CN111796600A - Object recognition and tracking system based on quadruped robot - Google Patents

Abstract

The invention discloses an object recognition and tracking system based on a quadruped robot, which relates to the technical field of robots and aims at solving the problems of low robustness of an existing robot to an environmental feature extraction algorithm, inaccurate updating of an environmental map and low real-time performance. According to the invention, through extraction of local invariant features of the image, uncertainty information processing and data association and updating of an environment map, efficient identification and processing of information by the robot are improved, and accuracy and efficiency of autonomous environment cognition of the robot are improved.

Description

An Object Recognition and Tracking System Based on Quadruped Robot

technical field

The invention relates to the field of robot technology, in particular to an object recognition and tracking system based on a quadruped robot.

Background technique

The cognition of the working scene and the target object of the quadruped robot is the premise and key to realize its applications such as autonomous navigation, object recognition and tracking, and environmental monitoring. At present, there are problems in the research on autonomous environment cognition of quadruped robots that the robustness of environmental feature extraction algorithms is not strong, as well as the inaccurate update of environment maps and low real-time performance. tracking system.

SUMMARY OF THE INVENTION

The object recognition and tracking system based on the quadruped robot proposed by the present invention solves the problems that the robot is not robust to the environmental feature extraction algorithm, and the environmental map update is inaccurate and the real-time performance is not high.

In order to achieve the above object, the present invention adopts the following technical solutions:

An object recognition and tracking system based on a quadruped robot, comprising a task unit, a map information unit, a feature processing unit, a feature extraction unit, an image preprocessing unit and an original data unit, the task unit establishes a task to be executed, the original The data unit acquires data through an odometer and a binocular vision sensor, the raw data unit transmits the acquired data to the image preprocessing unit for scene image preprocessing, and the image preprocessing unit transmits the preprocessed data Extracting the feature points of the scene image to the feature extraction unit, the feature extraction unit transmits the extracted data to the feature processing unit, the feature processing unit transmits the information to the map information unit, the map information The map information generated by the unit is compared with the information generated by the task to be executed established by the task unit.

Preferably, the tasks to be performed include autonomous navigation, object recognition and tracking, environmental monitoring and environmental cognition.

Preferably, the map information unit generates a geometric map, a topological map, a grid map and a mixed geometric and topological map.

Preferably, the feature processing unit is used to process geometric feature information, associated data and uncertain information in the environment.

Preferably, the feature extraction unit extracts the feature points of the scene image, and then generates the scene image feature descriptor from the extracted feature points, and the scene image feature descriptor matches the information with the scene image feature, and finally matches the matched image. information for local feature description.

Preferably, the feature extraction unit establishes an accurate correspondence between image local invariant features and map database features through the influence of extraction and matching, and uncertainty information associated with environmental feature database data, and updates the corresponding geometric- Topological hybrid map library to achieve accurate cognition of environmental information.

Preferably, the environment cognition is to construct a complete unknown environment map according to acquired sensor information and cognitive strategies when the sensor device autonomously passes through a completely unknown environment.

Compared with the prior art, the beneficial effects of the present invention are: the present invention improves the efficient identification and processing of information by the robot through the extraction of local invariant features of the image, the processing of uncertain information and data association, and the updating of the environment map. Improve the accuracy and efficiency of the robot's autonomous environmental cognition.

Description of drawings

1 is a schematic diagram of the workflow of an object recognition and tracking system based on a quadruped robot proposed by the present invention;

FIG. 2 is a schematic diagram of information processing of an object recognition and tracking system based on a quadruped robot proposed by the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, but not all of the embodiments.

Referring to Figure 1-2, an object recognition and tracking system based on a quadruped robot includes a task unit, a map information unit, a feature processing unit, a feature extraction unit, an image preprocessing unit and a raw data unit, and the task unit establishes tasks to be executed. , the original data unit acquires data through odometer and binocular vision sensor, the original data unit transmits the acquired data to the image preprocessing unit for scene image preprocessing, and the image preprocessing unit transmits the preprocessed data to the feature extraction unit Extract the feature points of the scene image, the feature extraction unit transmits the extracted data to the feature processing unit, the feature processing unit transmits the information to the map information unit, the map information generated by the map information unit and the information generated by the task to be executed established by the task unit Compare.

In this embodiment, the tasks to be performed include autonomous navigation, object recognition and tracking, environmental monitoring, and environmental cognition.

