The preservation of the environment has become a priority and a subject that is receiving more and more attention. This is particularly important in the field of precision agriculture, where pesticide and herbicide use has become more controlled. In this study, we propose to evaluate the ability of the deep learning (DL) and convolutional neural network (CNNs) technology to detect weeds in several types of crops using a perspective and proximity images to enable localized and ultra-localized herbicide spraying in the region of Beni Mellal in Morocco. We studied the detection of weeds through six recent CNN known for their speed and precision, namely, VGGNet (16 and 19), GoogLeNet (Inception V3 and V4) and MobileNet (V1 and V2). The first experiment was performed with the CNNs architectures from scratch and the second experiment with their pre-trained versions. The results showed that Inception V4 achieved the highest precision with a rate of 99.41% and 99.51% on the mixed image sets and for its version from scratch and its pre-trained version respectively, and that MobileNet V2 was the fastest and lightest with its size of 14 MB.

In each communication system an identification of the transmission channel between the transmitter and the receiver, it is necessary to identify the parameters of channel.Several methods exist, the most commonly used methods are learning by sending occasionally a known sequence between the transmitter and the receiver. In order to solve the problem of channels identification and save the resource of bandwidth, we use blind techniques which are a great interest to have the best compromise between a suitable bit rate and quality of the information retrieved.In this paper we describe tree blind algorithms witch are based on high order cumulant (HOC). In order to identify the impulse of two selective frequency fading channels called Broad Radio Access Network (BRAN A and BRAN E). Our contribution in this work is to make a comparative study between different algorithms of blind identification, compared with a supervised such as RLS algorithm. The simulation results are in a noisy environment with different SNR, demonstrate that the proposed algorithm is better to estimate blindly the impulse response of these channels (without any information about the input).

The modern telecommunication systems require very high transmission rates, in this context, the problem of channels identification is a challenge major. The use of blind techniques is a great interest to have the best compromise between a suitable bit rate and quality of the information retrieved.In this paper, we are interested to learn the algorithms for blind channel identification. We propose a hybrid method that performs a trade-off between two existing methods in order to improve the channel estimation.

In this paper we are focused on the Multi-Carrier Code Division Multiple Access (MC-CDMA) equalization problem. The equalization is performed using the Minimum Mean Square Error (MMSE) and Zero Forcing (ZF) equalizer based on the identified parameters representing the indoor scenario (European Telecommunications Standards Institute Broadband Radio Access Networks (ETSI BRAN A) channel model), and outdoor scenario (ETSI BRAN E channel model). These channels are normalized for fourth-generation mobile communication systems. However, for such high-speed data transmissions, the channel is severely frequencyselective due to the presence of many interfering paths with different time delays. The identification problem is performed using the Least Mean Squares (LMS) algorithm and the Takagi-Sugueno (TS) fuzzy system. The comparison between these techniques, for the channel identification, will be made for different Signal to Noise Ratios (SNR).

No abstract