CN-121994205-A - Multi-agent positioning and graph optimizing method, system, terminal and storage medium based on multi-feature matching

CN121994205ACN 121994205 ACN121994205 ACN 121994205ACN-121994205-A

Abstract

The invention relates to the technical field of information matching and positioning, and discloses a multi-agent positioning and graph optimizing method, a system, a terminal and a storage medium based on multi-feature matching, wherein the method comprises the steps of obtaining a plurality of groups of environment images and constructing corresponding text key frames; the method comprises the steps of calculating text similarity of a plurality of positions to screen candidate matching pairs, further calculating MAC address similarity and signal strength similarity to screen a plurality of true matching position pairs, and therefore updating the pose graph of the intelligent agent. The invention carries out multi-round matching on the simulation data and the real data, and identifies the shared area in a large, complex and repeated indoor environment, thereby realizing reliable map fusion and globally consistent collaborative map construction.

Inventors

CHEN SHOUBIN
ZHANG BAIYANG
LI CHUNYU
LI QINGQUAN

Assignees

人工智能与数字经济广东省实验室(深圳)

Dates

Publication Date: 20260508
Application Date: 20251203

Claims (10)

1. The multi-agent positioning and graph optimizing method based on multi-feature matching is characterized by comprising the following steps of: Acquiring a plurality of groups of environment images shot by a plurality of agents respectively, and constructing a plurality of text key frames corresponding to each agent according to each group of environment images; Calculating a plurality of text similarities of all the agents at a plurality of positions, and constructing a plurality of candidate matching pairs according to each text similarity; Acquiring Wi-Fi fingerprints of each candidate position in each candidate matching pair, and calculating the MAC address similarity between all the candidate positions according to all the Wi-Fi fingerprints; calculating signal strength similarity among a plurality of candidate positions according to all the MAC address similarity, and determining a plurality of true matching position pairs according to all the signal strength similarity; If the two positions of the true matching position pair come from the same agent, optimizing the pose graphs of the agents, and if the two positions of the true matching position pair come from different agents, fusing the pose graphs of the two agents.
2. The multi-agent positioning and image optimizing method based on multi-feature matching according to claim 1, wherein the obtaining a plurality of groups of environment images respectively shot by a plurality of agents, and constructing a plurality of text key frames corresponding to each agent according to each group of environment images, further comprises: Each intelligent body acquires a corresponding motion track and a plurality of environment images by utilizing a radar and a measuring unit of the intelligent body; And constructing a corresponding local pose graph according to each motion trail.
3. The multi-agent positioning and image optimizing method based on multi-feature matching according to claim 1, wherein the obtaining a plurality of groups of environment images respectively shot by a plurality of agents, and constructing a plurality of text key frames corresponding to each agent according to each group of environment images, specifically comprises: acquiring each group of environment images of each intelligent agent, and identifying each group of environment images to obtain text features detected by each intelligent agent and pose features when each text feature is recorded; And constructing a plurality of text key frames of each intelligent agent according to each text feature and the corresponding pose feature.
4. The multi-feature matching-based multi-agent localization and mapping optimization method of claim 3, wherein the calculating a plurality of text similarities for all agents at a plurality of locations and constructing a plurality of candidate matching pairs according to each of the text similarities specifically comprises: Calculating similarities between the text features of all the agents at all locations: ; Wherein, the And Respectively represent the first Position and the first The text characteristics of the individual locations are used, Representation of And The similarity between the two is set to be similar, And Respectively represent And Is provided with a die for the mold, Representing the presentation to be Conversion to The minimum number of edits required; and if the similarity is not smaller than a first preset value, defining the corresponding two positions as candidate matching pairs until all the candidate matching pairs are obtained.
5. The multi-agent localization and mapping optimization method based on multi-feature matching according to claim 1, wherein the acquiring Wi-Fi fingerprints of each candidate location in each candidate matching pair, and calculating the MAC address similarity between all the candidate locations according to all the Wi-Fi fingerprints, specifically comprises: Scanning Wi-Fi signals at all candidate positions to obtain the MAC address and a plurality of corresponding signal intensities of each candidate position; performing standardization processing on all the signal intensities corresponding to each MAC address to obtain average signal intensities, and constructing Wi-Fi fingerprints of each candidate position according to each average signal intensity and the corresponding MAC address: ; Wherein, the Represent the first Wi-Fi fingerprints of the candidate locations, And MAC addresses respectively representing the 1 st and 2 nd candidate positions, And Average signal strengths of the 1 st and 2 nd candidate positions are respectively represented; Calculating the same MAC address of the MAC address between each group of candidate matching pairs, and calculating the MAC address similarity between each group of candidate matching pairs according to the same MAC address: ; Wherein, the Represent the first And The MAC address similarity of the candidate locations, Representation of And Is used for the transmission of data to the mobile station, And Respectively represent the first And MAC addresses of the candidate locations.
