Spatial filtering velocimetry (SFV) has the advantages of simple structure, good stability, and wide applications. However, the traditional linear CCD-based SFV method requires an accurate angle between the direction of linear CCD and the direction of moving object, so it is not suitable for measuring a complex flow field or two-dimensional speed in a granular media. In this paper, a new extension of spatial filtering method (SFM) based on high speed array CCD camera is proposed as simple and effective technique for measuring two-dimensional speed field of granular media. In particular, we analyzed the resolution and range of array CCD-based SFV so that the reader can clarify the application scene of this method. This method has a particular advantage for using orthogonal measurement to avoid the angle measurement, which were problematic when using linear CCD to measure the movement. Finally, the end-wall effects of the granular flow in rotating drum is studied with different experimental conditions by using this improved technique.

Aiming at the problem of defect image character recognizing on steel ball, the improved method of run length coding was used in registering bug area of steel ball surface to disjoin defect region and irrelevant region after image was segmented. As a result, steel ball was detected by area character parameter. The program of image processing, segmentation, recognizing was developed in LabVIEW platform. This means has already been tested and it shows the algorithm has practicability, reliability and precision.

The effect of image edge detection and segmentation for damaged area is poor during tool detection. Aiming at that, Robert and Sobel and Prewitt characteristic segmentation arithmetic operators are compared in this paper. The arithmetic improved of Canny operator is put forward by calculating amplitude value and direction of grads with two rank partial derivative's finite difference. Smoothing edge information is reduced. Software of segmentation and recognition is developed by using VC 6.0. The practical test indicates that tool damage characteristic states under different lights can be recognized by this arithmetic and its precision is up to 96%.