Target detection method and device

Abstract

The invention discloses a target detection method and device and relates to the technical field of computers. A specific embodiment of the method comprises the following steps of receiving a video stream; performing target tracking processing on each frame of image in the video stream to obtain a target tracking result of each frame of image; performing target detection processing on a single frame of image in the video stream every n1 frames of images to obtain a target detection result of the frame of image, n being a quotient obtained by dividing the frame rate of the tracking processing bythe frame rate of the detection processing, and n being a positive integer; and if the target detection result exists in the image, performing calibration processing on a target tracking result according to the target detection result, taking the calibrated target tracking result as a final target detection result of the image for output, and otherwise, taking the target tracking result as a final target detection result of the image for output. According to the method, the detection frame rate can be further improved while the detection precision is ensured.

Description

technical field [0001] The invention relates to the field of computer technology, in particular to a target detection method and device. Background technique [0002] Vision-based object detection has always been a very important technical approach required for autonomous vehicles to perceive the surrounding environment. Different from passenger cars, due to mass production and cost considerations, logistics distribution robots often have limited computing power of on-board processors, which are limited by computing resources. How to deploy a large-scale target detection network on the intelligent platform of logistics distribution robots , Improving the detection frame rate has become the key to improving the environmental perception ability of logistics distribution robots. The currently widely used technical solutions include: model compression based on the algorithm level, such as network pruning, low-rank decomposition, weight quantization, fixed-point instead of float...

AI Technical Summary

Problems solved by technology

[0003] In the process of realizing the present invention, the inventors found that there are at least the following problems in the prior art: Although the engineering deployment of large-scale networks can be realized through the above two technical solutions, and the frame rate can be effectively improved, to a certain extent it is at the expense of a certain model At the cost of prediction accuracy, in actual engineering, it is often necessary to make a trade-off between increasing the frame rate and losing model prediction accuracy. Therefore, there is a limit to accelerating the model, and it cannot be guaranteed to meet the needs of actual engineering.

Images