6. The multi-agent localization and mapping optimization method based on multi-feature matching of claim 5, wherein the calculating the signal strength similarity between the candidate locations according to all the MAC address similarities, and determining the true matching location pairs according to all the signal strength similarities, specifically comprises: Screening all target candidate matching pairs with the MAC address similarity not smaller than a second preset value from all the candidate matching pairs; Calculating the Euclidean distance of each target candidate matching pair at the corresponding same MAC address: ; Wherein, the Represent the first And The euclidean distance of the individual target candidate locations, The index variable is represented by a number of indices, Represent the first And The same MAC address of the individual destination candidate locations, And Respectively represent the first And Signal strength of the target candidate locations; and determining the signal intensity similarity of each target candidate matching pair of the signal according to each Euclidean distance, and defining all target candidate matching pairs with the signal intensity similarity not smaller than a third preset value as true matching position pairs.
7. The multi-agent localization and mapping optimization method based on multi-feature matching of claim 1, wherein if the two positions of the true matching position pair are from the same agent, the pose map of the agent is optimized, and if there are two positions of the true matching position pair from different agents, the pose maps of the two agents are fused, specifically comprising: Judging whether two positions of each true matching position pair are derived from the same intelligent agent; If yes, loop constraint is added in the pose graph of the intelligent body so as to globally optimize the pose graph of the intelligent body and obtain a target pose graph; If not, acquiring a plurality of point clouds of the two positions, constructing transformation matrixes between the two pose graphs of the two intelligent bodies according to all the point clouds, adding the transformation matrixes into the corresponding two pose graphs, and unifying the two pose graphs into a world coordinate system to obtain an aggregate pose graph.
8. A multi-feature-matching-based multi-agent localization and mapping optimization system, wherein the multi-feature-matching-based multi-agent localization and mapping optimization system is applied to the multi-feature-matching-based multi-agent localization and mapping optimization method as set forth in any one of claims 1 to 7, and the multi-feature-matching-based multi-agent localization and mapping optimization system includes: the image acquisition module is used for acquiring a plurality of groups of environment images shot by a plurality of intelligent agents respectively, and constructing a plurality of text key frames corresponding to each intelligent agent according to each group of environment images; The first position matching module is used for calculating a plurality of text similarities of all the agents at a plurality of positions and constructing a plurality of candidate matching pairs according to each text similarity; The second position matching module is used for acquiring Wi-Fi fingerprints of each candidate position in each candidate matching pair, and calculating the MAC address similarity between all the candidate positions according to all the Wi-Fi fingerprints; the third position matching module is used for calculating signal strength similarity among a plurality of candidate positions according to all the MAC address similarity and determining a plurality of true matching position pairs according to all the signal strength similarity; And the pose graph updating module is used for optimizing the pose graphs of the intelligent agents if the two positions of the true matching position pair come from the same intelligent agent, and fusing the pose graphs of the two intelligent agents if the two positions of the true matching position pair come from different intelligent agents.
9. A terminal comprising a memory, a processor and a multi-feature matching-based multi-agent localization and mapping optimization program stored on the memory and executable on the processor, the multi-feature matching-based multi-agent localization and mapping optimization program implementing the steps of the multi-feature matching-based multi-agent localization and mapping optimization method of any one of claims 1-7 when executed by the processor.
10. A computer readable storage medium storing a multi-feature matching based multi-agent localization and mapping optimization program which when executed by a processor performs the steps of the multi-feature matching based multi-agent localization and mapping optimization method of any one of claims 1-7.