In this embodiment, the map information unit generates a geometric map, a topological map, a grid map, and a mixed geometric and topological map.

In this embodiment, the feature processing unit is used to process geometric feature information, associated data and uncertain information in the environment.

In this embodiment, the feature extraction unit extracts the feature points of the scene image, and then generates the scene image feature descriptor from the extracted feature points. The scene image feature descriptor matches the information with the scene image features, and finally the matched information is processed. Local feature description.

In this embodiment, the feature extraction unit establishes an accurate correspondence between the local image invariant features and the map database features by extracting and matching the uncertainty information associated with the environmental feature database data, and at the same time updates the corresponding geometric- Topological hybrid map library to achieve accurate cognition of environmental information.

In this embodiment, the environment cognition is to construct a complete unknown environment map according to the acquired sensor information and the cognitive strategy when the sensor device autonomously passes through the completely unknown environment.

Working principle, the cognition and understanding of the environment of the quadruped robot is the primary prerequisite for its autonomous navigation, including path planning, autonomous positioning and map creation. In the autonomous navigation of the vision-based quadruped robot, the visual sensor is mainly used to perceive its working scene, and the image features are extracted to represent the typical environmental information describing the robot. And extract stable image feature points to represent the actual physical points in 3D space, use them as natural features to construct a geometric map of the environment, and at the same time match with the features in the environment map (natural feature library) created before the current moment, The robot's current pose is estimated and the natural feature library is updated, so as to realize the autonomous cognitive process of the quadruped robot. By dealing with the influence of uncertain information such as feature extraction and matching, environmental feature database data association, etc., the accurate correspondence between image local invariant features and map database features is established, and the corresponding geometry-topology hybrid map Accurate cognition of environmental information. When the robot independently passes through the completely unknown environment according to the sensor equipment it carries, it constructs a complete or nearly complete unknown environment map according to the acquired sensor information and cognitive strategies. In specific applications, the tasks of quadruped robots include self-positioning, path planning and navigation, autonomous environment cognition, etc. In order to meet the needs of different tasks, the robot needs to create corresponding environment maps in the process of environment cognition, such as geometric Topological maps, raster maps, and geometric, topological hybrid maps, etc.

The above description is only a preferred embodiment of the present invention, but the protection scope of the present invention is not limited to this. The equivalent replacement or change of the inventive concept thereof shall be included within the protection scope of the present invention.

Claims (7)

1. An object recognition and tracking system based on a quadruped robot comprises a task unit, a map information unit, a feature processing unit, a feature extraction unit, an image preprocessing unit and an original data unit, it is characterized in that the task unit establishes a task to be executed, the original data unit acquires data through a speedometer and a binocular vision sensor, the original data unit transmits the acquired data to the image preprocessing unit for scene image preprocessing, the image preprocessing unit transmits the preprocessed data to the feature extraction unit to extract the scene image feature points, the feature extraction unit transmits the extracted data to the feature processing unit, the feature processing unit transmits the information to the map information unit, and the map information generated by the map information unit is compared with the information generated by the task to be executed established by the task unit.

2. The system of claim 1, wherein the tasks to be performed comprise autonomous navigation, object recognition and tracking, environmental monitoring and environmental awareness.

3. The quadruped robot-based object recognition and tracking system of claim 1, wherein the map information unit generates a geometric map, a topological map, a grid map, and a mixed geometric and topological map.

4. The quadruped robot-based object recognition and tracking system of claim 1, wherein the feature processing unit is configured to process the geometric feature information and the associated data and uncertainty information in the environment.

5. The system of claim 1, wherein the feature extraction unit extracts scene image feature points, generates scene image feature descriptors from the extracted feature points, matches the scene image feature descriptors with information, and performs local feature description on the matched information.

6. The system for identifying and tracking objects based on the quadruped robot as claimed in claim 1, wherein the feature extraction unit establishes an accurate correspondence between the local invariant features of the image and the features of the map database through the influence of the uncertainty information associated with the extraction and matching and the data of the environmental feature database, and updates the corresponding geometric-topological mixed map library to realize accurate cognition on the environmental information.

7. The system of claim 2, wherein the environment recognition is to construct a complete unknown environment map according to the acquired sensor information and the recognition strategy when the sensor device autonomously passes through the completely unknown environment.

Images