Anomaly Detection Method Based on Deep Learning in Complex Environment

Abstract

The invention provides an anomaly detection method based on deep learning in a complex environment. By inputting the spatiotemporal features of objects extracted by the convolutional neural network regression method into the LSTM model, the trajectory of multiple objects in the complex environment can be tracked; In this case, capture the nonlinear spatiotemporal actions of adjacent individuals, evaluate the dependence of motion trajectories between adjacent individuals, and predict their future trajectories; complete anomaly detection based on the abnormal probability of individual future motion trajectories. This method can reduce the false detection rate of images. The existing technology mainly detects the spatiotemporal characteristics of a single object, and does not consider the mutual interference of adjacent individual motion trajectories in a complex environment. The LSTM model described in the present invention, by evaluating Dependencies between coherent individuals, using the encoding and decoding framework to predict the future trajectory of objects, so that more accurate results can be obtained when performing anomaly detection on multi-object motion.

Description

technical field [0001] The invention relates to an anomaly detection method based on deep learning in a complex environment. Background technique [0002] In general, anomaly detection refers to the detection of abnormal behavior in the environment or data that does not conform to expected behavior. With the promotion of deep learning in the field of artificial intelligence, computer vision technology has been widely used in anomaly detection in complex environments such as subways, stadiums, and airports. However, such high-density environments have brought great challenges to anomaly detection. . In the face of the continuous irregular movement of a large number of objects, how to solve the problem of mutual interference between objects and how to detect anomalies under the condition that the motion trajectories of multiple objects influence each other have become an important problem in the current anomaly detection research. [0003] Abnormalities can be rare shapes or...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to provide an anomaly detection method based on deep learning in a complex environment, which can reduce the false detection rate of images under the condition that adjacent individual motions affect each other, especially in a crowded environment. Good performance provides a new idea for solving the problem of anomaly detection in complex environments, and solves the existing technology that does not consider the mutual interference of adjacent individual motion trajectories in complex environments, making the anomaly detection effect unsatisfactory The problem

Images