In this paper, a quality detection method for battery FPC (Flexible Printed Circuit) connectors based on active shape model template matching is proposed. It can deal with different kinds of connector appearance defects. Firstly, construct template data set of connector, acquire test images and apply cutting operation to original image, then execute tilt correction and image reconstruction by means of least square method and affine transformation to fulfil the pre-processing stage. Then, match and locate connector region in per-processing image with the help of the active shape model (ASM) based template matching method. To deal with different kinds of defect (soldering offset/tilt, exposed copper clad layer in FPC, broken edge in FPC, defects in center area of connector, defects on metal and plastic components), independent detection algorithm units are integrated in the system. Template can also be real-timely updated according to detection result. Finally, the defects will be classified, located and marked in detection image. In addition, aimed at the need of battery industry, a set of detection system with low cost, high performance and strong stability has been designed. It can be concluded from online and offline experiments that the proposed method is of high detection rate, good real-time performance and strong robustness.
For solving the problem of programming measurement path when inspecting Blade Profile, and improving the efficiency and precision, the self-adaptive dynamic path planning model of blade profile measurement is proposed, using Back Propagation Neutral Network, based on the blade measuring characteristic and non-contact measurement system of blade profile detecting in laser triangular principle. For the feature of blade profile measuring, with the factors affecting the Probe precision and efficiency (the range of depth of field, incident angle), we plan the probe position of next measurement point by selecting 3 layer BP networks of , using practically measured blade profile as Training Sample, and regarding probe coordinates of corresponding profile measuring point as networks input. This paper discusses and explains the factors affecting measurement path planning, the creating and training of the BP networks profile measurement path planning in details. Because of the use of Neural Network learning the ability of approximating nonlinear mapping in any precision and the application of regarding blade profile data as BP network training sample, measuring movement path can combine with the blade curvature variation closely. Then, practically measuring precision and efficiency are improved, and the Path Planning problem is solved, brought by curvature varying greatly of blade profile in large blade profile measurement. At last, a group of experimental data is given, and the results of experiment are analyzed in detail.