Image-based Lane Recognition And Research On Automobile Lane Departure Warning

As an important part of automobile,body-in-white(BIW)plays an important role,including protecting the safety of driver and passengers and forming a good dynamic environment.The manufacturing quality of BIW directly affects the safety,carrying capacity and economy of automobile.And in recent years,more and more traffic accidents also put forward high requirements for car body.Spot welding is the most commonly used connection mode on BIW,and the quality of the welding spot directly affects the safety of the whole BIW,so it is of great significance to detect the quality of the welding spots on BIW.However,the current inspection of the quality of the welding spots of the BIW mainly relies on the manual inspection of hand-held inspection equipment.However,this manual inspection method has the disadvantages of low efficiency,high labor cost,and inability to timely feedback the production line.Therefore,an automatic and intelligent inspection method of welding spot quality is urgently needed.In order to meet the high requirements for the position of the welding spot in the automated inspection task of welding spot quality,this paper studies the precise positioning method of the solder joint.The main research contents of this article are as follows:1.Region of interest(ROI)selection of welding spots.In this paper,a deep learning object detection algorithm is used to automatically extract the ROI of the welding spot.The object detection algorithm based on convolutional neural network has strong robustness and accurate target recognition ability.Taking into account the size distribution of the welding spots and the real-time requirements of the welding spot quality inspection scene,the Tiny-YOLOv3 algorithm was finally chosen to automatically select the ROI of the welding spots.Experiments showed that the trained Tiny-YOLOv3 algorithm has a strong ability to identify and locate welding spots.Through the expansion of the prediction box,it successfully selected accurate ROI that retain all the contour information of the welding spots and almost contain no noise,which provided the good enough contour information for subsequent fine positioning of welding spots.2.Welding spot contour extraction.According to the circle-like property of the welding spot,this paper adopted the circle detection algorithm based on traditional image processing as the basic algorithm for welding spot contour extraction.In order to reduce the noise on the basis of retaining enough welding spot contour information as much as possible,different image denoising algorithms,image enhancement algorithms and image edge detection algorithms were compared to select the suitable preprocessing algorithm for welding spot images;In order to improve the sampling efficiency and detection accuracy of the Randomize Hough Transform(RHT)circle detection algorithm,this paper proposed a right triangle circle detection(RTCD)algorithm,which reduced the invalid accumulation of sampling and improved the sampling efficiency of RHT algorithm.In order to further improve the accuracy and robustness of the RTCD,this paper proposed an optimization strategy based on a pixel distance constraint and a pixel angle constraint to enhance the robustness of the algorithm.Experiments showed that the proposed RTCD algorithm performed well on the task of extracting welding spots contours,and the contour extraction efficiency was nearly 10 times higher than CHT and RHT.3.Research on precise positioning method of welding spots.Since the ultrasonic inspection task of solder joint quality has very precise requirements for the position of welding spots,and neither the Tiny-YOLOv3 algorithm nor the RTCD algorithm can meet the actual positioning needs,so this article explored the combination of the two methods to achieve precise positioning of welding spots.The Tiny-YOLOv3 algorithm was used to select the ROI of the welding spot,and then the RTCD algorithm was used to finely locate the welding spot based on the ROI box.Experiments showed that this combined method can achieve more accurate welding spot positioning than the Tiny-YOLOv3 algorithm or the RTCD algorithm alone,and the method was very robust and was not easily affected by complex environmental factors,such as uneven light,rust and oil stain.In order to further verify the effectiveness of the combined method,this article used the ultrasonic method to verify it.The experiment showed that the combined method used in this article has a better ultrasonic waveform than the Tiny-YOLOv3 algorithm alone,and the positioning of the welding spot is more accurate.