Image processing method and device

Abstract

The invention provides an image processing method and device. The method comprises steps of detecting a photographed image, and determining a target region in the image; and acquiring a target local image corresponding to the target area, and transmitting the target local image to target equipment, or performing feature recognition on the target local image to obtain a recognition result, and transmitting the recognition result to the target equipment. According to the method and the device, feature recognition is carried out on the local image when the requirement of normal image recognitionis met, so that the calculated amount and the transmission bandwidth are saved, a relatively high frame rate can be obtained, and a guarantee is provided for real-time image recognition and control.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to an image processing method and device. Background technique [0002] Three-dimensional gesture recognition needs to perceive the depth of the scene. The current technical methods for depth perception include binocular stereo vision, binocular active stereo vision, structured light stereo vision, light time-of-flight distance measurement, and laser radar distance measurement; current technology Under horizontal conditions, structured light, light time-of-flight, and the depth resolution that lidar can measure are low; affected by natural light and limited by sensor power, structured light and light time-of-flight methods can only be used at close range indoors; at the same time, structured light , light time of flight, and lidar have a high cost; gesture recognition in the far field (3 ~ 5m) needs to sense the depth of hand details such as each joint of the finger...

AI Technical Summary

Problems solved by technology

[0002] Three-dimensional gesture recognition needs to perceive the depth of the scene. The current technical methods for depth perception include binocular stereo vision, binocular active stereo vision, structured light stereo vision, light time-of-flight distance measurement, and laser radar distance measurement; current technology Under horizontal conditions, structured light, light time-of-flight, and the depth resolution that lidar can measure are low; affected by natural light and limited by sensor power, structured light and light time-of-flight methods can only be used at close range indoors; at the same time, structured light , light time of flight, and lidar have a high cost; gesture recognition in the far field (3 ~ 5m) needs to sense the depth of hand details such as each joint of the finger, and the existing depth perception method is structured light , light time-of-flight, and lidar cannot meet the resolution requirements

[0003] In the existing technology, binocular stereo vision has the following disadvantages compared with structured light, which makes it a great challenge to perform far-field gesture recognition: the calculation of depth calculation for the entire image area is very large, requiring a lot of computing resources , it is difficult to achieve real-time; the binocular data volume is large and difficult to transmit, limited by the transmission bandwidth, it is difficult to achieve high-resolution and high frame rate real-time transmission

Images