Wall crack intelligent identification method based on image processing

Abstract

The invention discloses a wall crack intelligent identification method based on image processing. The method comprises the following steps of acquiring a wall image signal through a CCD camera; stretching the gray scale of a wall crack area; constructing a wall crack image scale space, performing scale transformation on the stretched wall crack image, and obtaining a representation sequence of thewall crack image multi-scale space; selecting the key points; counting the direction of the field gradient of the key point, and representing the main direction of the field gradient by a maximum value to form a feature vector; using a weak learning algorithm to train the wall crack image training set; obtaining t wall crack weak classifiers after T cycles, finally overlapping the T wall crack weak classifiers together according to the updated weight to obtain a wall crack strong classifier, and using the wall crack strong classifier for carrying out intelligent recognition on the wall cracks. According to the method, the image data redundancy is greatly reduced, the accuracy of the wall crack image recognition is remarkably improved, the wall image calculation redundancy is reduced, thecalculation efficiency is improved, the real-time recognition can be achieved, the calculation speed and the real-time performance are greatly enhanced, and the recognition efficiency is improved.

Description

technical field [0001] The invention relates to the technical field of image recognition, in particular to an image processing-based intelligent recognition method for wall cracks. Background technique [0002] my country's construction industry has achieved rapid development, but due to the late start of my country's construction safety mechanism, especially in natural environments such as long-term wind, rain, and sun, buildings will be affected by factors such as temperature stress, and wall materials will produce gaps. The gap may be catalyzed and enlarged to generate cracks on the surface. As the cracks continue to expand, it will cause safety hazards. Therefore, it is necessary to effectively detect cracks and evaluate their risks in order to prevent potential hazards. At present, the inspection of walls mainly depends on manual work. Patrol inspection, heavy workload, difficult to complete on time in harsh environment, and strong manual subjectivity, low reliability, lo...

AI Technical Summary

Problems solved by technology

[0002] my country's construction industry has achieved rapid development, but due to the late start of my country's construction safety mechanism, especially in natural environments such as long-term wind, rain, and sun, buildings will be affected by factors such as temperature stress, and wall materials will produce gaps. The gap may be catalyzed and enlarged to generate cracks on the surface. As the cracks continue to expand, it will cause safety hazards. Therefore, it is necessary to effectively detect cracks and evaluate their risks in order to prevent potential hazards. At present, the inspection of walls mainly depends on manual work. Patrol inspection, heavy workload, difficult to complete on time in harsh environment, and strong manual subjectivity, low reliability, low work efficiency, dangerous for high-rise buildings

[0003] Although the existing wall crack recognition part adopts image processing technology, and the existing wall crack recognition method is slow and inefficient. For example, the patent CN106651893A solves the problem of using image processing or neural network algorithm, but, In fact, the processing of image RGB three-channel data greatly increases the amount of calculation, which cannot reach the level of intelligent automatic recognition; moreover, it leads to more calculation time, and the steering method of this patent has high redundancy and low recognition accuracy; and In the prior art, the wall crack identification processing method involving image processing is relatively simple. For example, in the patent CN106203351A, the crack identification is only performed for horizontal and vertical projections, and the extraction of feature parameters is less, and the utilization rate of image information is low, so accurate identification cannot be realized.

Images