In recent years,minimally invasive cochlear implant surgery has become the main way to treat hearing impairment.As the premise of minimally invasive cochlear implantation,the robotic temporal bone drilling is particularly important.Due to the differences in visual perception mechanism between human and machine,some traditional surgical markers cannot be well used in robotic surgery.Therefore,this thesis proposes a visual localization method of temporal bone based on auxiliary ring in order to solve the temporal bone vision localization problem in cochlear implant robotic drilling surgery.The main work is as follows:Firstly,the existing problems of cochlear implant drilling were studied and analyzed,and a robotic system of cochlear implant drilling based on binocular vision was built according to the surgical requirements.In order to optimize the operation process,a terminal execution fixture that can automatically switch the operation drill is designed.The temporal bone auxiliary ring was designed as the temporal bone localization marker,and the calculation method of surgical portal was designed based on the marker.In addition,a rapid temporal bone localization method based on the temporal bone auxiliary ring was proposed.Secondly,the temporal bone auxiliary ring object detection and feature extraction method of YOLOv4-tiny was proposed by adding penalty weight to the classification loss function of YOLOv4-tiny to reduce the false detection rate of titanium sphere.Experimental results show that the proposed method improves the Recall rate of titanium sphere by 5.53% compared with the original algorithm,and can more accurately identify the temporal bone auxiliary ring titanium sphere.Based on the results of target detection,the features of temporal bone auxiliary ring image were extracted by image processing method,and the inner ring features were extracted by ellipse fitting method with contour similarity discrimination.At the same time,an automatic multi-seed point region growth method combined with target detection results was proposed to accurately segment the target region of titanium sphere and extract the characteristics of titanium sphere center.Then,binocular vision was used to calculate the three-dimensional position and posture information of the temporal bone auxiliary ring.A feature point matching method based on pole line restriction was proposed for fast matching of left and right image feature pairs to optimize the matching efficiency.According to the principle of binocular measurement,the three-dimensional coordinate of each feature point of the auxiliary ring was calculated based on the feature point pair.The plane equation of the temporal bone auxiliary ring was calculated by multi-point plane fitting method,and the attitude Angle of the auxiliary ring was calculated.Finally,the overall temporal bone visual localization method was validated.The final experimental results show that the temporal bone auxiliary ring localization method proposed in this thesis can quickly and accurately detect the temporal bone auxiliary ring target,extract the characteristic points of the temporal bone auxiliary ring and measure the position of the temporal bone auxiliary ring.The temporal bone auxiliary ring positioning time was less than 0.5s,the spatial position measurement error was less than 0.1mm,and the spatial attitude Angle measurement error was less than 0.15°,which met the requirements of real-time temporal bone auxiliary ring positioning. |