Method and device for detecting object in image

Abstract

The embodiment of the invention discloses a method and device for detecting an object in an image for improving the real-time performance of target detection. The method comprises dividing the image to be detected into a plurality of grids according to a preset division method, inputting the divided image into a convolutional neural network (CNN) trained in advance, obtaining the eigenvectors, output by the CNN, corresponding to respective grids of the image, determining the maximum value of a category parameter in each eigenvector, determining the positional information of an object with a category corresponding to the category parameter according to the central point position parameter and the boundary dimension parameter in the eigenvector when the maximum value is greater than a set threshold. By determining the category and the position of the object in the image with the CNN trained in advance, the method may simultaneously detect the category and the position of the object, does not need to select multiple feature regions, saves detection time, improves detection real-time performance and detection efficiency, and facilitates overall optimization.

Description

technical field [0001] The invention relates to the technical field of machine learning, in particular to an object detection method and device in an image. Background technique [0002] With the development of video surveillance technology, intelligent video surveillance is used in more and more scenarios, such as traffic, shopping malls, hospitals, communities, parks, etc. The application of intelligent video surveillance is to detect objects through images in various scenarios Foundation. [0003] When performing target detection in an image in the prior art, a region-based convolutional neural network (Region Convolutional Neural Network, R-CNN) and its extensions Fast RCNN and FasterRCNN are generally used. figure 1 It is a schematic diagram of the process of object detection using R-CNN. The detection process includes: receiving an input image, extracting a candidate region (region proposal) in the image, calculating the CNN feature of each candidate region, and using...

AI Technical Summary

Problems solved by technology

The above-mentioned detection of the category and position of the object is two independent steps. In both steps, the 256-dimensional data needs to be detected separately. Therefore, this process will also increase the detection time, thereby affecting the real-time performance of object detection.

Images