基于DSP的蝗虫视觉神经网络碰撞预警实现研究

Research on DSP-Based Locust's Visual Neural Network Collision Alarm Realization

针对运动场景下碰撞安全预警问题,研究一种基于DSP的蝗虫视觉神经网络碰撞预警系统。该系统通过视频采集器接收场景视频图像;借助图像处理器提取运动目标,利用此神经网络获取目标逼近行为量,以及利用已有预警检测方案和预警装置发出预警信号。实验结果表明,该预警系统能有效对视域中可能发生的碰撞进行预警,无附加雷达、红外线设备,具有灵敏度高、可靠性好、处理速度快和抗干扰能力强的优点,对运动场景下安全预警有应用潜力。

A locust＇s visual neural network collision alarm system,based on DSP was developed to solve the problem of collision alarm under movement environments. In this system,the video images of a moving scene were collected by a video collector; an image processor was used to extract moving objects in view,while the locust＇s visual neural network captures the approaching object＇s behaviors. A reported advance warning scheme was adopted to identify whether an alarm occurs. Experimental results have validated that the advance warning system cansend alarmsignals before a collision happens. It is of potential use for safety advance warning,with no additional radar or infrared instrumentations and the merits of low sensitivity,high reliability and efficiency as well asstrong anti-jamming.