Research On Embedded 3D Structured Light Sensor

Line structured light is widely used in three-dimensional contour detection of object surfaces due to its advantages of non-contact,large measurement range,and fast response.In particular,the rapid development of modern industry has put forward a series of requirements for non-contact,large-scale,high-precision,etc.on the three-dimensional contour detection technology of object surfaces.Although the current products for 3D contour detection of object surfaces can meet some of the detection requirements,the real-time nature of the detection process and the miniaturization of the measurement system require that the current products can not meet the requirements.Therefore,research on embedded three-dimensional structured light sensors is of great significance.In view of the above problems,the system optical path is first designed according to the Scheimpflug principle to obtain a larger depth of field and always maintain a clear image.After analyzing and comparing the principles and advantages and disadvantages of direct laser triangulation and oblique laser triangulation,the final solution is determined according to the needs of the project.Then from the imaging relationship,the conversion formula between the three-dimensional coordinates of the points on the object in the world coordinate system and the row and column coordinates in the pixel coordinate system is derived.Finally,the non-linear equations are solved by experimental data to obtain the object distance and distance between the reference plane and the angle between the line structured light and the optical axis of the lens to complete the calibration of the system structure parameters.The center of the online structured light strip extraction part,the current mainstream methods include the gray center of gravity method,Steger algorithm,and directional template method.In this paper,after comparing the extraction accuracy and speed of the above methods,the gray-center-of-gravity method with fast extraction speed and moderate extraction accuracy is selected.In order to meet the design requirements of the high frame rate output of the measurement system,this paper chooses to perform GPU(Graphics Processing Unit)-accelerated processing on the line structure light stripe center extraction of the algorithm core.The main difference between a GPU and a CPU(Central Processing Unit)is that the GPU has weak control and strong computing,has strong parallel computing capabilities,and is more suitable for occasions that have strict requirements on image processing time.In the GPU,first define the number of threads in each thread block,then allocate a twodimensional grid according to the image size,and take single task multi-thread processing for the pixels in the image.On the premise of ensuring computing power,the system volume is reduced,and the processing unit adopts NVIDIA's high-performance embedded computing platform TX2.During the experiment,GPU and CPU processing time comparison was used to verify the acceleration effect.Finally,according to the design principle,a measurement system is set up in a laboratory environment,the system structure parameters are calibrated,and the measurement accuracy of the system is verified by using standard test objects.Compare CPU and GPU processing time to verify the time efficiency of the system.