Description

Multi-agent positioning and graph optimizing method, system, terminal and storage medium based on multi-feature matching Technical Field The present invention relates to the field of information matching and positioning technologies, and in particular, to a multi-agent positioning and graph optimization method, system, terminal and computer readable storage medium based on multi-feature matching. Background Synchronous positioning and mapping (Simultaneous Localization AND MAPPING, SLAM) is a core technology for autonomous navigation of mobile robots, unmanned aerial vehicles and autonomous vehicles. SLAM in the prior art scheme performs well in a specific scenario, but has serious defects in indoor environments (such as a gallery of an office building, a plurality of rooms with the same layout, a hospital, a mall, etc.) with highly similar and repetitive structures, and the system may misjudge similar positions as the same positions, resulting in map construction or position positioning errors. Accordingly, the prior art is still in need of improvement and development. Disclosure of Invention The invention mainly aims to provide a multi-agent positioning and graph optimizing method, a system, a terminal and a computer readable storage medium based on multi-feature matching, and aims to solve the problem that the graph construction error is caused by the fact that the prior art has a position identification error for similar environments of repeated semantics and structures. In order to achieve the above object, the present invention provides a multi-agent positioning and graph optimizing method based on multi-feature matching, the multi-agent positioning and graph optimizing method based on multi-feature matching comprising the steps of: Acquiring a plurality of groups of environment images shot by a plurality of agents respectively, and constructing a plurality of text key frames corresponding to each agent according to each group of environment images; Calculating a plurality of text similarities of all the agents at a plurality of positions, and constructing a plurality of candidate matching pairs according to each text similarity; Acquiring Wi-Fi fingerprints of each candidate position in each candidate matching pair, and calculating the MAC address similarity between all the candidate positions according to all the Wi-Fi fingerprints; calculating signal strength similarity among a plurality of candidate positions according to all the MAC address similarity, and determining a plurality of true matching position pairs according to all the signal strength similarity; If the two positions of the true matching position pair come from the same agent, optimizing the pose graphs of the agents, and if the two positions of the true matching position pair come from different agents, fusing the pose graphs of the two agents. Optionally, in the multi-agent positioning and graph optimizing method based on multi-feature matching, the acquiring a plurality of groups of environment images respectively shot by a plurality of agents, and constructing a plurality of text key frames corresponding to each agent according to each group of environment images, before further including: Each intelligent body acquires a corresponding motion track and a plurality of environment images by utilizing a radar and a measuring unit of the intelligent body; And constructing a corresponding local pose graph according to each motion trail. Optionally, in the multi-agent positioning and graph optimizing method based on multi-feature matching, the acquiring multiple groups of environment images respectively shot by multiple agents, and constructing multiple text key frames corresponding to each agent according to each group of environment images specifically includes: acquiring each group of environment images of each intelligent agent, and identifying each group of environment images to obtain text features detected by each intelligent agent and pose features when each text feature is recorded; And constructing a plurality of text key frames of each intelligent agent according to each text feature and the corresponding pose feature. Optionally, the multi-feature matching-based multi-agent localization and graph optimization method calculates a plurality of text similarities of all the agents at a plurality of positions, and constructs a plurality of candidate matching pairs according to each text similarity, including: Calculating similarities between the text features of all the agents at all locations: ; Wherein, the AndRespectively represent the firstPosition and the firstThe text characteristics of the individual locations are used,Representation ofAndThe similarity between the two is set to be similar,AndRespectively representAndIs provided with a die for the mold,Representing the presentation to beConversion toThe minimum number of edits required; and if the similarity is not smaller than a first preset value, defining the correspo