

[0036] Step 4: Result sorting and optimization

[0037] Step 41, area and ratio relationship; according to the principle of spatial locality, results with smaller regional areas usually have higher matching accuracies. Given that the sketch shape is fixed, the area of the matching result r is proportional to the side length ratio ρ:

[0038]

[0039] where d(e' i ) is the side length of the sketch, and d(e' i ) is the side length in the matching result.

[0040] Step 42, sorting rule; sort the result set R in ascending order of the ratio ρ to ensure that the matching results with higher accuracies are returned first.

[0041] Step 43, final return; return the sorted result set R for the user to view or further operate. The time complexity of this search algorithm is O(C(p 1 ) × (K - 1)). Where K is the number of points in the sketch, and C(p 1 ) is the number of points of the corresponding type of the first point p 1 in the map point set P.

[0042] The present invention has the following obvious innovations and prominent advantages: The present invention proposes an innovative map search method, which makes full use of the powerful expression ability of the sketch to make up for the limitations of traditional keyword map search. During the search and matching process, points with fewer type numbers are locked first, thus significantly accelerating the search process. At the same time, on the basis of preferentially searching for points with fewer type numbers, the position of the longest side is determined first, effectively solving the problem of error magnification caused by sketch drawing. In addition, by means of the principle of spatial locality, by preferentially determining the positions of two points, the search difficulty caused by the user's inability to clarify the due north direction when drawing the sketch is overcome. Finally, based on the principle of spatial locality, the search result set is sorted to ensure that the returned results are more accurate and meet the actual requirements. Description of the Drawings

[0043] Figure 1 This is a display of the number of map data types provided by the present invention.

[0044] Figure 2 Flowchart of the attribute map sketch search method based on the principle of spatial locality provided by the present invention.

[0045] Figure 3 An application example demonstration of the present invention. Detailed implementation manners

[0046] The present invention will be explained and elaborated below in conjunction with relevant attached drawings:

[0047] The map data set adopted by the present invention mainly comes from the data of the Fifth Ring Road in Beijing of OpenStreetMap (including 32,170 points). The points in the data set contain detailed longitude and latitude as well as type information. The types of the map data points are divided into 17 categories in total, such as Figure 1 shown. The flowchart of the attribute map sketch search method based on the principle of spatial locality is as Figure 2 shown, and it is characterized by including the following steps.

[0048] Step 1: Map data preprocessing and feature quantization. Spatial features are extracted from the original map data set, and a topological relationship matrix is constructed by calculating the Euclidean distance and azimuth angle between each node. To improve the robustness of the algorithm, a discretized binning technique is used to normalize continuous spatial parameters: the distance quantization unit is set to 50 meters (using the rounding-up strategy, such as 62.4 meters → 2 units), and the angle quantization unit is set to 5° (such as 23° → 5 units). This feature quantization mechanism effectively absorbs the drawing errors of hand-drawn sketches (±25-meter distance error, ±2.5° angle deviation) by constructing a tolerance interval, reduces the sensitivity of feature matching while maintaining the spatial topological features, and lays a foundation for subsequent elastic matching.

[0049] Step 2: Sketch feature parsing and anchor point optimization. A feature parsing engine based on graph theory is used to structurally process the user-input sketch. First, semantic-spatial dual feature analysis is performed: by statistically analyzing the distribution density of each node type, the feature point with the highest rarity of semantic categories is selected as the anchor point (when there are multiple instances of the same type of feature point, the first-encounter priority strategy is adopted). Taking the anchor point as the center, a spatial relationship feature vector between this feature point and adjacent nodes is constructed, including a set of normalized distance-azimuth angle binary group parameters. To improve the retrieval efficiency, feature vector sorting optimization is implemented: using the longest spatial baseline as the main feature vector direction, and the remaining associated edges are organized according to the ascending order arrangement strategy based on the semantic richness of the associated nodes. This mechanism of composite sorting based on semantic weights and spatial baselines not only ensures the priority matching degree of key spatial relationships, but also effectively reduces the search space by establishing multi-dimensional constraint conditions, significantly improving the operation efficiency and result accuracy in the subsequent graph matching stage.

[0050] Step 3: The spatial topology matching engine based on the reference edge establishes the mapping relationship between the sketch and the geographical database through a spatial-semantic dual-constraint mechanism. First, retrieve the candidate anchor point set that matches the start point type of the sketch reference edge in the map dataset, and find the point closest to each anchor point that has the same type as the other endpoint based on the principle of spatial locality to form a matching pair. Using the successfully matched reference edge as the spatial reference system, establish an affine transformation model to deduce the theoretical coordinate distribution interval of the associated nodes, and perform multi-dimensional verification through the constraint propagation mechanism: synchronously verify during the breadth-first traversal, the discretized angle binning rule, and the semantic compatibility of the node type combinations, and combine the dynamic pruning algorithm to eliminate the candidate solutions with topological conflicts in real time. Finally, the matching instances that pass the full constraint satisfaction detection form the result set R. This mechanism realizes the efficient sublinear search of large-scale spatial databases through the synergistic effect of reference edge space locking and multi-constraint joint reasoning.

