一种具有尺度协调机制的实时视觉目标识别方法

• Format: Patent
• Language: Chinese
• Published: 18.04.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS

一种具有尺度协调机制的实时视觉目标识别方法，以目标分割确定待识别主体的大小，协调实时目标跟踪中待识别区域的尺度，从而提高识别精度。该方法首先对采集到的图像按对比度作图像分割，将待识别主体从背景中分割出来以确定待识别主体的大小。此外，在普通的实时视觉目标识别流程中，目标跟踪算法计算出每一帧中跟踪目标主体位置，并将主体位置提供给目标识别算法完成识别。当目标分割后的主体大小与目标跟踪后的主体大小之比高于阈值时，启动尺度协调机制，以目标分割大小为准，调整识别框基于跟踪框的扩大倍数。所提方法能够有效应对目标主体形变、旋转等问题。 A real-time visual target identification method with a scale coordination mechanism determines the size of a to-be-identified subject by target segmentation, and coordinates the scale of a to-be-identified region in real-time target tracking, thereby improving the identification precision. The method comprises the following steps: firstly, carrying out image segmentation on an acquired image according to a contrast ratio, and segmenting a to-be-identified main body from a background to determine the size of the to-be-identified main body; besides, in a common real-time visual target identification process, a target tracking algorithm calculates the position of a tracking target main body in each frame, and provides the position of the main body for th