Ball machine monitoring anomaly detection method based on PSPNet-RCNN

Abstract

The invention discloses a PSPNet-RCNN-based dome camera monitoring anomaly detection method, and the method specifically comprises the steps: collecting multi-period exposure images at the same position, fusing the exposure images to generate an HDR image, carrying out the SIFT feature registration of the HDR image and an original position image, obtaining a registered HDR image, and carrying out the enhancement processing; the voidage of void convolution of a PSPNet network pyramid pooling module is increased, coding and decoding convolutional neural network branches are added to PSPNet, a residual convolutional network is added to an entrance of a feature extraction module for inputting double images, an improved multi-scale PSPNet network is obtained, the improved multi-scale PSPNet network is fused with a Mask R-CNN network for extracting target information in an HDR image, and the target information in the HDR image is extracted. And a PSPNet-RCNN image anomaly detection network is obtained and is used for extracting an image anomaly region. According to the invention, the universality and accuracy of the dome camera monitoring anomaly detection method are improved, and the real-time performance and robustness of the detection result are enhanced.

Description

technical field [0001] The invention relates to the field of image anomaly detection, in particular to a PSPNet-RCNN-based dome machine monitoring anomaly detection method. Background technique [0002] As an important equipment in the monitoring field, PTZ Dome has been widely used in various open scenes that require dynamic monitoring. It has the characteristics of flexible control mode, fast operation speed, and precise zoom positioning. It has been used in electric power, water affairs , industry, education and other industries have a good performance. With the continuous enhancement and development of network, digitization, and high-definition in the monitoring industry, the investment in product research and development of the supporting software system for PTZ monitors and the development of the promotion market have also entered a new level. [0003] It is an important purpose of monitoring behavior to detect the abnormality of the monitored scene through monitoring...

AI Technical Summary

Problems solved by technology

[0003] It is an important purpose of monitoring behavior to detect the abnormality of the monitored scene through monitoring equipment to achieve the early warning effect. In order to achieve this purpose, before the introduction of computer-related technologies, the judgment and recording of monitoring have been carried out manually, but this method is limited by manual The working hours and energy of the monitoring personnel may be unable to continue large-scale monitoring and mental fatigue due to the long-term high concentration of the monitoring personnel, resulting in false detections and missed detections

After the introduction of computer automatic detection technology, the detection method of traditional surveillance image processing utilizes and analyzes the statistical information such as the probability density of pixels and the number of patterns in the image frames in the continuously collected surveillance video stream within a certain period of time to represent the background, and then uses Statistical differences or feature probabilities are used to distinguish foreground objects from background images, so as to achieve the purpose of monitoring image anomaly detection. This research direction establishes background models for images through statistical probability or time series, etc. The judgment of binary classification has the advantages of high real-time performance, high target detection rate, and long target detection distance. Good performance, in addition, changes in environmental factors or imaging conditions, such as light, vibration, exposure, etc., also have a greater impact on the accuracy of this method

[0004] The dome camera has the characteristics of wide monitoring range, fixed-point shooting of the lens according to the inspection requirements, dynamic exposure, and long detection intervals, etc., but it is impossible to directly use traditional algorithms for real-time analysis of image changes and differences

Images