Photoacoustic tomography (PAT) is an emerging hybrid biomedical imaging technique based on photoacoustic effects with high spatial resolution and high contrast. In practical photoacoustic imaging, the data acquired by the detector is usually incomplete (limited detection view and sparse sampling) due to geometric limitations or accelerated data acquisition, which leads to blurred reconstructed images with artifacts. In order to solve the problem of limited data PAT, an optoacoustic image reconstruction algorithm based on gradient minimisation is studied. The algorithm first uses the gradient descent method to obtain an initial solution, then projects it onto the positive constraint set, and finally uses a quadratic smoothing regularization term to further improve the reconstruction quality. The simulation results show that the algorithm studied in this paper can effectively suppress the image noise and artifacts in the case of limited viewing angle and sparse sampling, and has a better reconstruction effect, and improves the peak signal-to-noise ratios (PSNRs) by about 8.09 dB, 5.12 dB and 13.84 dB, compared with those of the classical TR, ITR and FBP reconstruction algorithms, respectively.