Through the fast r-cnn target detection algorithm, the position of normal bucket teeth in the image can be obtained, and the predicted number of frames can be calculated to preliminarily judge whether there is a fault. Image segmentation and image processing technology will be combined to further estimate the position of the faulty bucket teeth and preliminarily judge the wear degree of the bucket teeth.
On the basis of fast r-cnn bucket tooth target detection, combined with the actual needs of bucket tooth fault detection, the problems of fault location and wear degree judgment in bucket tooth fault detection are further solved. Firstly, the bucket tooth image obtained by target detection is taken as the input, and the bucket tooth image is segmented combined with full convolution neural network. After the segmentation image is obtained, the contour is extracted, and the standardized Fourier descriptor is used to distinguish the bucket teeth at both ends and the bucket teeth in the middle, so as to help realize the calibration of bucket tooth sequence and provide the basis for detecting the position of faulty bucket teeth. At the same time, this chapter also designs the calculation method of bucket tooth length, and realizes the preliminary judgment of bucket tooth wear degree by comparing other normal bucket teeth.