Research On The Bridge Crack Detection Method Based On Image Identification Technology

As the current main building material,concrete is widely used in the construction of bridge facilities.Under the influence of unfavorable factors such as temperature change,vehicle load,uneven foundation settlement,etc.,the bridge concrete will have cracks that are not conducive to its normal operation.If the bridge crack exceeds its cracking limit,reduce its life cycle,and cause serious harm to people’s lives and properties.Periodic inspection of bridge cracks is a basic method of bridge maintenance,and image inspection of cracks is the most intuitive method of inspection.In this paper,by establishing a bridge crack collection system,combined with image processing technology to detect bridge cracks,the cracks under the complex concrete background are extracted with high noise resistance and their characteristic information is extracted.The research content mainly includes the following four aspects:(1)Research work on image acquisition equipment.Making an optimal solution between CCD image sensor and CMOS image sensor,because CCD image sensor has relatively high requirements for stability,it is not suitable for the field of engineering inspection.Due to its low cost and low energy consumption,CMOS image sensor is convenient for network scanning in engineering inspection,and CMOS image sensor has rich colors and richer details,which is more suitable for engineering inspection field.(2)Crack detection based on deep residual network.Py Torch is used to build a deep learning framework,and the deep residual network Res Net-50 is selected as the model to train the crack detection neural network system to test and evaluate the generalization ability of the model to further enhance the adaptability of the model.(3)Fracture extraction based on dynamic threshold method.Using preprocessing methods such as gray scale transformation,filtering and denoising,and image enhancement to remove most of the noise of the image,and then use the OTSU method to binarize the preprocessed image to obtain a connected domain that only contains cracks and large noises.set.Sort by the size of the connected domains,set the second largest connected domain area as the threshold variable,extract the crack connected domains from the image,and eliminate the large noisy areas of the image that are not cracks.(4)Crack feature recognition based on image morphology method.By counting the number of pixels in the crack area and calibrating it,the actual area of the crack area is calculated.The image morphology calculation method is used to refine the skeleton of the crack area,so that the crack is arranged in a single pixel along the length of the crack,the number of pixels is counted,and the true length is obtained according to the calibration coefficient.The crack area is divided by finite element,rectangular grid is divided,and the average width of the crack is obtained according to the information of the area and length of the crack.