Grape quality inspection needs the identification of multiple categories of grapes, and there are many scenes such as depth of field changes and multiple strings in the grape images. Grape recognition is ineffective due to the limitations of single pretreatment method. The research objects were 15 kinds of natural scene grape images collected in the greenhouse, and the corresponding image dataset Vitis-15 was established. Aiming at the large intra-class differences and small inter-class of differences grape images, a multi-scale grape image recognition method based on Convolutional Neural Network (CNN) was proposed. Firstly, the data in Vitis-15 dataset were pre-processed by three methods, including the image rotating based data augmentation method, central cropping based multi-scale image method and data fusion method of the above two. Then, transfer learning method and convolution neural network method were adopted to realiize the classification and recognition. The Inception V3 network model pre-trained on ImageNet was selected for transfer learning, and three types of models-AlexNet, ResNet and Inception V3 were selected for convolution neural network. The multi-scale image data fusion classification model MS-EAlexNet was proposed, which was suitable for Vitis-15. Experimental results show that with the same learning rate on the same test dataset, compared with the augmentation and multi-scale image method, the data fusion method improves nearly 1% testing accuracy on MS-EAlexNet model with 99.92% accuracy, meanwhile the proposed method has higher efficiency in classifying small sample datasets.

Concerning the problem that the existing synthesis-analysis dictionary learning method can not effectively eliminate the differences between the samples of the same class and ignore the different effects of different features on the classification, an image classification method based on Multi-view Feature Projection and Synthesis-analysis Dictionary Learning (MFPSDL) was put forward. Firstly, different feature projection matrices were learned for different features in the process of synthesis-analysis dictionary learning, so the influence of the within-class differences on recognition was reduced. Secondly, discriminant constraint was added to the synthesis-analysis dictionary, so that similar sparse representation coefficients were obtained for samples of the same class. Finally, by learning different weights for different features, multiple features could be fully integrated. Several experiments were carried out on the Labeled Faces in the Wild (LFW) and Modified National Institute of Standards and Technology (MNIST) database, the training time of MFPSDL method on LFW and MNIST databases were 61.236 s and 52.281 s. Compared with Fisher Discrimination Dictionary Learning (FDDL), Lable Consistent K Singular Value Decomposition (LC- KSVD) and Dictionary Pair Learning (DPL), the recognition rate of MFPSDL method on LFW and MNIST was increased by at least 2.15 and 2.08 percentage points. The experimental results show that MFPSDL method can obtain higher recognition rate while keeping low time complexity, and it is suitable for image classification.

Dynamic network topology is one of the important characteristics of vehicle Ad Hoc networks (VANET). Based on Intelligent Driver Model with Lane Changes (IDM-LC), the VanetMobiSim was used to study the dynamic centrality of topology for VANET in detail. The temporal network model of VANET was built. The evaluation method of dynamic centrality based on the attenuation factor and information store-and-forward index was established, which not only can describe the relation between the current network topology and the historical one, but also can depict the store-and-forward mechanism of information transmission in VANET. Finally, the dynamic centrality of VANET was analyzed through the simulation experiment. The results show that although the dynamic centrality of VANET topology varies with time and parameters, the ranking of important nodes remains relatively stable. This conclusion not only can help us identify the relay nodes of information transmission better to achieve the successful delivery of information, but also provides guidance for invulnerability of VANET topology.

Concerning the problem of invulnerability in Vehicular Ad Hoc NETwork (VANET), the invulnerability characteristics for VANET under random attacks and intentional attacks were analyzed. Firstly, the largest connected component, average size of components, critical point removal ratio as well as network efficiency were proposed to be used for the invulnerability evaluation metrics for VANET. Then, based on intelligent driver model with lane changes, the VANET was established through VanetMobisim software. Finally, the influence of the number of nodes, transmission ranges and patterns of attack on VANET invulnerability were given. The experimental results show that the VANET has a strong invulnerability faced with random attacks while its invulnerability to intentional attacks is fairly low as a result of the uneven degree distribution of vehicles; the intentional attacks based on node betweenness destroy the networks more quickly and strongly. The derived rules can provide the optimization of VANET topology control, protocol development and network management with new guidance.

Aiming at the problem of connectivity in Vehicular Ad Hoc Network (VANET), the evolution characteristics of connectivity characteristics for VANET were analyzed. Firstly, the number of connected components, connectivity probability and connectivity length were proposed to be used for the evaluation connectivity metrics for VANET. Then, based on Intelligent Driver Model with Lane Changes (IDM-LC), the VANET was set up through VanetMobiSim software. Finally, the relation of node communication radius and the average number of connected components, average connectivity probability and average connectivity length were given. At the same time, the statistical distribution of the number of connected components was also analyzed. The results show that number of connected components follows a normal distribution by using Q-Q plot and T-test. Moreover, the results also show that the statistical distribution of the number of connected components is independent of the node communication radius.

In view of the requirement of convenience and fast speed of automobile ECU online upgrading, by researching the CAN bus communication as well as the IAP technology, the online upgrading method is designed based on software trigger. This method achieves the rapid online upgrading of ECU in automobile network by sending the instruction through online upgrading software to communicate with CAN bus. Thus it solves the operation inflexibility brought by the hardware trigger during the online upgrading. The upgrading system uses STM8AF51AA micro controller as the platform, and is implemented and applied in automotive BCM controller, which verifies the feasibility and reliability of this technology.

With the rapid development of automotive electronic market, more and more Electronic Control Units (ECU) for vehicle controller appear and the functional test also becomes more complex. In order to solve the problem of ECU functional test, the ECU's automatic test method based on Controller Area Network (CAN) was studied. The system included the software and hardware platform of National Instrument (NI) and communication platform of CAN bus, by which the system and ECU formed a closed-loop structure. To transmit the test message through CAN bus, the system could achieve batch test of ECUs with the same type. By using the new test method, the system can reduce the test errors, and support assembly line test of ECU, which greatly reduces the complexity of ECU functional test and test work. At the same time, the system can also apply to other types of ECU functional test by improving the generation module of simulated signal and use case library.

Current three-dimensional reconstruction algorithms of the computer vision field have limitations that they need to deploy and calibrate the cameras around the scene, or they need a structure light. Furthermore, these algorithms are not robust enough to every object. A new kind of four camera array reconstruction algorithms which properly combined the image registration algorithm and the camera array method was proposed to solve the robustness and limitation problems. It does not need calibration or structure light support. The experiments based on complex indoor sense with shadows demonstrate that this method is able to do dense point cloud reconstruction robustly and can overcome the shortcomings of current reconstruction algorithms.