车辆碰撞检测方法、电子设备及计算机可读存储介质
By converting video and audio data into feature data and using multiple deep neural network models for comprehensive judgment, the problem of low accuracy in vehicle collision detection has been solved, achieving higher detection accuracy and stability.
CN117593725BActive Publication Date: 2026-07-17SHENZHEN XIAOJING TECH CO LTD
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHENZHEN XIAOJING TECH CO LTD
- Filing Date
- 2023-11-28
- Publication Date
- 2026-07-17
AI Technical Summary
Technical Problem
Existing vehicle collision detection methods rely on single image or sound data, which cannot fully reflect information about the vehicle in motion, resulting in low detection accuracy.
Method used
The raw video and audio data are transformed into feature data, including audio energy spectrum and video feature data. Multiple deep neural network models are used for comprehensive judgment, including an audio detection model based on EfficientNet and a video detection model based on InceptionV3 and GRU. Collision detection is performed by combining audio and video features.
Benefits of technology
It improves the accuracy and stability of vehicle collision detection, especially performing better in complex scenarios, and enhances the accuracy and applicability of detection.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN117593725B_ABST
Abstract
本申请公开了一种车辆碰撞检测方法、电子设备及计算机可读存储介质，涉及车辆安全技术领域，所述车辆碰撞检测方法包括：将原始视频数据和原始音频数据转换成符合预设格式的特征数据，其中，所述特征数据至少包括音频能量频谱和视频特征数据；将所述音频能量频谱输入预设的第一音频检测模型，通过所述第一音频检测模型进行车辆碰撞检测，得到第一检测结果；若所述第一检测结果为碰撞，则将所述视频特征数据输入到预设的目标视频检测模型中，通过所述视频检测模型进行车辆碰撞检测，得到第二检测结果。本申请与使用单一数据流、单个深度神经网络的检测方案相比，融合了视频数据和音频数据。来进行碰撞检测，大幅度提升了车辆碰撞检测的准确率。
Need to check novelty before this filing date? Find Prior Art