基于双目视觉的台区测量算法研究

Research on the algorithm of station area measurement based on binocular vision

配电网台区建设质量的优劣,直接影响着台区的供电可靠性。针对传统配电网台区建设中人工验收存在的效率低、成本高、周期长、安全风险大等问题,设计了一种基于双目视觉和人工智能的台区量化指标测量算法。设计的多阶段检测算法能够实现对台区关键部件的高精度定位识别。基于定位区域进而利用双目测量点匹配策略实现对台区验收量化指标的高精度测量。试验结果表明,所设计的测量设备能够在5 m工作距离内实现对15 m标准台区的拍摄测量,算法测量相对人工测量的误差范围为0.45%~1.97%,满足了配电网台区标准化建设验收中量化指标的精度需要。

The quality of the construction of the Distribution Transformer Base Plate(DTBP)directly affects the power supply reliability.Aiming at solving problems in traditional manual acceptance of DTBP construction,such as low efficiency,high cost,long cycle,and high safety risk,a measurement algorithm and equipment based on binocular vision and artificial intelligence is proposed.The designed multi-stage detection algorithm can locate and identify the key components of DTBP with high precision.Based on the located area,a binocular measurement point matching strategy is used to achieve high-precision measurement of the quantitative indicators.The experimental results show that the designed measurement equipment can realize the photographic measurement of the 15 m standard DTBP within a working distance of 5 m,and the error range of the algorithm measurement compared with the manual measurement is 0.45%~1.97%,which meets the accuracy requirements of quantitative indicators in the standard construction of DTBP.