Method and apparatus for detecting object in image and electronic device

Abstract

Embodiments of the invention disclose a method and an apparatus for detecting an object in an image and an electronic device. The method comprises the following steps: locating a first sliding window at a starting sampling point in the image; repeating the following steps: calculating a response value according to a matching degree of a pixel point in the first sliding window with a preset feature of the object; determining the position of a next sampling point in the image according to a preset sliding direction depending on the response value; and locating the first sliding window at the next sampling point until the image is retrieved to obtain a first group of sampling points comprising the sampling points where the first sliding window is located; and detecting the position of the object according to the first group of sampling points and the corresponding response values thereof. The method can be used for effectively shortening the image search time and improving the efficiency of computer vision.

Description

technical field [0001] Embodiments of the present invention relate to the field of computer vision, and in particular, relate to a method, device and electronic device for detecting objects in an image. Background technique [0002] With the proliferation of mobile devices, mobile interactions and services are facilitating the wider use of computer vision for mobile solutions and mobile user interfaces. For example, gesture tracking and recognition is becoming an important interaction enabler for wearable devices. [0003] However, compared to human vision, computer vision is still inefficient and cannot meet the needs of many practical mobile use cases. The reason mainly lies in the following two aspects: [0004] (1) Object detection is very time-consuming. Unlike human vision, which can always search for objects in the field of view in a parallel manner, most existing computer vision searches for objects in a traversal manner. Because the size and direction of the obj...

AI Technical Summary

Problems solved by technology

[0005] (2) Target recognition is unstable and recognition errors often occur

Because the size and direction of the object to be detected are usually unknown, if the object's model happens to miss the object, the object cannot be correctly identified

[0006] It can be seen that, on the one hand, many calculations in a large number of exhaustive searches are redundant, and using exhaustive searches is very time-consuming

On the other hand, although many models are trained with different scales and orientations, there are always some scales and orientations that cannot be covered, which makes some objects missed in the exhaustive search.

Images