Human body action detection method suitable for mobile robot platform

Abstract

The invention provides a human body action detection method suitable for a mobile robot platform, which comprises the following steps: step 1, collecting a characteristic quantity: editing an input video, and respectively extracting N target character characteristics A1, A2,..., AN belonging to RC and a group of background environment characteristic mapping X belonging to RC * H * W from a key frame by adopting a character detector and a backbone network, c represents a channel, H represents height, and W represents width; r represents a number field. The high-order interaction relation is modeled in the form of target person-background environment-target person relation (OCOR), the indirect relation between multiple target persons and the background environment is deduced, then action positioning is achieved more accurately and efficiently, the whole design is simple and flexible, information of the background environment and other objects is fully utilized, and the method has the advantages of being high in practicability and easy to popularize. The accuracy of target action detection can be effectively improved.

Description

technical field [0001] The invention relates to the technical field of robot application, in particular to a human body motion detection method suitable for a mobile robot platform. Background technique [0002] As an important branch in the field of video understanding, human action detection technology is being more and more widely used. At present, the obstacle avoidance of mobile robots is mostly based on passive methods such as laser radar and infrared induction. Once there is an emergency (such as a passerby suddenly appearing on the road of the robot), the mobile robot will suddenly brake, which will greatly reduce the use of the robot motor. Life expectancy; at the same time, in some complex environmental scenarios, unsafe things such as pickpocketing and people falling may occur from time to time. At present, only relying on artificial video monitoring judgments still has the defects of incomplete monitoring range and low efficiency. In view of the above problems, ...

AI Technical Summary

Problems solved by technology

The current relatively new technology in this field is to directly establish a model of the paired interaction relationship between two target objects, and then continue to determine the actions of the two, but the relationship between objects in reality is not always available. The clues that can be presented in pairs and can provide more accurate information often exist in the implicit interactive relationship between the target and surrounding things (that is, the high-order relationship derived from the direct first-order relationship); The predecessors have also done a lot of work to model the high-order interaction relationship. Most of them need to add a pre-trained object detector on the basis of the original network, which will make the network structure more complex and use has many limitations

Images