Target detection method, target detection model and system based on cascaded detectors

Abstract

The invention discloses a target detection method, target detection model and system based on cascaded detectors. The detection method includes the following steps: S1: using a training data set with target labels to train the target detection model; S2: The image to be tested is input into the trained target detection model, and feature maps P of different sizes are obtained after feature extraction, upsampling, and feature fusion 2 ,P 3 ,...,P n ; S3: According to the feature map P n Predicted target position B n and class C n ; According to the target position Bi from the corresponding feature map P i‑1 Extract the corresponding features to predict the target, and get the target position B i‑1 and class C i‑1 , i=3~n; S4: take target position B 2 As the final prediction result; the present invention obtains a more accurate predicted target position through multiple iterative regression predictions of target candidate frames, improves target detection accuracy, and has higher detection accuracy.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and more specifically relates to a cascaded detector-based high-precision target detection model, target detection system and method. Background technique [0002] In the field of computer vision, target detection is a very active and challenging research direction. Unlike classification tasks, target detection not only needs to identify the category of the target, but also locates the position of the target, so as to achieve the task of positioning and recognition. , It is closely related to many real-life applications, such as the detection and recognition of license plates in the field of traffic security, the detection and recognition of driverless car traffic lights and traffic signal signs, the algorithm of camera face positioning, etc. Target detection is usually divided into two steps: the first step is to locate the target area, and the second step is to identify the content in t...

AI Technical Summary

Problems solved by technology

[0005] Aiming at at least one defect or improvement requirement of the prior art, the present invention provides a cascaded detector-based target detection method, target detection model and system, and performs multiple target bounding boxes on feature maps of different scales during prediction Regression, in order to obtain a more accurate positioning of the bounding box, the purpose is to solve the problem of low target positioning accuracy existing in the existing target detection method

Images