Method for rapidly measuring distance between multiple moving objects based on machine vision

Abstract

The invention relates to a method for rapidly measuring a distance between multiple moving objects based on machine vision. The method comprises the following steps: fitting a ground truth box in an image training set and actual distance, storing a distance measuring model, performing height extraction on a bounding box generated by a specific target by a real-time detection algorithm SSD (singleshot multibox detector) and inputting into the distance measuring model so as to acquire the measuring distance between multiple objects. According to the multi-moving-object rapid identification anddistance measurement method, the specific multiple objects can be subjected to identification and distance measurement efficiently in real time. Compared with other ETA (electronic travel aids) distance-measuring systems, the method is more practical and has a good application prospect and an application value of actual engineering.

Description

technical field [0001] The invention relates to a machine vision-based fast distance measuring method for multiple moving targets. Background technique [0002] Today, there are 75 million visually impaired people in our country, whose life and travel are greatly troubled by their visual impairment. The emergence of the Electronic Travel Aids (ETA, Electronic Travel Aids) system to help the blind has brought great convenience to the blind. Among them, obstacle detection and distance measurement are the key technologies in the ETA system, and the accuracy and real-time performance of detection and recognition are the measures. An important symbol of the practicability of the ETA system is also a difficult problem to be solved urgently. [0003] In the past non-computer vision fields, most of them used laser, infrared, and ultrasonic ranging methods. However, its ranging distance is short, the beam is narrow, obstacles cannot be identified, and multiple targets can only be d...

AI Technical Summary

Problems solved by technology

In addition, other existing ETA systems based on machine vision still rely on an additional rangefinder for distance measurement, and cannot simultaneously measure distances for multiple moving targets

Aiming at the deficiencies of the existing ETA system, the present invention solves the real-time and recognition problems in the core detection and distance measurement module in ETA from the new angle of combining target detection algorithm and regression distance measurement

Images