Video frame loss and frame clamping anomaly detection method and system based on Bi-LSTM (Bidirectional Long Short Term Memory)
Through the Bi-LSTM-based video anomaly detection method, convolution operation and bidirectional long and short-term memory networks, the cumbersome problems of video detection in the existing technology are solved, automated detection and high accuracy are achieved, and labor and time costs are saved.
CN119071565AInactive Publication Date: 2024-12-03GUANGZHOU BAOLUN ELECTRONICS CO LTD
Patent Information
- Application Number
- CN202410987412.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-23
- Publication Date
- 2024-12-03
- Estimated Expiration
- Not applicable · inactive patent
Smart Images

Figure CN119071565A_ABST
Abstract
The invention discloses a Bi-LSTM-based video frame loss and frame clamping anomaly detection method and system, and the method comprises the steps: obtaining a video data set which comprises a plurality of calibration videos, and the content between frames in each calibration video is logical; a video recognition neural network model is constructed, the video recognition neural network model comprises a Bi-LSTM layer, a video data set is input into the video recognition neural network model, and a trained video recognition neural network model is obtained; and loading the calibration video into a scene to be detected, detecting the scene to be detected carrying the calibration video by using the video identification neural network model, and recording a position where contents between frames of the calibration video have logicality, the position being a frame loss or frame clamping position. According to the method, the logicality of the video content is learned by training the Bi-LSTM, so that whether the video is lost or blocked or not is detected when the playing equipment plays the video, and the problems that an existing detection method wastes time and labor and is tedious in operation are solved.
Need to check novelty before this filing date? Find Prior Art
Citation Information
Patent Citations
Tensor-model-based infrared dim target detecting method
CN103729854A
Video equipment image frame test method and delay measurement method
CN113676722A
Audio frame loss detection method and training method and device of frame loss detection model
CN116312457A
Pipeline magnetic flux leakage image recognition method based on improved residual shrinkage network
CN117058443A
Video frequency broadcast quality detecting method for medium broadcast terminal device
CN1859584A