[0051] Step 4: Spatial clustering optimization and result ranking. Based on the cognitive model of spatial aggregation effect, re-rank the priority of the candidate result set and establish a sorting mechanism driven by geometric similarity. According to the spatial proximity characteristics of sketch drawing (users usually draw geographical elements with high spatial coupling first), take the geometric scale of the matching result as the core sorting index: through the positive correlation between the reference edge length and the matching area (area ∝ reference edge length^2), design an ascending sorting strategy based on the spatial baseline of the reference edge. This algorithm directly extracts the quantified length of the successfully matched reference edge as the proxy variable of the area scale, avoiding the performance loss caused by complex polygon area calculation, and at the same time ensuring that the sorting result is consistent with the scale perception characteristics of human spatial cognition (that is, users are more likely to accept small-scale matching areas as the primary candidates). The finally output ordered result set R significantly improves the visual interpretability of the matching result and the system interaction efficiency through the dual mapping of geometric scale and cognitive priority.

[0052] The following gives an example of using the present invention for sketch search, as Figure 3 shown. Among them, Canvas is the sketch, Map is the map, and the two Candidates are the two returned results.

[0053] The present invention proposes an innovative map search method, which makes full use of the powerful expression ability of sketches and makes up for the limitations of traditional keyword map search. It will play a certain role in fields such as criminal investigation, revisiting old places by memory, and urban and rural planning in the future.

Claims

1. A method for searching attribute map sketches based on the principle of spatial locality, characterized in that: By processing the basic data, a map with point features as the main feature is generated. After the user inputs a sketch, similar areas are searched on the map based on the content of the sketch and the characteristics of the proximity of the objects when the sketch was drawn. Finally, the search results are sorted by the spatial size of the returned area and the results are returned to the user. The traditional map is processed into an attribute map containing only point features, each of which contains type, coordinates and detailed information; the user's input sketch is then processed to obtain the relative position relationship between the objects in the sketch; the principle of spatial locality is then used to determine the corresponding set of any two points in the sketch in the map, and then all areas that meet the requirements are obtained when the two points are determined; The returned set is sorted by the area of ​​the returned region according to the locality principle.

2. The attribute map sketch search method based on the spatial locality principle according to claim 1, characterized in that: A method for searching attribute map sketches based on the principle of spatial locality. Includes the following steps: Step 1: Map data abstraction and preprocessing Step 11, abstraction of point elements; The objects in traditional maps are abstracted into point features, each of which contains the following attributes: Type t: the type of the feature, including buildings and road nodes; Coordinates (x, y): spatial position of a point feature; Detailed information: other features attributes; Step 12: Count the number of objects of each type N t , and record the relative position relationship between any two points, that is, angle and distance, to facilitate search and calculation; Step 13, two-point relationship calculation; for any two points p i =(x i ,y i ) and p j =(x j ,y j ), calculate the angle and distance between them: distance angle Step 2: Sketch parsing and edge set construction; Step 21, sketch point set extraction; The user inputs a sketch, from which the type t and coordinates (x, y) of each point feature are extracted; according to the number of features N counted in the map t , sort the points in the sketch in ascending order according to the number of types to generate a point set P; the purpose of sorting the points is to prioritize the points with fewer types in the subsequent search process, so as to first lock the two initial matching points, and then gradually determine the matching of the remaining points based on these two points; selecting points with smaller candidate sets as priority matching objects can significantly improve the search efficiency; Step 22, edge set construction; Calculate the first point p1 and the remaining points p in the sort i The angle and distance relationship is constructed to construct the edge set E: E=(e1,e2,...,e k },and i (p1,p i ,d(p1,p i ),θ(p1,p i )) In order to reduce the impact of scale error amplification, the longest edge e in the edge set is max Put it in the first place, and keep the order of the other edges unchanged; Step 23, error amplification effect; Assume that the length of edge e1 is d(e1), the length of another edge e2 is d(e2), and both satisfy d(e)2)=N·d(e1); if the drawing error of e1 is ∈, then the error of e2 is: When N>1, the error will be magnified, thus affecting the accuracy of the matching result; max Put it in the first place, and the ratio of the remaining edges to it will be less than 1, reducing the impact of the error on subsequent matching; Step 3: Search and match based on the principle of spatial locality; Step 31, determine the initial matching point; enumerate all points p'1 in the map that are of the same type as the first point p1 in the point set P; based on the principle of spatial locality, with p'1 as the center, find the point p'2 that is closest to it and has the same type as the other end point of the first edge e1 of the edge set; Step 32, deriving a new edge set: according to the positional relationship between p'1 and p'2, calculating the angle and distance in the actual map, and generating a new edge set E'; Step 33, gradually expand the matching points; use E' and the matched point pair (p'1, p'2) to continue searching for other point features in the map until the entire point set is matched; during the matching process, appropriately relax the conditions to tolerate sketch errors; Step 34, generating a result set; Each complete match generates a result r, and finally the result set R = {r1, r2, ..., r m }; Step 4: Sorting and optimizing results Step 41, area and ratio relationship; according to the principle of spatial locality, the result with a smaller area usually has a higher matching accuracy; it is known that the shape of the sketch is fixed, and the area of ​​the matching result r is proportional to the edge length ratio ρ: Where d(e' i ) is the sketch side length, d(e' i ) is the edge length in the matching result; Step 42, sorting rules: sort the result set R from small to large according to the ratio ρ, and return the matching results with priority to ensure their accuracy; Step 43, finally return; return the sorted result set R for the user to view or further operate; the time complexity of this search algorithm is O(C(p1)×(K-1)); where K is the number of points in the sketch, and C(p1) is the number of points of the corresponding type in the map representing the first point p1 in the sketch point set P.

Images