The invention discloses a bridge crack detection method. Due to the fact that manual marking has certain subjectivity, the pavement crack detection precision depends on experience knowledge of experts, and experience lacks objectivity in quantitative analysis. The method comprises the following steps: 1, collecting images of a detected bridge floor one by one to obtain a bridge floor image set; 2,splicing images; 3, detecting crack; 4, extracting crack parameter. According to the invention, the image acquisition and processing technology replaces human eyes to complete automatic nondestructive detection of the bridge crack, and has very important practical significance for research of the bridge crack detection technology in a complex terrain environment. On one hand, the construction safety is enhanced, and on the other hand, the operation maneuverability and flexibility are improved. According to the invention, fidelity splicing of multiple groups of bridge deck images is realized,the image splicing precision and efficiency are improved, a working foundation is laid for subsequent bridge crack image detection, and a technical reference is provided for image splicing detection in other fields.

Bridge crack detection method

The embodiment of the invention provides a drilling leakage crack width prediction method based on neural network data mining. The method comprises the steps that historical drilling data, a leakage stoppage case, an imaging logging real crack width and other data information of a target block are collected; data preprocessing is carried out on the collected data information, wherein the preprocessing content comprises data cleaning, integration and conversion; the preprocessed historical drilling data is used as input, the crack width is used as output, the imaging logging real crack width isused as a standard value, and a crack width prediction neural network model is obtained through supervision training and optimization; and the drilling instant data related to the target positive drilling well is imported into the neural network model, and the model automatically judges the corresponding well depth crack width at the moment. By utilizing the technical scheme provided by the embodiment of the invention, the crack width of the positive drilling well can be conveniently and accurately predicted in real time, so that decision support is provided for leakage stoppage constructionpersonnel, and the one-time leakage stoppage success rate of crack leakage is improved.

Drilling leakage crack width prediction method based on neural network data mining

The invention discloses a method for detecting cracks in a bridge image. Due to the fact that manual marking has certain subjectivity, detection precision depends on experience knowledge of experts, and experience lacks objectivity in quantitative analysis. The method comprises the following steps: 1, detecting the position of a crack in a bridge floor image, and completely extracting the morphological characteristics of the crack; 2, extracting real parameters of cracks in the image. According to the invention, an image processing technology is used to replace human eyes to complete automaticnondestructive detection of bridge cracks. Aiming at the unique spatial ductility and gray distinguishability of bridge cracks, the improved median filtering algorithm is provided, according to the characteristics of crack edge similar gray linear distribution, on the basis of original median filtering, the thought of the similar gray scale extension direction is introduced, interference of bridge floor image composite noise can be restrained more effectively, the signal-to-noise ratio of the image is increased, the continuity of detail information of the crack edge is kept, and the reliability of fidelity filtering of a target crack is improved.

Method for detecting cracks in bridge image

The invention provides a crack identification method and system based on ABC-PCNN. The method comprises the following steps: extracting a characteristic vector of a pavement crack image, and further comprising the following steps: training a test sample by using an optimized BP neural network; and adopting the cross entropy of the predicted value and the true value of the test sample as an objective function of the vector characteristic to complete the recognition of the pavement crack image. The invention provides the rack identification method and system based on ABC-PCNN. An adaptive factoris added to improve the search position and probability selection of an artificial bee colony (ABC) algorithm; an improved ABC algorithm is used for optimizing the weight and the threshold value of the BP neural network, an improved ABC-BP hybrid neural network pavement crack recognition algorithm is established, and it is verified through experiments that the algorithm has good universality andeffectiveness.

Crack identification method and system based on ABC-PCNN

The invention provides a crack identification method and system based on an ABC-BP neural network. The method comprises the following steps: extracting a feature vector of a pavement crack image, andfurther comprising the following steps: training a test sample by using the optimized BP neural network; and adopting the cross entropy of the predicted value and the true value of the test sample asan objective function of the vector characteristic to complete the recognition of the pavement crack image. The invention provides a crack identification method and system based on an ABC-BP neural network. An adaptive factor is added to improve the search position and probability selection of an artificial bee colony (ABC) algorithm; an improved ABC algorithm is used for optimizing the weight andthe threshold value of the BP neural network, an improved ABC-BP hybrid neural network pavement crack recognition algorithm is established, and it is verified through experiments that the algorithm has good universality and effectiveness.

Li Shenzhen

Papers

Crack identification method and system based on ABC-PCNN

Crack identification method and system based on an ABC-BP neural network