Target detection method and device in image, electronic equipment, storage medium

Abstract

The invention discloses a target detection method in an image, comprising: acquiring an image to be detected; establishing a first-layer pyramid image feature data set of an image pyramid according to the image to-be-detected, and building a pyramid image feature data set on the first layer Detect the target object in the data set; if the target object is detected, stop the detection process and output the detection result; if the target object is not detected, then continue to build the next layer of pyramid image feature data set, and in the next The target object is detected in the layer pyramid image feature data set; the feature resolution corresponding to the pyramid image feature data set of the next layer is greater than the feature resolution corresponding to the pyramid image feature data set set up in the previous step; repeat the above-mentioned establishment of pyramid image feature data The steps of collecting and detecting target objects are performed until the target object is detected. The invention also discloses a target detection device in an image, electronic equipment and a storage medium.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a method and device for detecting objects in an image, electronic equipment, and a storage medium. Background technique [0002] Image pyramid is a representation of multi-scale representation of images, and it is an effective but conceptually simple structure to explain images at multiple resolutions. In image processing, image pyramids are used by a wide range of algorithms as a data structure that stores image information at multiple scales. However, when the scale space is relatively dense and the amount of stored data is large, the establishment and processing of the data structure takes a long time. In scenarios such as real-time video processing or embedded solutions, the processing speed of image or machine vision algorithms will be reduced. [0003] Conventional image pyramid building adopts the strategy and structure of "build first, then use". This pr...

AI Technical Summary

Problems solved by technology

[0006] Time is wasted, and more time is wasted due to the higher resolution of features of distant targets, which take longer to set up

However, the strategy of target detection starts from the pyramidal layer of the larger small-scale features, thus creating a contradiction of time loss, such as figure 2 shown

[0007] When the algorithm works on an embedded platform, using figure 2 The static pyramid data structure establishment mode shown will greatly increase the processing time of the detection algorithm, resulting in low operating efficiency of the algorithm, thereby reducing the user experience of the product

Images