Research On Dam Crack Detection Algorithm Based On Image Processing

China has the most number of dams in the world.However the dam failure rate and the aging rate of the dam goes the same high which makes a real big problem.The traditional way of monitoring the dam is manually.And it's easy to find that this way of monitoring dams has too many disadvantages.It has the disadvantage of being dangerous,not accurate enough,and wasting human and material resources.The dam crack image has problems of low contrast,uneven illumination,and severe surface noise interference.Due to the features of the dam,it is quite hard to get the good result.Besides,the texture noise interference problem and the blurred image and the skeleton extraction of complex cracks increase the difficulty of crack detection.According to the dam image signal model established by analysis,this paper systematically studies the denoising preprocessing,image enhancement,crack skeleton extraction and parameter acquisition.Aiming at the problem of texture noise interference of dam image,this paper analyzes the spectral characteristics of noise signal and obtains the spectral relationship between noise signal and image signal.An image denoising algorithm based on wavelet transform is proposed.The algorithm first performs Discrete wavelet processing on the image signal and uses the wavelet function to perform two-layer decomposition.Next,the threshold is set to filter out the noise signal,and then the final denoised image is obtained by two reconstructions.Aiming at the problem that the dam image is blurred and the underwater dam image is difficult to identify in foggy weather,this paper studies the imaging process of dam image based on the image characteristics of dam image,and abstracts the image of dam image.Scattering and refraction model.The dark channel image enhancement algorithm proposed in this paper uses the dark channel prior theory to derive the scattering and refraction model,and obtains the restored expression of the image.The algorithm quickly obtains the atmospheric light value and transmittance through down sampling and guided filtering,and optimizes the transmittance by weighted summation.The enhanced image is obtained based on the atmospheric light value and the transmittance value.A dam image homogenization algorithm based on mask principle is proposed to solve the uneven illumination problem.The algorithm uses a rectangular filter mask to simulate the light intensity change of the image.A preliminary homogenized image is obtained by calculating the mask model.The final homogenization result is obtained by stretching reinforcement.Aiming at the problem that PCNN algorithm needs a lot of time to adjust parameters,this paper proposes a PCNN image segmentation algorithm based on genetic algorithm.By searching the threshold matrix required by PCNN optimally by genetic algorithm,it can automatically complete the image segmentation function and greatly reduce the cost of manual adjustment.Finally,the parameters of the obtained skeleton cracks are calculated according to the mathematical morphology and the connected domain principle,so that the crack detection can be completed automatically.