Real-time target detection method based on online learning strategy

Abstract

The invention discloses a real-time target detection method based on an online learning strategy. The method comprises the following steps: firstly, acquiring an input target window and a scanning window from a sensor image, dividing the scanning window into a positive sample or a negative sample according to an overlapping degree parameter of the two windows, calculating a posterior probability, and establishing a random forest detector; traversing all scanning windows in a new frame of image, establishing a training sample set, and calculating a random forest feature vector and a posterior probability of each sample; executing target tracking, automatically obtaining latest target window information, dividing a training sample set into a positive sample set and a negative sample set according to overlapping degree parameters of a scanning window and a new target window, and training a random forest detector by using the training sample set; and finally, obtaining a trained detector through continuous iterative training, inputting a frame of image into the detector, and outputting a target detection result. According to the invention, fine detection of a small target can be realized, the calculation amount is small, and real-time processing can be realized on an embedded platform.

Description

technical field [0001] The invention belongs to the technical field of automatic target detection, and relates to a real-time target detection method based on an online learning strategy. Background technique [0002] Automatic target detection is an important function in modern weapons and equipment. It can provide reliable target type and location information, and provide a strong guarantee for tasks such as battlefield reconnaissance, border patrol, and precision strikes. It is a key factor in realizing the intelligence of weapons and equipment. Object detection based on machine learning is a key direction in the field of automatic object detection technology. [0003] Object detection methods based on machine learning mainly include support vector machines and deep learning. Support vector machine is based on statistics and has intuitive geometric interpretation and good generalization ability. However, its disadvantages are high computational complexity, poor adaptabil...

AI Technical Summary

Problems solved by technology

[0004] To sum up, the problems existing in the existing technology are: the construction of the data set and the training of the model take a lot of time, and the training can only be completed offline. When performing the target detection task, only the specific target category detection based on the offline training model can be completed. , it is difficult to detect any object of interest in the scene; the computational complexity is high, and the operational efficiency is low; the generalization of the training model is too high, and the fine-grained recognition of small objects is not ideal

Images